JP2015152417A - Object recognition device, object recognition method, and object recognition program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce erroneous recognition of a shape of an object when recognizing the object on the basis of an image obtained by imaging a projected pattern.SOLUTION: An object recognition device includes: a recording unit which records distances to a plurality of light spots which are respectively obtained from a plurality of images obtained by imaging a pattern which is projected on an area including an object and includes the plurality of light spots, at time intervals for respective light spots; a first calculation unit which calculates variances of the plurality of distances recorded by the recording unit, for respective light spots included in the pattern; an evaluation unit which evaluates degrees of reliability of distances recorded for respective light spots on the basis of the variances calculated by the first calculation unit; and a correction unit which corrects distances recorded for a first light spot, of which the degrees of reliability are determined to be equal to or lower than a first threshold by the evaluation unit, by using distances recorded for other light spots adjacent to the first light spot.

Description

  The present invention relates to an object recognition device, an object recognition method, and an object recognition program.

  For example, a pattern including a plurality of light spots arranged in a grid pattern is projected onto a predetermined area, and light spots are projected based on the deviation of each light spot from the reference position in an image obtained by photographing the projected pattern. There is a technique for recognizing the shape of an object in a specified area (for example, see Patent Document 1).

  In this kind of technology, the object on which the pattern is projected is obtained by obtaining the distance between the light source used for projecting the pattern and the light spot projected on the object from the magnitude of deviation from the reference position of each light spot. Recognize the shape of objects such as surface irregularities.

JP 2004-96457 A

  By the way, when the light source used for pattern projection and the object are separated from each other, the brightness of each light spot in the image obtained by photographing the projected pattern is lowered. For this reason, it may be difficult to discriminate between a light spot included in the pattern and noise other than the light spot in an image obtained by photographing the projected pattern. That is, with this type of technology, noise other than the light spot included in the image may be erroneously detected as a light spot. Then, when recognizing the shape of the object on which the pattern is projected, if the distance obtained from the erroneously detected light spot is used, the shape of the object may be erroneously recognized.

  An object recognition device, an object recognition method, and an object recognition program of the present disclosure are intended to reduce erroneous recognition of an object shape in object recognition based on an image obtained by photographing a projected pattern.

  According to one aspect, the object recognizing device obtains a plurality of light spots obtained from each of a plurality of images obtained by photographing a pattern including a plurality of light spots projected on a region including an object at time intervals. A recording unit that records the distance to each corresponding to each light spot, a first calculation unit that calculates a variance of a plurality of distances recorded in the recording unit for each light spot included in the pattern, and a first calculation An evaluation unit that evaluates the reliability of the distance recorded corresponding to each light spot based on the variance calculated by the unit, and the first light whose reliability is determined to be less than or equal to the first threshold by the evaluation unit And a correction unit that corrects the distance recorded corresponding to the point using the distance recorded corresponding to another light spot adjacent to the first light spot.

  According to another aspect, the object recognition method includes a plurality of light spots obtained from each of a plurality of images obtained by capturing a pattern including a plurality of light spots projected on a region including an object at time intervals. The distance to each is recorded corresponding to each light spot, the variance of a plurality of distances recorded for each light spot included in the pattern is calculated, and based on the calculated dispersion, the distance corresponding to each light spot is calculated. The reliability of the recorded distance is evaluated, and the recorded distance corresponding to the first light spot whose reliability is equal to or lower than the first threshold is associated with another light spot adjacent to the first light spot. Then, the correction is made using the recorded distance.

  According to another aspect, the object recognition program stores a plurality of light spots obtained from each of a plurality of images obtained by capturing a pattern including a plurality of light spots projected on a region including an object at time intervals. The distance to each is recorded corresponding to each light spot, the variance of a plurality of distances recorded for each light spot included in the pattern is calculated, and based on the calculated dispersion, the distance corresponding to each light spot is calculated. The reliability of the recorded distance is evaluated, and the recorded distance corresponding to the first light spot whose reliability is equal to or lower than the first threshold is associated with another light spot adjacent to the first light spot. Then, the computer executes the process of correcting using the recorded distance.

  The object recognition device, the object recognition method, and the object recognition program disclosed herein can reduce erroneous recognition of the shape of an object in object recognition based on an image obtained by photographing a projected pattern.

It is a figure which shows one Embodiment of an object recognition apparatus. It is a figure which shows the example of the image which image | photographed the pattern shown in FIG. It is a figure which shows the example of the distance obtained by the measuring apparatus shown in FIG. It is a figure which shows the example of the reliability calculated | required by the evaluation part shown in FIG. It is a figure which shows the example of the distance correct | amended by the correction | amendment part shown in FIG. It is a figure which shows operation | movement of the object recognition apparatus shown in FIG. It is a figure which shows another embodiment of an object recognition apparatus. It is a figure which shows the example of the distance obtained by the measuring apparatus shown in FIG. It is a figure which shows another embodiment of an object recognition apparatus. It is a figure which shows another embodiment of an object recognition apparatus. It is a figure which shows the example of the characteristic for every kind of object. It is a figure which shows an example of the holding | maintenance part shown in FIG. It is a figure which shows the hardware structural example of the object recognition apparatus shown in FIG. It is a figure which shows operation | movement of the object recognition apparatus shown in FIG.

  Hereinafter, embodiments will be described with reference to the drawings.

  FIG. 1 shows an embodiment of an object recognition apparatus. The object recognition apparatus 10 illustrated in FIG. 1 includes a recording unit 11, a first calculation unit 12, an evaluation unit 13, a correction unit 14, and a recognition unit 15. The object recognition device 10 is connected to a measurement device ME that receives reflected light from the pattern PT projected from the projection device PR onto the object TG. In FIG. 1, for ease of explanation, the case where the pattern PT is projected onto an object TG having a flat surface is shown. However, the shape of the object TG is not limited to the shape shown in FIG. The surface may be uneven.

  Prior to the description of the function and operation of each unit included in the object recognition device 10, the measurement device ME and the projection device PR will be described.

  The projection device PR projects a pattern PT including a plurality of light spots onto a region including the object TG. The pattern PT projected by the projection device PR includes, for example, a plurality of light spots arranged in a grid pattern at predetermined intervals when projected onto a plane PL that is separated from the projection device PR by a distance D0. The projection apparatus PR, for example, diffracts a light beam emitted from a light emitting element such as a semiconductor laser included in the projection apparatus PR by a diffraction grating included in the projection apparatus PR, thereby grating a plurality of light spots. The pattern PT arranged in a shape is projected. Note that the plane PL shown in FIG. 1 is shown for explanation, and the plane PL is not included in the region where the pattern PT is projected.

  The measuring device ME includes an imaging unit CAM such as a camera and an image processing unit IMP. The imaging unit CAM, for example, repeatedly captures the pattern PT projected on the region including the object TG by the projection device PR at predetermined time intervals, thereby including a plurality of light spots including a plurality of light spots projected on the surface of the object TG. An image IMG is acquired. The imaging unit CAM has sensitivity to a wavelength band including the wavelength of light emitted from the light emitting element included in the projection apparatus PR. The image IMG taken by the imaging unit CAM is transferred to the image processing unit IMP.

  The imaging unit CAM included in the measuring apparatus ME captures the pattern PT projected on the object TG from the direction Drs inclined with respect to the direction Drp in which the projection apparatus PR projects the pattern PT. For this reason, in the image IMG taken by the imaging unit CAM, the position of each light spot included in the pattern PT projected on the surface of the object TG depends on the distance between the projection device PR, which is a light source, and each light spot. , It will deviate from the reference position. Here, the reference position of each light spot is, for example, the position of each light spot in the image IMG obtained when the pattern PT projected on the plane PL is photographed by the imaging unit CAM.

  The image processing unit IMP included in the measuring apparatus ME shifts between the position of each light spot in the image IMG obtained by capturing the pattern PT projected on the surface of the object TG and the reference position corresponding to each light spot. Is detected. Then, the image processing unit IMP obtains a distance from the projection device PR to each light spot projected on the surface of the object TG based on the detected deviation corresponding to each light spot. For example, when obtaining the distance D1 from the projection device PR to the light spot Sp1 projected on the surface of the object TG, the image processing unit IMP performs light corresponding to the light spot Sp1 when the pattern PT is projected onto the plane PL. The position of the point Sp0 in the image IMG is used as the reference position. The image processing unit IMP obtains the height h1 of the light spot Sp1 projected on the surface of the object TG from the plane PL based on the deviation of the position of the light spot Sp1 included in the image IMG from the reference position. Then, the image processing unit IMP obtains the distance D1 from the obtained height h1 and the distance D0 between the projection device PR and the plane PL.

  Each time the image processing unit IMP receives an image IMG taken by the imaging unit CAM, the image processing unit IMP obtains a distance between each light spot included in the received image IMG and the projection device PR, and sequentially displays information indicating the obtained distance. To the object recognition device 10.

  The recording unit 11 included in the object recognition apparatus 10 sequentially records the distance between each light spot indicated by the information received from the measurement apparatus ME and the projection apparatus PR, for example, corresponding to each light spot. For example, when the imaging unit CAM of the measuring device ME has captured the pattern PT projected onto the object TG n times (n is an integer of 2 or more), n distances corresponding to each light spot are set. The numerical value shown is recorded in the recording unit 11.

  For example, the first calculation unit 12 uses a numerical value indicating n distances recorded in the recording unit 11 corresponding to each light spot by using the formula (1), and uses the numerical value indicating the distance for each light spot during a predetermined time. The distance variance V obtained by the measuring device ME is calculated. Here, the predetermined time is a time elapsed while n images are captured by the imaging unit CAM included in the measurement apparatus ME. In Expression (1), a numerical value j is a positive integer equal to or smaller than n and indicates the imaging order of the image IMG, and a variable D (j) is the jth imaging of a certain light spot included in the pattern PT. The distance obtained by the measuring device ME from the measured image IMG is shown. The numerical value Dav indicates an average of numerical values indicating n distances recorded in the recording unit 11 corresponding to the same light spot.

  Note that the dispersion of the distances obtained for each light spot by the first calculation unit 12 is not limited to the statistical dispersion represented by the equation (1), but n pieces recorded in the recording unit 11 corresponding to each light spot. Any value may be used as long as it indicates the magnitude of the variation of the numerical value. For example, the first calculation unit 12 considers that the distance to each light spot changes when the object TG moves during a predetermined time, and calculates the distance to each light spot obtained for each image IMG. You may obtain | require the average value of variation | change_quantity as dispersion | distribution of the distance of each light spot.

  The dispersion of the distance calculated for each light spot by the first calculation unit 12 described above is passed to the evaluation unit 13.

  The evaluation unit 13 evaluates the reliability of the distance recorded in the recording unit 11 corresponding to each light spot based on the dispersion of the distance received from the first calculation unit 12. The evaluation unit 13 uses a function F (V (V) that gives a value closer to the numerical value 1 that is the maximum value of reliability as the value V indicating the variance calculated by the first calculation unit 12 is smaller, for example, using Equation (2). ) To calculate the reliability r. Hereinafter, a case will be described in which the reliability obtained by the evaluation unit 13 is a real number with a numerical value of 0 or more and a value of 1 or less, and the value becomes closer to the maximum value of 1 as the reliability increases.

The function used for obtaining the reliability in the evaluation unit 13 is not limited to the function F (V) shown in the equation (2), and is a function that gives a reliability having a larger value as the value V indicating the variance is smaller. I just need it.

  The correction unit 14 determines the distance recorded in the recording unit 11 corresponding to the first light spot, which is the light spot whose reliability is determined to be equal to or less than the predetermined first threshold, by the evaluation unit 13. And a distance recorded corresponding to another light spot having a predetermined positional relationship. For example, the correction unit 14 compares the reliability obtained by the evaluation unit 13 corresponding to each light spot with a predetermined first threshold value, and thereby selects the reliability from a plurality of light spots included in the pattern PT. A first light spot having a degree equal to or less than a predetermined first threshold is detected. Then, the correction unit 14 records the distance recorded in the recording unit 11 corresponding to the detected first light spot, and the distance recorded in correspondence with the plurality of light spots adjacent to the first light point in the pattern PT. The distance corresponding to the detected first light spot is corrected by replacing with the average value of. Here, it is desirable that the correction unit 14 sets a small value of, for example, about 0.02 as the first threshold value used for detection of the first light spot whose reliability of the distance measured by the measuring device ME is low. The correction unit 14 may obtain an average value of the reliability obtained for each light spot by the evaluation unit 13, and may use a value about half of the obtained average value as the first threshold value.

  The recognition unit 15 reads the distance corresponding to each light spot including the first light spot corrected by the correction unit 14 from the recording unit 11, and based on the read distance, the shape of the object TG on which the pattern PT is projected. recognize. For example, the recognizing unit 15 obtains the position in the three-dimensional space of each light spot projected on the surface of the object TG based on the distance read corresponding to each light spot, and interpolates the obtained positions with each other. Then, a curved surface indicating the shape of the surface of the object TG is obtained. The recognition unit 15 may detect the movement of the object TG based on the obtained time change of the curved surface.

  Note that the measurement device ME may be included in the object recognition device 10, or the object recognition device 10 may be included in the measurement device ME. Furthermore, the measurement device ME, the projection device PR, and the object recognition device 10 may be included in a single device.

  FIG. 2 shows an example of an image IMG obtained by photographing the pattern PT shown in FIG. The image IMG illustrated in FIG. 2 is an example of one of the n images IMG captured by the imaging unit CAM illustrated in FIG.

  In the image IMG illustrated in FIG. 2, a rectangular region TG indicates the outer shape of the object TG illustrated in FIG. 1. In FIG. 2, each of white circles S1, S2, S3, S4, S5, S6, S7, S8, and S9 represents a light spot included in the pattern PT projected on the surface of the object TG. In the example of FIG. 2, the light spots other than the light spots S1 to S9 among the light spots included in the pattern PT are not shown. In the example of FIG. 2, the rectangle indicating the outline of the object TG is for explaining the positional relationship between the object TG and the light spots S1 to S9, and is not included in the image IMG.

  In FIG. 2, a rectangular region R5 indicated by a broken line indicates a range in which the light spot S5 is detected from the image IMG in the image processing unit IMP shown in FIG. Here, the range used for detecting the light spot S5 from the image IMG includes, for example, the position in the image IMG when the light spot S5 is projected onto the plane PL shown in FIG. A range having a width W1 smaller than the interval W0 is set. Although not shown in FIG. 2, the image processing unit IMP shown in FIG. 1 corresponds to all the light spots included in the pattern PT, and the region R5 shown in correspondence with the light spot S5 in FIG. A similar range is set in advance, and the corresponding light spot is detected inside the set range. Note that the rectangle indicating the region R5 illustrated in FIG. 2 is for indicating the range in which the light spot S5 is detected by the image processing unit IMP illustrated in FIG. 1, and is not included in the image IMG.

  Here, consider the case where noise N5 other than the light spot S5 appears inside the region R5 shown in FIG. The noise N5 appears due to, for example, a pattern on the surface of the object TG or light from a light source other than the projection apparatus PR shown in FIG.

  For example, when the image processing unit IMP shown in FIG. 1 detects a circle having a luminance of a predetermined value or more within a correspondingly set range as each light spot included in the pattern PT, it is shown in FIG. There is a possibility that the noise N5 is detected as a light spot in the region R5 instead of the light spot S5. When there is fluctuation in the brightness of the light spot S5 and the brightness of the noise N5 between the n images, the image processing unit IMP shown in FIG. 1 performs the light spot in the region R5 in each of the n images IMG. As a result, either the light spot S5 or the noise N5 may be detected at random. In such a case, for example, the distance calculated based on the position of the light spot detected from the region R5 in each of the n images IMG is the noise in the corresponding region as shown in FIG. It will vary greatly compared to the case where is not included.

  FIG. 3 shows an example of the distance obtained by the measuring apparatus ME shown in FIG. In FIG. 3, the horizontal axis t represents time, and the vertical axis D represents the distance from the projection device PR shown in FIG. 1 to the light spot projected onto the object TG. In the example of FIG. 3, a case where an image IMG is acquired by photographing by the imaging unit CAM shown in FIG. 1 at each of times T1, T2, T3, T4, T5, T6, T7, T8, T9, and T10 will be described. To do. Note that the number of images IMG acquired by imaging by the imaging unit CAM is not limited to 10 which is the number of times shown in FIG. 3, and may be less than 10 or more than 10 as long as it is 2 or more.

  In FIG. 3, the white circles shown corresponding to the times T1 to T10 are detected from the region R5 shown in FIG. 2 by the image processing unit IMP shown in FIG. 1 from the image IMG taken at each time. The example of the distance calculated based on the position of a light spot is shown. On the other hand, the black circles shown corresponding to the times T1 to T10 are taken at each time when the noise corresponding to the light spot S4 shown in FIG. 2 does not contain noise such as the noise N5, for example. An example of the distance calculated based on the position of the light spot S4 detected from the obtained image IMG is shown.

  In the example of FIG. 3, the distance indicated by the white circle varies in the range of the width Dv5, while the range of variation in the distance indicated by the black circle is a width Dvc smaller than the width Dv5. The reason why the variation in the distance indicated by the white circle is larger than the variation in the distance indicated by the black circle is that the light spot in the region R5 illustrated in FIG. Is, for example, randomly detected. In the example of FIG. 3, the light spot S5, which is the original light spot, is detected from the region R5 of the image IMG photographed at times T1, T3, T4, T7, and T8, and images photographed at other times. A case where noise N5 is detected instead of the light spot S5 from the region R5 of the IMG is shown.

  That is, the greater the variance of the distance calculated based on the position of a certain light spot in the n images IMG taken during a predetermined time, the more the light spot and noise are detected from each image IMG. It is likely that For example, when the noise N5 is erroneously detected as the light spot S5 included in the pattern PT, the distance obtained based on the position of the detected noise N5 is the distance to the light spot S5 projected on the object TG. Is different. Therefore, the average value of the distances obtained for the light spot S5 indicated by the white circle in FIG. 2 is a value that does not indicate the distance between the light spot S5 projected on the object TG and the projection device PR, and the distance measurement result The reliability is reduced as compared with the case where there is no noise N5. That is, the reliability obtained by the evaluation unit 13 shown in FIG. 1 using the formula (1) or the like is shown in FIG. 2 in the position of each light spot used for calculating the distance by the measuring device ME. The low possibility that noise like the shown noise N5 is included is shown.

  FIG. 4 shows an example of the reliability obtained by the evaluation unit 13 shown in FIG. FIG. 4A shows a correspondence relationship between the distance obtained by the measuring apparatus ME shown in FIG. 1 and the reliability obtained by the evaluation unit 13 for the light spots S1 to S9 shown in FIG. 4B recognizes the shape of the object TG onto which the pattern PT is projected based on the distances obtained for the light spots S1 to S9 shown in FIG. 2 by the measuring device ME shown in FIG. An example of the result is shown.

  In the example of FIG. 4A, the value [1500] shown corresponding to the light spot S1 indicates the distance obtained by the measuring device ME for the light spot S1, and is also shown corresponding to the light spot S1. The value [0.10] indicates the reliability obtained by the evaluation unit 13. Similarly, in the table of FIG. 4A, the distance [1510], the distance [1490], the distance [1480], and the distance [2860] are measured by the measuring device ME for each of the light spots S2, S3, S4, and S5. It shows that it was obtained. In the table of FIG. 4A, the distance [1490], the distance [1500], the distance [1480], and the distance [1510] are obtained for each of the light spots S6, S7, S8, and S9 by the measuring device ME. Indicates that 4A shows that the reliability [0.11], the reliability [0.12], the reliability [0...] Are obtained by the evaluation unit 13 for each of the light spots S2, S3, S4, and S5. 08], indicating that the reliability [0.01] was obtained. Similarly, the table of FIG. 4A shows the reliability [0.10], reliability [0.11], reliability [0] by the evaluation unit 13 for each of the light spots S6, S7, S8, and S9. .09], indicating that the reliability [0.11] was obtained.

  In FIG. 4A, if the distances and the reliability indicated for the respective light spots S1 to S9 are compared with each other, the distance and the reliability obtained for each light spot other than the light spot S5 are almost equal. It can be seen that the values are equivalent. On the other hand, the distance [2860] shown corresponding to the light spot S5 is a value close to twice the average value of the distances corresponding to the other light spots. On the other hand, the reliability [0.01] indicated corresponding to the light spot S5 is less than half of the average value of reliability corresponding to the other light spots (0.09 in the example of FIG. 4A). Value. That is, the light spot S5 shown in FIG. 2 has a reliability set to about half of the average value of the reliability obtained for each adjacent light spot by the evaluation unit 13 shown in FIG. It is an example of the 1st light spot considered to be below a threshold value.

  Then, like the distance corresponding to the light spot S5 shown in FIG. 4A, the distance with the reliability lower than the first threshold is not distinguished from the distance obtained for the other light spots. When used for shape recognition, the shape of the object TG may be recognized as a shape different from the actual shape.

  For example, in FIG. 4B, a distance close to twice the distance obtained corresponding to the other light spots is obtained for the light spot S5, so that the recognition unit 15 shown in FIG. Is shown as an example of a shape Rb having a dent at the center. In FIG. 4B, the recess in the center is shown in white.

  FIG. 5 shows an example of the distance corrected by the correction unit 14 shown in FIG. 5A shows the distances obtained by the measuring apparatus ME for the light spot S5 shown in FIG. 2 by the correction unit 14 shown in FIG. 1, and the light spots S1 to S4 and the light spot adjacent to the light spot S5. The example corrected based on the distance obtained about S6-S9 is shown. FIG. 5B shows the result of recognizing the shape of the object TG onto which the pattern PT is projected by the recognition unit 15 shown in FIG. 1 using the corrected distance shown in FIG. An example is shown.

  In the table shown in FIG. 5 (A), the distance shown corresponding to each light spot other than the light spot S5 is equivalent to the distance shown in FIG. 4 (A). In the table of FIG. 5A, the distance [1495] shown corresponding to the light spot S5 corresponds to each light spot adjacent to the light spot S5 by the correction unit 14 shown in FIG. The corrected distance calculated from the obtained distance is shown. Here, the light spots adjacent to the light spot S5 are the light spots S1 to S4 and the light spots S6 to S9 shown in FIG. 2. In the example of FIG. As the distance after correction, the average value of the distances corresponding to these light spots is shown.

  The corrected distance corresponding to the light spot S5 obtained by the correction unit 14 described above is compared with the value obtained by averaging the distances obtained for the light spots S1 to S9 without considering the reliability. This is a certain value. This is because the corrected distance corresponding to the light spot S5 obtained by the correcting unit 14 is not affected by the low reliability distance obtained corresponding to the light spot S5. In other words, by performing the correction by the correction unit 14 illustrated in FIG. 1, the recognition unit 15 is more sure about each light spot included in the pattern PT than in the case where simple smoothing is performed without considering the reliability. An object can be recognized based on the measurement result of a probable distance.

  Note that the information used for correcting the distance obtained for the light spot whose reliability is evaluated to be equal to or lower than the predetermined threshold is not limited to the example of FIG. 4, and low reliability is obtained in the pattern PT shown in FIG. Any information can be used as long as it is obtained from a light spot arranged closer to the light spot than other light spots. For example, the correction unit 14 uses the distances obtained for the light spots S2, S4, S6, and S8 adjacent to the light spot S5 to calculate a correction value for replacing the distance obtained corresponding to the light spot S5. May be. More preferably, the correction unit 14 corresponds to a light spot having a reliability equal to or higher than a predetermined threshold among light spots arranged close to a light spot having a low reliability. It is desirable to selectively use the obtained distance to obtain the corrected distance corresponding to the light spot that is determined to have low reliability.

  The distance corresponding to the corrected light spot S5 as shown in FIG. 5A is used for the recognition of the shape of the object TG by the recognition unit 15 shown in FIG. 1 together with the distance obtained for the other light spots. It is done. In this case, as shown in FIG. 5B, the recognition unit 15 can recognize the shape of the object TG as a shape having no dent in the center.

  FIG. 6 shows the operation of the object recognition apparatus 10 shown in FIG. The processing of step S301 to step S304 shown in FIG. 6 shows the operation of the object recognition apparatus 10 shown in FIG. 1, and shows an example of the object recognition method and the object recognition program. For example, the process illustrated in FIG. 6 is realized by a processor mounted on the object recognition apparatus 10 executing an object recognition program. Note that the processing illustrated in FIG. 6 may be executed by hardware mounted on the object recognition apparatus 10.

  In step S301, the recording unit 11 illustrated in FIG. 1 records each of the distances measured a plurality of times during a predetermined time by the measuring device ME for each light spot included in the pattern PT.

  In step S302, the first calculation unit 12 illustrated in FIG. 1 disperses a plurality of distances recorded in the recording unit 11 corresponding to each light spot during a predetermined time for each light spot included in the pattern PT. Is calculated.

  In step S303, the evaluation unit 13 illustrated in FIG. 1 evaluates the reliability of the distance recorded in the recording unit 11 corresponding to each light spot based on the variance calculated by the first calculation unit 12.

  In step S304, the correction unit 14 illustrated in FIG. 1 uses the evaluation unit 13 to calculate the distance recorded in the recording unit 11 corresponding to the first light spot whose reliability is equal to or less than the first threshold value. Correction is performed using the distance recorded in the recording unit 11 corresponding to another light spot adjacent to one light spot.

  Then, after the processing of step S304 is completed, the recognition unit 15 illustrated in FIG. 1 includes an area within the region where the pattern PT is projected based on the corrected distance recorded in the recording unit 11 corresponding to each light spot. The process of recognizing the shape of the object TG is performed.

  As described above, the object recognizing device 10 shown in FIG. 1 determines each light spot from the dispersion of distances obtained by a plurality of measurements by the measuring device ME for each light spot projected on the surface of the object TG. The reliability of the distance measured corresponding to is evaluated. Then, the object recognition device 10 records the distance recorded corresponding to the first light spot whose reliability is equal to or less than the first threshold value, corresponding to another light spot adjacent to the first light spot. Correct using the measured distance. Thereby, the recognition unit 15 can recognize the shape of the object TG using a probable value as the distance to each light spot projected on the surface of the object TG, so that the shape of the object TG is erroneously recognized. The probability of being reduced is reduced compared to not considering the reliability of the measured distance.

  FIG. 7 shows another embodiment of the object recognition apparatus. 7 that are equivalent to the components shown in FIG. 1 are denoted by the same reference numerals and description of the components may be omitted.

  The object recognition apparatus 10a illustrated in FIG. 7 includes a first calculation unit 12a and an evaluation unit 13a, as well as the recording unit 11, the correction unit 14, and the recognition unit 15 illustrated in FIG. A measuring device ME is included. The first calculation unit 12 a is a component corresponding to the first calculation unit 12 illustrated in FIG. 1, and includes a variance calculation unit 121, an average calculation unit 122, and a difference calculation unit 123. The evaluation unit 13a is a component corresponding to the evaluation unit 13 illustrated in FIG. 1 and includes a reliability calculation unit 131, a detection unit 132, and an adjustment unit 133.

  The variance calculation unit 121 included in the first calculation unit 12a replaces the equation (1) with the calculation of the variance V of n distances recorded in the recording unit 11 corresponding to each light spot included in the pattern PT. For example, equation (3) is used. As a result, the dispersion calculating unit 121 calculates the sum of absolute values of differences between each of the n distances recorded in the recording unit 11 corresponding to each light spot and the distance obtained in the previous measurement. The dispersion V of the distance recorded corresponding to each light spot is obtained. In Equation (3), the numerical value j is a positive integer less than or equal to n and indicates the order of measurement in n measurements, and the variable D (j) is a certain light spot included in the pattern PT. , Indicates the distance obtained in the j th measurement.

  Moreover, the average calculation part 122 calculates the average value Av of n distances recorded on the recording part 11 corresponding to each light spot for each light spot included in the pattern PT. For example, the average calculation unit 122 calculates the average of the distances corresponding to each of the light spots S1 to S9 from the n distances recorded in the recording unit 11 corresponding to each of the light spots S1 to S9 illustrated in FIG. The value Av is calculated.

  Then, the difference calculation unit 123 calculates the average distance Av calculated for the light spots sequentially selected from the light spots included in the pattern PT and the distance calculated for the other light spots adjacent to the selected light spot. A difference d from a value obtained by further averaging the average value Av is calculated. For example, the difference calculation unit 123 calculates the average distance Av calculated for the light spot S5 illustrated in FIG. 2 and the average value of eight distances calculated for each of the eight light spots adjacent to the light spot S5. The difference d from the value obtained by further averaging Av is calculated.

  That is, the average calculation unit 122 and the difference calculation unit 123 recorded the average value of the distance recorded in the recording unit 11 corresponding to each light spot and the plurality of light spots adjacent to the light spot. It is an example of the 2nd calculation part which calculates | requires the difference d with the average value of distance.

  The distance variance V calculated by the variance calculation unit 121 corresponding to each light spot is passed to the reliability calculation unit 131 of the evaluation unit 13a. On the other hand, the difference d calculated corresponding to each light spot by the difference calculation unit 123 is passed to the detection unit 132 of the evaluation unit 13a. FIG. 8 shows an example of the distance variance V calculated by the variance calculator 121, the average distance Av calculated by the average calculator 122, and the difference d calculated by the variance calculator 121. Will be described later.

  The reliability calculation unit 131 included in the evaluation unit 13a receives the dispersion V of the distance to each light spot from the dispersion calculation unit 121, and the recording unit 11 corresponds to each light point based on the received dispersion V of the distance. The reliability r of the distance recorded in is calculated. The reliability calculation unit 131 calculates the reliability r of the distance recorded corresponding to each light spot, for example, using the function F (V) of the distance variance V shown in Expression (4).

  Note that the function F (V) used for calculating the reliability r in the reliability calculation unit 131 is not limited to the function F (V) shown in Expression (4). The reliability calculation unit 131 may use any function as long as the function F (V) gives the reliability r having a smaller value as the variance V of the distance obtained by the variance calculation unit 121 is larger. For example, the function shown in Formula (2) may be used. The reliability r calculated by the reliability calculation unit 131 corresponding to each light spot is passed to the adjustment unit 133.

  On the other hand, the detection unit 132 included in the evaluation unit 13a receives the difference d calculated by the difference calculation unit 123 corresponding to each light spot, and detects the light spot having the received difference d equal to or greater than the second threshold as the surroundings. It is detected as a second light spot from which a distance different from the distance recorded corresponding to the light spot is obtained. Here, in the detection unit 132, the second threshold value for determining whether or not a certain light spot is the second light spot is calculated by, for example, the average calculation unit 122 for a plurality of light spots adjacent to the light spot. It is desirable to set the value to about 20% of the average value Av of the measured distance.

  The adjustment unit 133 illustrated in FIG. 7 receives information indicating the second light spot detected by the detection unit 132 and the reliability r calculated for each light spot by the reliability calculation unit 131. Then, the adjustment unit 133 sets the reliability r corresponding to the light spot detected as the second light spot among the reliability r received from the reliability calculation unit 131 for each light spot, for example, a predetermined value smaller than 1. By multiplying the coefficient, the reliability r obtained for the second light spot is adjusted. That is, the adjustment unit 133 performs adjustment to change the reliability r obtained for the second light spot in a direction to reduce the reliability r of the light spot detected as the second light spot. The value is lower than the reliability r corresponding to the point. Then, the reliability r adjusted by the adjusting unit 133 is passed to the correcting unit 14 and used for correcting the distance of each light spot by the correcting unit 14.

  FIG. 8 shows an example of the distance obtained by the measuring apparatus ME shown in FIG. Of the elements shown in FIG. 8, the elements equivalent to those shown in FIG. 3 are denoted by the same reference numerals and description of the elements may be omitted. FIG. 8 shows the case where the object TG shown in FIG. 7 moves in a direction away from the projection device PR, and the measurement device ME performs the light spot shown in FIG. 7 at each time from time T1 to time T10. The distance obtained by performing the measurement is shown.

  In FIG. 8, white circles shown corresponding to the times T1 to T10 indicate that each of the regions used for detecting the light spot S5 shown in FIG. 7 includes noise similar to the noise N5 shown in FIG. The example of the distance obtained by the measurement of the measuring device ME at the time is shown. On the other hand, the black circles shown corresponding to the times T1 to T10 in FIG. 8 show examples of distances measured by the measuring apparatus ME based on the positions of the light spots detected from the area not including noise. For example, when no noise is included in the area corresponding to each light spot other than the light spot S5 included in the pattern PT shown in FIG. 7, the distances obtained for the light spots detected from each area are all FIG. 8 shows a distribution similar to the distribution indicated by black circles.

  As can be seen from the distribution of the black circles shown in FIG. 8, the distance obtained based on the position of the light spot detected from the area not including noise is along the straight line Lc indicating the change in distance due to the movement of the object TG. Distributed. On the other hand, as can be seen from the distribution of the white circles shown in FIG. 8, the distance obtained by the measurement for the light spot S5 is either the noise in the region used for the detection of the light spot S5 or the light spot S5 randomly. In order to be detected, it changes irregularly like a broken line L5.

  Therefore, based on the distances indicated by white circles in FIG. 8, the value of the distance variance V calculated by using the equation (3) by the variance calculation unit 121 shown in FIG. Based on each distance shown, the distance becomes larger than the value of the variance V of the distance calculated by the variance calculator 121. Therefore, for each distance indicated by white circles in FIG. 8, the value of the reliability r calculated by using the equation (4) by the reliability calculation unit 131 shown in FIG. 7 is the distance indicated by the black circle in FIG. Is smaller than the reliability r calculated for.

  That is, the variance calculation unit 121 calculates the distance variance V using, for example, Equation (3), and the reliability calculation unit 131 calculates the reliability r using the calculated distance variance V. The reliability r in consideration of the movement of the object TG shown in FIG. 1 can be calculated.

  By the way, when the surface of the object TG shown in FIG. 1 is a flat surface or a smooth curved surface, the distances measured for each light spot included in the pattern PT projected on the surface of the object TG are substantially equal or each light It changes continuously according to the position of the point in the image IMG. Therefore, for each light spot shown in FIG. 2, when the distances obtained by the measuring device ME shown in FIG. 7 are likely, the average value Av of the distances obtained corresponding to each light spot is They are close to each other or show a smooth change depending on their position in the image IMG. In this case, the average value Av of the distance obtained by the average calculation unit 122 for a certain light spot is substantially the same as the average value Av of the distance obtained for the light spots around the light spot. In other words, when the average value Av of distances obtained for a certain light spot does not approximate the average value Av of distances obtained for a plurality of adjacent light spots, that is, obtained by the difference calculation unit 123 shown in FIG. When the difference d is large, the reliability r of the distance obtained for the light spot is lower than that of the adjacent light spot. Here, the difference d obtained for each light spot by the difference calculation unit 123 shown in FIG. 7 is obtained by calculating the average distance Av obtained for a certain light spot from the average distance Av obtained for surrounding light spots. The degree of separation, ie the spatial dispersion of the distance obtained for a certain light spot. Therefore, the second threshold value used for detection of the second light spot in the detection unit 132 is desirably set to a value indicating a limit at which it can be determined that there is continuity with the position of the adjacent light spot, for example. In this case, the detection unit 132 sets the light spot whose difference d obtained by the difference calculation unit 123 is equal to or greater than the second threshold value to the second light spot whose reliability of the distance measured by the measurement device ME is lower than the other light spots. It can be detected as two light spots.

  In the example of FIG. 8, the difference between the average value Da5 of the distance obtained for the light spot S5 shown in FIG. 2 and the average value Dac of the distance obtained for each light spot adjacent to the light spot S5 in FIG. d is larger than the second threshold Th2 set to about 20 percent of the average value Dac. In this case, the detection unit 132 passes information indicating that the light spot S5 has been detected as one of the second light spots to the adjustment unit 133 illustrated in FIG. Then, the adjustment unit 133 multiplies the reliability r obtained by the reliability calculation unit 131 for the light spot S5 by a predetermined coefficient smaller than 1, thereby adjusting the reliability adjusted to a value smaller than the original reliability r. The degree r is obtained.

  In other words, the evaluation unit 13a having the detection unit 132 and the adjustment unit 133 sets the reliability of the second light spot that does not satisfy the continuity with the distance obtained from other light spots distributed in the pattern PT to other values. It can be evaluated lower than the reliability of the light spot. Thereby, for example, even when noise is erroneously detected as the light spot S5 in all of the n measurements by the measuring device ME, the evaluation unit 13a determines the reliability r of the distance obtained for the light spot S5. It can be evaluated lower than the reliability r of the distance of the surrounding light spot.

  As described above, the first calculation unit 12a of the object recognition apparatus 10a illustrated in FIG. 7 detects noise as each light spot belonging to the pattern PT even when the object TG illustrated in FIG. 1 moves. The dispersion of the distance indicating the high possibility of being The evaluation unit 13a including the detection unit 132 and the adjustment unit 133 determines the light spot when the distance of each light spot included in the pattern PT does not satisfy the continuity with the distance obtained from the surrounding light spots. The reliability of the distance is low. Therefore, the object recognition device 10a can obtain the reliability of the distance recorded corresponding to each light spot in consideration of the spatial continuity of the distance obtained for each light spot as the object TG moves. it can. And the object recognition apparatus 10a recognizes the shape of the object TG by the recognition part 15 based on the distance to each light spot corrected by the correction part 14 according to the calculated reliability. Thereby, the object recognition apparatus 10a can recognize the shape of the object TG with higher accuracy than in the case where the spatial continuity of the distance obtained by the movement of the object TG and the measurement for each light spot is not considered. it can.

  Next, a method of using the reliability r obtained corresponding to each light spot by the evaluation unit 13 shown in FIG. 1 or the evaluation unit 13a shown in FIG. 7 for recognizing the shape of the object TG will be described.

  FIG. 9 shows another embodiment of the object recognition apparatus. 9 that are equivalent to the components shown in FIG. 1 or FIG. 7 are denoted by the same reference numerals and description of the components may be omitted.

  The object recognition apparatus 10b illustrated in FIG. 9 includes a recognition unit 15b in addition to the recording unit 11, the first calculation unit 12, the evaluation unit 13, and the correction unit 14 illustrated in FIG. The recognition unit 15b illustrated in FIG. 9 is a component corresponding to the recognition unit 15 illustrated in FIG. 1 and includes a setting unit 151 and a shape calculation unit 152.

  The setting unit 151 illustrated in FIG. 9 receives the reliability r obtained by the evaluation unit 13 corresponding to each light spot, and the recording unit 11 corresponds to each light spot based on the received reliability r. A smaller weight is set to the recorded distance as the reliability is lower. For example, the setting unit 151 sets the reliability r received from the evaluation unit 13 corresponding to each light spot as the weight of the distance recorded in the recording unit 11 corresponding to each light spot.

  In addition, the shape calculation unit 152 obtains the shape of the object TG from the distance recorded in the recording unit 11 corresponding to each light spot, using the weight set corresponding to each light spot by the setting unit 151. For example, the shape calculation unit 152 obtains a plane or curved surface representing the shape of the surface of the object TG by regression analysis or the like in which the weights set by the setting unit 151 are added to the distances obtained by the measurement for each light spot. .

  Thereby, the shape calculation unit 152 faithfully reflects the distance obtained for the light spot with the higher reliability r than the distance obtained for the light spot with the lower reliability r by the evaluation unit 13. The shape to be obtained can be obtained as the shape of the object TG.

  In addition, as one piece of information indicating the shape of the object TG, the shape calculation unit 152 sets, for example, the centroid of the figure indicating the surface of the object TG on which the pattern PT is projected, by the setting unit 151 for each light spot. You may obtain | require in consideration of a weight. The shape calculation unit 152 uses, for example, Expression (5) for calculating the distance c to the center of gravity of the graphic indicating the surface of the object TG. In Expression (5), for example, the vector X (k) is a position indicating the position where the kth light spot included in the pattern PT is projected when the pattern PT is projected onto the plane PL shown in FIG. Is a vector. In Expression (5), symbol h (k) indicates a distance obtained by measurement for the kth light spot, and symbol w (k) is set by the setting unit 151 for the kth light spot. Weight (for example, reliability r). Here, the numerical value k is a positive integer equal to or less than the number N of light spots included in the pattern PT.

  The recognizing unit 15b described above can obtain the shape of the surface of the object TG on which the pattern PT is projected, using the weight set based on the reliability r obtained corresponding to each light spot. . As a result, the recognition unit 15b can recognize the shape of the object TG while reflecting the distance corrected by the correction unit 14 while emphasizing the distance determined to be highly reliable by the evaluation unit 13. Therefore, the recognition unit 15b can reduce the possibility of erroneously recognizing the shape of the object TG compared to the case where the reliability of the distance obtained for each light spot is not considered in the recognition of the shape of the object TG. it can.

  Next, a method of using the reliability r obtained corresponding to each light spot by the evaluation unit 13 shown in FIG. 1 or the evaluation unit 13a shown in FIG. 7 as a feature amount used for recognition of the object TG will be described. To do.

  FIG. 10 shows another embodiment of the object recognition apparatus. 10 that are equivalent to the components shown in FIG. 1 are denoted by the same reference numerals and description of the components may be omitted.

  The object recognition apparatus 10c shown in FIG. 10 includes a recognition unit 15c together with the recording unit 11, the first calculation unit 12, the evaluation unit 13, and the correction unit 14 shown in FIG. The recognition unit 15c illustrated in FIG. 10 is a component corresponding to the recognition unit 15 illustrated in FIG. 1 and includes an extraction unit 153, a holding unit 154, and a specifying unit 155.

  The extraction unit 153 extracts features of the object TG from the reliability obtained by the evaluation unit 13 corresponding to a plurality of light spots included in the pattern PT projected on the object TG. For example, the extraction unit 153 projects the reliability obtained by the evaluation unit 13 for each light spot projected onto the object TG for each of a plurality of classes into which the reliability range is divided, thereby projecting the reliability onto the object TG. A frequency distribution of reliability obtained corresponding to each light spot is obtained as one of the features of the object TG. In addition, the extraction unit 153 aggregates the frequency distribution of the distance recorded in the recording unit 11 corresponding to each light spot for each of a plurality of classes into which the distance range is divided, so that the light projected on the object TG The frequency distribution of the distance to the point is obtained as another feature of the object TG. Note that the frequency distribution of distances and the frequency distribution of reliability obtained as the features of the object TG by the extraction unit 153 will be described later with reference to FIGS. 11 and 12.

  The holding unit 154 holds information indicating reliability of distances obtained in advance for each of a plurality of light spots projected on each of the plurality of types of objects as information indicating the characteristics of the plurality of types of objects. . For example, when the pattern PT is projected onto each of a plurality of types of objects, the holding unit 154 obtains the frequency distribution of distances obtained by measurement based on the positions of a plurality of light spots included in the pattern PT and the measurements. Information indicating the frequency distribution of distance reliability is stored in advance for each type of object. Note that the distance frequency distribution and the reliability frequency distribution previously held in the holding unit 154 are combined with the distance frequency distribution and the reliability frequency distribution obtained as features of the object TG by the extraction unit 153 in FIG. And it mentions later using FIG.

  The identification unit 155 receives the frequency distribution of distances and the frequency distribution of reliability obtained by the extraction unit 153 as information indicating the characteristics of the object TG. Then, the identifying unit 155 uses the characteristics of the object TG extracted by the extraction unit 153 in the information indicating the characteristics of each of the multiple types of objects held by the holding unit 154 rather than the characteristics of the other types of objects. A type of object having similar characteristics is specified as the type of object TG. Then, the specifying unit 155 outputs information indicating the type of the specified object as a recognition result by the object recognition device 10c.

  Here, the relationship between the characteristics of the object TG in the area where the pattern PT is projected and the reliability of the distance obtained by the measuring apparatus ME shown in FIG. 10 will be described.

  FIG. 11 shows an example of features for each type of object. Each of FIGS. 11A and 11B shows the measurement by the measuring apparatus ME shown in FIG. 10 when the pattern PT is projected onto each of the person QA wearing glasses and the person QB wearing glasses. An example of one of n images IMG obtained in the process is shown. Further, FIGS. 11C and 11D show the distances obtained from the positions of the light spots included in the images IMG shown in FIGS. 11A and 11B, respectively. The example of the reliability calculated | required by the evaluation part 13 is shown.

  11A and 11B, the pattern PT is composed of 20 light spots S1, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, arranged in 4 rows and 5 columns. S12, S13, S14, S15, S16, S17, S18, S19, S20 are included. The number of light spots included in the pattern PT and the way of arranging the light spots are limited to the case where 20 light spots are arranged in 4 rows and 5 columns as shown in FIGS. However, any arrangement may be used as long as N (N is an integer of 2 or more) light spots are distributed two-dimensionally. For example, the pattern PT may be a pattern in which N light spots are arranged concentrically. Further, the outlines of the persons QA and QB indicated by broken lines in FIGS. 11A and 11B are shown for explaining the positional relationship between the persons QA and QB and each light spot included in the pattern PT. Is not included in the actual image IMG.

  In FIGS. 11A and 11B, light spots S2, S3, S4, S6, S7, S8, S9, S10, S11, S12, S13, S14, S15, S17, S18, and S19 indicated by black circles, respectively. Indicates a light spot projected on each of the persons QA and QB. That is, in FIGS. 11A and 11B, each of the light spots S2 to S4, S6 to S15, and S17 to S19 indicated by black circles is a light spot included in the image IMG. On the other hand, light spots S1, S5, S16, and S20 indicated by white circles in FIGS. 11A and 11B indicate light spots projected on the space behind the persons QA and QB. Therefore, each of the light spots S1, S5, S16, and S20 shown in FIGS. 11A and 11B is not included in the actual image IMG.

  The shape of the human head is almost the same. Therefore, based on the position of each light spot indicated by the black circle in FIG. 11A, the distance obtained by the measuring device ME shown in FIG. 10 and the light spot indicated by the black circle in FIG. The distance obtained by the device ME is a value approximate to each other.

  Here, the light spots S6, S7, S9, and S10 shown in FIG. 11B are projected on the surface of the glasses worn by the person QB. Therefore, the reliability obtained by the evaluation unit 13 shown in FIG. 10 with respect to the distances obtained corresponding to the light spots S6, S7, S9, and S10 included in the image IMG shown in FIG. It may be lower than the reliability obtained for other light spots. Because the surface of the glasses has a higher reflectance than the surface of the face of the person QB, the luminance in the image IMG of the light spots S6, S7, S9, and S10 projected onto the surface of the glasses becomes small, and the light spot This is because it becomes difficult to detect S6, S7, S9, and S10. In this case, there is a higher possibility that noise that appears due to reflection of external light or the like is erroneously detected as each of the light spots S6, S7, S9, and S10 as compared to other light spots. For this reason, the reliability obtained corresponding to each of the light spots S6, S7, S9, and S10 is lower than the reliability corresponding to the other light spots.

  Each of the rectangular areas shown in FIGS. 11C and 11D corresponds to the light spots S1 to S20 shown in FIGS. 11A and 11B, respectively. In FIGS. 11C and 11D, the cross marks shown in the areas corresponding to the light spots S1, S5, S16, and S20 cannot be measured by the measuring device ME shown in FIG. Is shown.

  In the examples of FIGS. 11C and 11D, the range of reliability values is divided into four classes, and the classes to which the reliability levels corresponding to the light spots S1 to S20 belong are assigned to the corresponding areas. Shown with different types of hooks. In the examples of FIGS. 11C and 11D, each of the four classes obtained by dividing the range of reliability values is shown with dark shading in order from the class including the larger value.

  That is, in the example of FIG. 11C, the reliability obtained for the light spots S12 and S14 belongs to the class including the largest value, and is obtained for the light spots S2 to S4, S7 to S9, S13, and S18. The reliability is shown to belong to a class including the next largest value. In the example of FIG. 11C, it is indicated that the reliability obtained for the light spots S6, S10, S11, S15, S17, and S19 belongs to the class including the third largest value.

  On the other hand, in the example of FIG. 11D, the reliability obtained for the light spots S12 and S14 belongs to the class including the largest value, and the reliability obtained for the light spots S2 to S4, S8, S13, and S18. Is shown to belong to the class containing the next largest value. In the example of FIG. 11D, the reliability obtained for the light spots S11, S15, S17, and S19 belongs to the class including the third largest value, and is obtained for the light spots S6, S7, S9, and S10. The given reliability is shown to belong to the class including the smallest value.

  As can be seen from a comparison between FIG. 11C and FIG. 11D, the reliability distribution obtained for each light spot projected onto the person QB wearing glasses is the person who is not wearing glasses. A characteristic different from the reliability distribution obtained for each light spot projected onto the QA is shown. For example, in the distribution of the reliability corresponding to the person QB shown in FIG. 11D, the reliability belonging to the class corresponding to the smallest value is distributed corresponding to the light spots S6, S7, S9, and S10. Indicates that

  The reliability distribution feature described with reference to FIG. 11 includes, for example, the reliability obtained corresponding to each light spot included in the pattern PT for each of a plurality of classes into which the range of reliability values is divided. It can be represented by a frequency distribution obtained by counting. Here, as described with reference to FIG. 11, the frequency distribution obtained for the reliability reflects the characteristics of the object on which the pattern PT is projected. Therefore, if a frequency distribution of reliability is obtained in advance for a plurality of types of objects, and the obtained frequency distribution of reliability is held in the holding unit 154 shown in FIG. 10, the pattern PT is projected onto the region. It can be used to specify the included object TG.

  FIG. 12 shows an example of the holding unit 154 shown in FIG. The holding unit 154 illustrated in FIG. 12 holds a feature vector including a distance frequency distribution and a reliability frequency distribution for each type of object. In FIG. 12, as an example of feature vectors corresponding to various types of objects, a feature vector corresponding to person QA and a feature vector corresponding to person QB are shown, and other objects are held in holding unit 154. Illustration of feature vectors is omitted.

  Here, the frequency distribution of distances is a total of distances obtained corresponding to each light spot included in the pattern PT shown in FIG. 10 for each of a plurality of classes obtained by dividing a range of values obtained as distances. This is the frequency distribution obtained.

  In the example of FIG. 12, it is obtained for each light spot included in the pattern PT for each of the four classes Cd1, Cd2, Cd3, and Cd4 obtained by dividing the distance range obtained as a measurement result by the measurement apparatus ME shown in FIG. The case where the frequency distribution of distance is obtained by totaling the distance is shown.

  In the example of FIG. 12, the light spot included in the pattern PT is obtained for each of the four classes Cr1, Cr2, Cr3, Cr4 obtained by dividing the reliability range obtained by the evaluation unit 13 shown in FIG. The case where the frequency distribution of reliability is obtained by totaling the reliability is shown. Here, each of the four classes Cr1, Cr2, Cr3, and Cr4 shown in FIG. 12 corresponds to each of the four classes shown in FIGS. 11C and 11D by changing the type of shading. . For example, the class Cr1 shown in FIG. 12 is the class shown in white in FIG. 11 (D), indicating the class including the smallest value, and the class Cr4 shown in FIG. 12 is the most in FIG. 11 (D). It is a class indicated by dark shading and indicates a class including the largest value. Further, the class Cr3 shown in FIG. 12 is a class indicated by the darkest shade after the class Cr4 in FIG. 11D, and indicates a class including the second largest value. And the class Cr2 shown in FIG. 12 is a class shown by the darkest shade next to the class Cr3 in FIG. 11D, and indicates a class including the third largest value.

  The feature vector held in correspondence with the person QA in the holding unit 154 shown in FIG. 12 aggregates the reliability shown corresponding to each light spot in FIG. 11C for each of the classes Cr1 to Cr4. The frequency distribution [0, 6, 8, 2] of the reliability obtained by this is included. In addition, the feature vector corresponding to the person QA is the frequency of the distance obtained by totaling the distances obtained by the measurement based on the positions of the respective light spots shown in FIG. 11A for each of the classes Cd1 to Cd4. The distribution [2, 7, 5, 2] is included.

  Further, the feature vector held in correspondence with the person QB in the holding unit 154 shown in FIG. 12 has the reliability shown in correspondence with each light spot in FIG. 11D for each of the classes Cr1 to Cr4. The frequency distribution [4, 4, 6, 2] of the reliability obtained by the aggregation is included. Further, the feature vector corresponding to the person QB is the frequency of the distance obtained by totaling the distances obtained by the measurement based on the positions of the respective light spots shown in FIG. 11B for each of the classes Cd1 to Cd4. The distribution [3, 8, 4, 1] is included.

  The frequency distribution of the distances included in the feature vectors corresponding to each of the persons QA and QB shown in FIG. 12 shows the characteristics of the shapes of the persons QA and QB. On the other hand, the frequency distribution of the reliability corresponding to each of the persons QA and QB shown in FIG. 12 is a surface indicated by, for example, the degree that a highly reflective part such as glasses is included in the surfaces of the persons QA and QB. The characteristics of

  As described with reference to FIG. 11, since the shapes of the human head are similar to each other, each element included in the frequency distribution of distance shown in FIG. 12 corresponding to the person QA and the person QB correspond to each other. The elements included in the frequency distribution of distances shown in FIG.

  On the other hand, the frequency of the class Cr1 included in the frequency distribution of the reliability shown corresponding to the person QA in FIG. 12 and the frequency of the class Cr1 included in the frequency distribution of the reliability shown corresponding to the person QB are: It is very different from the frequency of other classes. Therefore, if feature vectors including the frequency distribution of reliability are compared, for example, whether the object TG on which the pattern PT is projected is the person QA regardless of the similarity in shape between the person QA and the person QB. It is possible to determine whether it is QB.

  As described in FIG. 10, the extraction unit 153 obtains the frequency distribution of the distance recorded in the recording unit 11 for each light spot projected on the object TG and the information about each light spot as information indicating the characteristics of the object TG. The frequency distribution of the reliability of the given distance is obtained. Then, the extraction unit 153 passes the feature vector including the obtained frequency distribution of distance and the frequency distribution of reliability to the specifying unit 155.

  For example, the specifying unit 155 collates the feature vector received from the extraction unit 153 with each of the feature vectors held in the holding unit 154 for each type of object. Then, the specifying unit 155 finds a feature vector similar to the feature vector corresponding to the other type to the feature vector received from the extracting unit 153, and specifies the type of the object corresponding to the found feature vector. The specifying unit 155 outputs the type of the object specified based on the feature vector collation as a recognition result for the object TG in the area where the pattern PT is projected.

  As described above, the recognizing unit 15c shown in FIG. 10 has the reliability of the distance obtained for each light spot together with the frequency distribution of the distance obtained by the measuring device ME for each light spot projected on the object TG. A frequency distribution of degrees is used as a feature of the object TG for recognition of the object TG. Thereby, the object recognition apparatus 10c having the recognition unit 15c illustrated in FIG. 10 discriminates objects having similar shapes, such as the person QA and the person QB, based on the reliability information included in the feature vector. be able to.

  Note that the feature of the object TG extracted by the extraction unit 153 shown in FIG. 10 is not limited to the frequency distribution of reliability, but may be any information that can be extracted from the reliability obtained corresponding to each light spot. For example, the extraction unit 153 uses information indicating how the first light spots described with reference to FIGS. 1 to 4 are distributed in the pattern PT projected on the object TG as information on the object TG onto which the pattern PT is projected. You may extract as one of the characteristics. In this case, for example, information indicating the distribution of reliability as illustrated in FIGS. 11C and 11D is obtained in advance when the pattern PT is projected onto each of a plurality of types of objects. Information indicating the distribution of reliability is held in the holding unit 154 shown in FIG. Information indicating the distribution of reliability held in the holding unit 154 corresponding to each of a plurality of types of objects includes, for example, the recording unit 11, the first calculation unit 12, and the like of the object recognition device 10c illustrated in FIG. It can be obtained in advance using the functions of the evaluation unit 13 and the extraction unit 153.

  The object recognition device 10 of the present disclosure described above can be realized using a computer device.

  FIG. 13 shows a hardware configuration example of the object recognition apparatus 10c shown in FIG. Note that among the constituent elements shown in FIG. 13, the same constituent elements as those shown in FIG. 10 are denoted by the same reference numerals, and description thereof is omitted.

  The computer device COM shown in FIG. 13 includes a processor 21, a memory 22, a storage device 23, a general-purpose interface 24, a display device 25, an optical drive device 26, and a network interface 28. The processor 21, the memory 22, the storage device 23, the general-purpose interface 24, the display device 25, the optical drive device 26, and the network interface 28 shown in FIG. 13 are connected to each other via a bus. Further, the general-purpose interface 24 is connected to a projection device PR1 having a function equivalent to that of the projection device PR shown in FIG. 1 and a camera CAM1 having a function equivalent to the imaging unit CAM shown in FIG.

  The processor 21, the memory 22, the storage device 23, and the general-purpose interface 24 illustrated in FIG. 13 are included in the object recognition device 10c. The processor 21, the memory 22, the storage device 23, the general-purpose interface 24, and the camera CAM1 are included in the measurement device ME.

  For example, the projection apparatus PR1 shown in FIG. 13 projects a pattern PT1 including a plurality of light spots on a predetermined region including the person Q1 and the bed BD, similar to the pattern PT described in FIGS. . Further, the camera CAM1 shown in FIG. 13 captures, for example, a region at which the pattern PT1 is projected by the projection device PR1 at a predetermined time interval, and each of a plurality of images IMG obtained by the imaging is displayed on the general-purpose interface 24. To the processor 21.

  The optical drive device 26 shown in FIG. 13 can be loaded with a removable disk 27 such as an optical disk, and reads and records information recorded on the mounted removable disk 27.

  The computer device CAM is connected to a network NW such as the Internet via the network interface 28, and can exchange information with the server device SV connected to the network NW.

  The memory 22 shown in FIG. 13 is an application program for the processor 21 to execute the object recognition process described with reference to FIGS. 10 to 12 and the distance measurement process described with reference to FIG. 1 together with the operating system of the computer apparatus COM. Is stored. The application program for executing the object recognition process and the distance measurement process can be recorded and distributed on the removable disk 27 such as an optical disk, for example. The processor 21 may store the read application program in the memory 22 and the storage device 23 by attaching the removable disk 27 to the optical drive device 26 and performing read processing. Further, an application program for executing the object recognition process and the distance measurement process may be downloaded from the server device SV, for example, via the network interface 28 and the network NW. The downloaded application program is read into the memory 22 and the storage device 23 to become an application program that can be executed by the processor 21.

  The processor 21 executes an application program for object recognition processing stored in the memory 22 or the like, whereby the first calculation unit 12, the evaluation unit 13, the correction unit 14, and the recognition unit 15 illustrated in FIG. The functions of the extraction unit 153 and the identification unit 155 included in the above are performed. Further, the processor 21 executes an application program for distance measurement processing stored in the memory 22 or the like, so that the bed BD and the person can be detected from the position of each light spot included in the image IMG photographed by the camera CAM1. The distance to each light spot projected on Q1 is obtained. That is, the processor 21 performs the function of the image processing unit IMP shown in FIG. 1 by executing an application program for distance measurement processing stored in the memory 22 or the like. The processor 21 records the distance to each light spot obtained by executing an application program for distance measurement processing, for example, in the recording unit 12 provided in the storage area of the memory 22 or the storage device 23. .

  Further, the application program for object recognition processing may include a feature vector obtained in advance for each of a plurality of types of objects as information indicating the characteristics of each of the plurality of types of objects. In this case, the feature vector obtained in advance for each of the plurality of types of objects is stored in a part of the storage area of the storage device 23, for example. That is, the holding unit 154 illustrated in FIG. 10 is realized using a part of the storage area of the storage device 23.

  As described above, the object recognition device 10c shown in FIG. 10 includes, for example, the cooperation of the processor 21, the memory 22, the storage device 23, and the general-purpose interface 24 included in the computer device COM shown in FIG. It can be realized by working.

  FIG. 14 shows the operation of the object recognition apparatus 10c shown in FIG. Each process of step S300 to step S304 and step S311 to step S313 illustrated in FIG. 14 is an example of a process included in the application program for the object recognition process. In addition, each process of step S300 to step S304 and step S311 to step S313 is executed by the processor 21 shown in FIG. Note that the processing in steps S301 to S304 shown in FIG. 14 is the same as the processing in steps S301 to S304 shown in FIG.

  For example, every time the processor 21 receives n images IMG from the camera CAM1 illustrated in FIG. 13, the processor 21 performs the processes of steps S300 to S304 and steps S311 to S313 illustrated in FIG. 14.

  In step S300, the processor 21 executes an application program for distance measurement processing, and thereby reaches each light spot based on the position of the light spot included in each of the n images IMG received from the camera CAM1. Measure the distance.

  In step S301, the processor 21 calculates the distance measured using each of the n images IMG by the processing in step S300, and the recording unit 11 provided in the memory 22 or the storage device 23 corresponding to each light spot. To record.

  In step S302, the processor 21 calculates a variance of n distances recorded in the recording unit 11 corresponding to each light spot.

  In step S303, the processor 21 evaluates the reliability of the distance recorded in the recording unit 11 corresponding to each light spot, based on the variance calculated corresponding to each light spot in the process of step S302.

  In step S304, the processor 21 sets the distance recorded corresponding to the first light spot whose reliability is lower than the predetermined first threshold in the process of step S303 to the other adjacent to the first light spot. Correction is performed using the distance recorded corresponding to the light spot.

  In step S311, the processor 21 projects the light spots included in the pattern PT1, for example, based on the distances recorded in the recording unit 11 corresponding to the light spots and the positions of the light spots in the image IMG. Divide into groups for each object. For example, the processor 21 performs the clustering using the distance corresponding to each light spot and the position of each light spot in the image IMG to change the light spot included in the pattern PT1 to the group including the light spot projected on the person Q1. It classify | categorizes into the group containing the light spot projected on the bed BD.

  In step S312, the processor 21 obtains the feature vector described in FIGS. 11 and 12 for each group obtained in the process of step S311.

  In step S312, the processor 21 sets the types of objects associated with feature vectors similar to other feature vectors to the feature vectors obtained for each light spot group in the process of step S312 and corresponding to each group. Specified as the type.

  Here, the holding unit 154 provided in the storage device 23 includes, as the bed BD, a feature vector corresponding to the state in which the person Q1 is in various postures, and a feature vector indicating features of different types of objects. It is possible to hold it in advance. In this case, the processor 21 can determine the positional relationship between the person Q1 and the bed BD, the posture of the person Q1, and the like based on the type of object specified in the process of step S313.

  That is, the object recognition apparatus 10c shown in FIG. 13 can be used in the field of, for example, a service for watching a person in the room.

  From the above detailed description, features and advantages of the embodiment will become apparent. It is intended that the scope of the claims extend to the features and advantages of the embodiments as described above without departing from the spirit and scope of the right. Any person having ordinary knowledge in the technical field should be able to easily come up with any improvements and changes. Therefore, there is no intention to limit the scope of the inventive embodiments to those described above, and appropriate modifications and equivalents included in the scope disclosed in the embodiments can be used.

10, 10a, 10b, 10c ... object recognition device; 11 ... recording unit; 12, 12a ... first calculation unit; 13, 13a, 13b ... evaluation unit; 14 ... correction unit; 15, 15b, 15c ... recognition unit; ... dispersion calculation unit; 122 ... average calculation unit; 123 ... difference calculation unit; 131 ... reliability calculation unit; 132 ... detection unit; 133 ... adjustment unit; 151 ... setting unit; 152 ... shape calculation unit; 154: Holding unit; 155: Specific unit; 21 ... Processor; 22 ... Memory; 23 ... Storage device; 24 ... General-purpose interface; 25 ... Display device: 26 ... Optical drive device; 27 ... Removable disk; , PR1 ... projection device; ME ... measuring device; CAM ... imaging unit; CAM1 ... camera; IMP ... image processing unit; TG ... object; PT, PT1 ... pattern QA, QB, Q1 ... person; BD ... bed; COM ... computer equipment; NW ... network; SV ... server device

Claims (6)

Corresponding to each light spot is a distance to each of the plurality of light spots obtained from each of a plurality of images obtained by imaging a pattern including a plurality of light spots projected on a region including an object at time intervals. A recording section for recording,
A first calculation unit that calculates a variance of a plurality of distances recorded in the recording unit for each of the light spots included in the pattern;
An evaluation unit that evaluates the reliability of the distance recorded corresponding to each light spot, based on the variance calculated by the first calculation unit;
The distance recorded by the evaluation unit corresponding to the first light spot whose reliability is determined to be equal to or less than the first threshold is recorded corresponding to another light spot adjacent to the first light spot. A correction unit for correcting using the distance;
An object recognition apparatus characterized by comprising: The object recognition apparatus according to claim 1,
The first calculation unit includes an average value of the distance recorded in the recording unit corresponding to each light spot, and an average of the distance recorded corresponding to a plurality of light spots adjacent to the light spot. A second calculation unit for obtaining a difference from the value;
The evaluation unit is
A detector that detects a second light spot at which a difference obtained by the second calculator is equal to or greater than a second threshold;
An object characterized in that the reliability of the distance recorded corresponding to the second light spot detected by the detection unit is evaluated lower than the reliability of the distance recorded corresponding to the other light spot. Recognition device. In the object recognition device according to claim 1 or 2,
A recognition unit for recognizing the shape of the object included in the region based on the distance recorded in the recording unit corresponding to each light spot;
The recognition unit
A setting unit that sets a smaller weight to the distance recorded in the recording unit corresponding to each light spot as the reliability obtained by the evaluation unit is lower;
An object recognition apparatus, wherein the shape of the object is obtained from the distance recorded corresponding to each light spot using the weight set by the setting unit. In the object recognition device according to claim 1 or 2,
A recognition unit for recognizing the shape of the object included in the region based on the distance recorded in the recording unit corresponding to each light spot;
The recognition unit
An extraction unit that extracts features of the object from the reliability obtained by the evaluation unit for each light spot included in the pattern projected onto the object;
A holding unit that holds information indicating reliability of distances obtained in advance for each of the plurality of light spots projected on each of the plurality of types of objects as information indicating the characteristics of the plurality of types of objects;
Among the information indicating the characteristics of each of the plurality of types of objects held in the holding unit, the type having characteristics similar to the characteristics of the object extracted by the extraction unit than the characteristics of other types of objects An object recognizing apparatus comprising: a specifying unit that specifies the object as a type of the object included in the region. Corresponding to each light spot is a distance to each of the plurality of light spots obtained from each of a plurality of images obtained by imaging a pattern including a plurality of light spots projected on a region including an object at time intervals. And record
Calculating a variance of a plurality of distances recorded for each of the light spots included in the pattern;
Based on the calculated variance, evaluate the reliability of the recorded distance corresponding to each light spot,
The recorded distance corresponding to the first light spot whose reliability is less than or equal to the first threshold is used as the distance recorded corresponding to the other light spot adjacent to the first light spot. to correct,
An object recognition method characterized by the above. Corresponding to each light spot is a distance to each of the plurality of light spots obtained from each of a plurality of images obtained by imaging a pattern including a plurality of light spots projected on a region including an object at time intervals. And record
Calculating a variance of a plurality of distances recorded for each of the light spots included in the pattern;
Based on the calculated variance, evaluate the reliability of the recorded distance corresponding to each light spot,
The recorded distance corresponding to the first light spot whose reliability is less than or equal to the first threshold is used as the distance recorded corresponding to the other light spot adjacent to the first light spot. to correct,
An object recognition program that causes a computer to execute processing.

Images