JP4542666B2 - Foreign object detection method and apparatus by image processing - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a foreign matter detection method and a foreign matter detection apparatus that perform image processing to detect the presence or absence of mixed foreign matter by applying threshold processing to image data including physical property information of an object to be inspected.
[0002]
[Prior art]
In a production line for foods and various industrial products, foreign matter detection by image processing using X-rays or the like has been conventionally performed in order to eliminate foreign matters. Foreign matter detection by image processing is performed by irradiating an inspection object with X-rays from an X-ray source, etc., and converting a transmission image of the inspection object into digital data with an X-ray detector or the like, and using this as transmission image data with an image input means The transmission image data obtained and stored in the memory is logarithmically converted into absorption image data indicating the amount of absorption by the inspection object as necessary, and image processing is performed on these image data including physical property information of the inspection object. This is a technique for detecting presence / absence of a mixed foreign object by applying a thresholding process by applying an image filter for reducing noise or an image filter for enhancing the likelihood of a foreign substance as required. Here, X-rays, X-ray sources, etc., X-ray detectors, etc. represent X-rays that represent electromagnetic waves such as light, X-rays, etc. used for foreign object detection. It is for explaining to an example.
[0003]
FIG. 10 is a flowchart for explaining a conventional foreign object detection method. An image filter that reduces noise in step S20 is applied to the image data including the physical property information of the object to be inspected that is input in step S10, an image filter that emphasizes the likelihood of a foreign object is applied in step S30, and a threshold value in step S40. Value processing is applied to detect the presence or absence of mixed foreign matter. Here, when there is no need for noise reduction or foreign substance enhancement, one or both of steps S20 and S30 are omitted.
[0004]
For example, for an X-ray absorption image data of a food containing physical property information such as an X-ray absorption amount of a substance, an image filter that emphasizes a high-luminance portion based on the fact that a metal has an X-ray absorption amount larger than that of a food is applied. By emphasizing the image of the high brightness portion and applying threshold processing that uses the high brightness portion as the mixed metal to the obtained foreign matter emphasized image data, the mixed metal in the food can be detected. In addition, for the X-ray transmission image data of food containing physical property information such as the X-ray transmission amount of the substance, an image filter that emphasizes a high-luminance portion is applied on the basis that air has a larger X-ray transmission amount than food. By emphasizing the image of the high-luminance portion and applying threshold processing that uses the high-luminance portion as mixed air to the obtained foreign matter-emphasized image data, mixed air in the food can be detected.
[0005]
Various image filters can be applied to enhance the image of foreign objects, but the production line is required to have high speed capable of withstanding real-time processing. For this reason, a differential image filter using a square kernel of N × N pixel size is often used. For example, with respect to X-ray absorption image data, assuming a kernel area of N × N pixels around each pixel, the difference between the luminance of the pixel at the kernel center and the average luminance in the kernel area is evaluated, An image filter that outputs difference image data in which the pixel value of the pixel corresponding to the pixel at the center of the kernel is how much higher than the average luminance is used. Here, the reason why the square kernel is used is to perform image enhancement processing at high speed and isotropically. There are various other variations of the differential image filter, which increases the image processing load. However, the difference between the average luminance at the kernel center and the average luminance at the kernel periphery can be increased. Although it increases, the maximum or minimum brightness is used instead of the average brightness. When there is a considerable margin in image processing capability, X-ray absorption image data, a minimum value filter for outputting the minimum luminance in the kernel area to the image data, and a maximum value filter for outputting the maximum luminance in the kernel area are provided. It is also possible to apply a top hat filter that uses a difference from image data smoothed by continuous application.
[0006]
It is desirable that the kernel pixel size be larger than the pixel size of the foreign object to be detected. However, if the kernel pixel size is increased to detect a large foreign object, the image processing load increases and the image processing capability is insufficient. If you do. For this reason, in consideration of the image data size and the image processing capability determined by the measurement range and spatial resolution of the foreign object detection, it is necessary to limit the types of applicable image filters and the kernel pixel size. In addition, when the spatial resolution of foreign object detection is increased or the measurement range is expanded, it is necessary to greatly improve the image processing capability to meet the increase in image data size.
[0007]
[Problems to be solved by the invention]
In the conventional foreign object detection method, the type of image filter that can be applied to the image data of the object to be inspected and the pixel size of the kernel depend greatly on the image processing capability. Therefore, the size of detectable foreign matter is limited by the image processing capability. End up. This problem also contributes to making it difficult to increase the spatial resolution of foreign matter detection and widen the measurement range in a foreign matter detection device for a production line. An object of the present invention is to provide a foreign object detection method and a foreign object detection device that are useful for solving these problems and do not depend much on the image processing capability.
[0008]
[Means for Solving the Problems]
In order to solve the above-described problem, the foreign matter detection method and apparatus of the present invention irradiates an inspection object with X-rays or the like from an X-ray source or the like, and transmits a transmission image of the inspection object with digital data using an X-ray detector or the like. This is taken into the memory as transmission image data by the image input means, and the obtained transmission image data is logarithmically converted to absorption image data indicating the amount of absorption by the object to be inspected, if necessary, and the physical property information of the object to be inspected for these image data, including, of the image processing means, by equally dividing the image data into small areas Do that a plurality of pixels, likeness foreign matter local high luminance portion or edge portion on the divided small regions each And reducing the noise by newly generating reduced image data with the evaluation value of the foreign object likelihood as a pixel value and condensing the likelihood of the foreign object. It is performed whether detection of contaminating foreign matter and is applied image filter and threshold processes emphasize.
That is, the foreign matter detecting method according to claim 1, in the foreign matter detecting method for performing presence detection of mixed foreign substance by processing the image data including the property information of the object to be inspected, ing the image data from a plurality of pixels small Equally divide into regions and evaluate the likelihood of foreign matter at the local high-intensity part or edge part for each small region, and newly reduce image data using the evaluation value of the foreign matter likelihood obtained for each small region as a pixel value A foreign matter mixed by applying a threshold value process and an image filter that emphasizes a local high-intensity portion or an edge portion of the reduced data for evaluating the likelihood of the foreign matter with respect to the reduced image data (S2) (S3, S4).
The foreign matter detection method according to claim 2 is the foreign matter detection method according to claim 1, wherein the step (S2) of generating the reduced image data and the step (S3) of detecting the presence or absence of the mixed foreign matter are performed. It is executed multiple times while changing the size of the small area.
Further, in the foreign object detection method according to claim 3, the step of generating the reduced image data of the foreign object detection method according to claim 1 or claim 2 evaluates the first foreign object likelihood for each of the small regions, and First reduced image data having a pixel value as the first foreign object evaluation value obtained for each small area and a second foreign object obtained by evaluating the second foreign substance characteristic for each small area and obtained for each small area Generating the second reduced image data having the likelihood evaluation value as the pixel value, and lowering the pixels of the first reduced image data excluding the pixel position indicating the high luminance of the second reduced image data . Generating a third reduced image data in which the first reduced image data is masked with the second reduced image data by functioning as a foreign substance candidate point extraction mask by replacing with luminance . 3 of the reduced image data is the reduced image data. It is set to be a stage to.
The foreign object detection device according to claim 4 is an X-ray irradiation unit that irradiates the inspection object with X-rays, an X-ray detector that receives the X-rays transmitted through the inspection object and converts the X-rays into digital data, Image input means for capturing digital data output from the X-ray detector as image data, and presence / absence of foreign matter mixed in the inspection object by performing image processing on the image data captured by the image input means in the foreign matter detection device and an image processing means for detecting said image processing means, local high luminance portion or edge portion for each said small region by equally dividing the image data into small areas Do that a plurality of pixels in evaluates foreign likelihood, comprise image reduction means for newly generating a reduced image data to the pixel value the evaluation value of likeness foreign substances obtained for each the small region, the foreign matter ness for the reduced image data Comment The local high luminance portion or edge portion of said reduced data to emphasize image filter and by applying a threshold process for detecting the presence or absence of the foreign substance.
[0009]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a flowchart for explaining the foreign object detection method of the present invention. The image data including the property information of the object input in step S1, the noise reduction by generating a reduced image data to the pixel value the evaluation value of the foreign matter likeliness of each small area obtained by equally dividing in step S2 At the same time, the likelihood of foreign matter is concentrated, and an image filter that emphasizes the likelihood of foreign matter is applied to the reduced image data in step S3, and threshold processing is applied in step S4 to detect the presence or absence of mixed foreign matter. Here, step S3 may be omitted when foreign matter enhancement is not necessary. Further, steps S2 to S4 may be repeated by changing the size of the small area equally divided in step S2 as necessary.
[0010]
For example, when detecting the presence / absence of a metal foreign substance smaller than approximately N × N pixels mixed in food, the X-ray absorption image data of M × M pixels is equally divided into small regions of N × N pixels in step S2 of FIG. The X-ray absorption image data is reduced to 1 / N scale by setting the maximum luminance for each small area to a new pixel value, and newly reduced image data of (M / N) × (M / N) pixels. Is generated. If there is local contamination of small metallic foreign matter in food, it is observed as local high brightness on the X-ray absorption image data, which is observed on the X-ray absorption image data of food without contamination. Brighter and sharper than brightness. Therefore, by collecting the maximum luminance for each small area, it is possible to cut out fine noise in the small area and concentrate the foreignness. In addition, if necessary, you can detect parts that are much brighter than foreign objects, such as sausage aluminum clips, and cut their surroundings, or cut noise images in low-luminance parts such as packaging materials uniformly. Also good.
[0011]
Now, in step S4, an image filter that emphasizes the likelihood of foreign matter is applied to the reduced image data of the (M / N) × (M / N) pixels in which noise is reduced and the likelihood of foreign matter is concentrated, and in step S4. Now, let us move on to the description of the superiority when detecting the presence or absence of mixed foreign matter by applying threshold processing. First, in step S3, since the foreign substance likelihood is already concentrated in step S2 so that the foreign substance information is about 1 to 2 pixels, the larger the N, the smaller the number of pixels of the reduced image data, and the more likely the foreign substance is. The kernel size of the image filter for emphasizing is sufficient with about 3 × 3 to 5 × 5 pixels, and the load of image processing is lightened. For this reason, it is easy to apply an image filter that requires a high image processing capability, such as a top hat filter, or a combination of various image filters having different characteristics, and it is possible to realize foreign object detection with higher sensitivity than before. Next, in step S4, since the foreign substance likelihood has already been concentrated in step S2, the number of pixels of the reduced image data decreases as N increases, and the load of image processing decreases. For this reason, foreign object determination considering food characteristics and foreign object determination using a plurality of threshold values are easy, and more robust foreign object detection can be realized.
[0012]
Now, let us move on to the explanation of superiority when increasing the spatial resolution of foreign object detection or expanding the measurement range. First, consider the case where the spatial resolution of foreign object detection is doubled in the above example. In this case, the presence / absence of a metal foreign object smaller than 2N × 2N pixels is detected from the X-ray absorption image data of 2M × 2M pixels, and the reduced image data generated in step S2 is (2M / 2N) × (2M / 2N) Same as pixel. That is, if the size of the foreign matter to be detected is the same, the image processing loads in steps S3 and S4 are the same even if the spatial resolution is increased.
Next, consider the case where the measurement range of foreign object detection is doubled in the above example. In this case, the presence / absence of a metal foreign object smaller than N × N pixels is detected from 2M × 2M pixel X-ray absorption image data, and the reduced image data generated in step S2 is (2M / N) × (2M / N) Increase to pixels. Nevertheless, compared with the conventional detection method that detects foreign matter from X-ray absorption image data of M × M pixels with an N × N pixel kernel, the measurement range of foreign matter detection is doubled. It can be seen that the load of image processing is reduced. If it is changed so as to detect even a metal foreign substance that is twice as large as this, the reduced image data generated in step S2 is (2M / 2N) × (2M / 2N) pixels, and the image processing load is further reduced. . In other words, when detecting the presence / absence of a metal foreign object smaller than 2N × 2N pixels, the conventional detection method requires a kernel of 2N × 2N pixels, and the load of image processing is heavy, whereas the detection method of the present invention is reverse. In addition, the image processing load is reduced.
From the above consideration, it has been shown that the detection method of the present invention is extremely effective for detecting a foreign substance having a large pixel size. According to the detection method of the present invention, a larger foreign object can be detected in a shorter time. Therefore, if N is increased in order to detect a foreign object having a large pixel size, the detection sensitivity of the small foreign object is reduced. In order to detect in a short time and to detect the next small foreign substance with high sensitivity, N may be reduced and steps S2 to S4 in FIG. 1 may be repeated.
[0013]
FIG. 2 is a block diagram for explaining a foreign matter detection apparatus using X-rays, which is an embodiment of the foreign matter detection apparatus of the present invention. That is, a conveying means 1 that conveys a measurement object (inspection object) 3, an X-ray source 2 that irradiates the measurement object 3 with X-rays, and an X-ray transmission image of the measurement object 3 that is converted into digital data. It is a foreign object detection device including a line detector 4, an image input unit 5 that takes this into a memory as transmission image data, and an image processing unit 6 that processes the transmission image data and determines the presence or absence of a foreign object. . You may add the image display means 7 as needed.
[0014]
The transport means 1 is realized by, for example, a belt conveyor that transmits X-rays well, and transports the object to be measured 3 between the X-ray source 2 and the X-ray detector 4 that are arranged to face each other. The X-rays irradiated from the X-ray source 2 are absorbed by the object to be measured 3 and slightly absorbed by the belt conveyor, pass through them, and then reach the X-ray detector 4.
[0015]
The X-ray detector 4 is realized by, for example, an X-ray line sensor, and converts an X-ray transmission image of the inspection object 3 into digital data. This digital data is sampled in the transport direction by the image input means 5 at a sampling pitch substantially equal to the sampling pitch on one line by the X-ray line sensor, and is taken into the memory as X-ray transmission image data.
[0016]
Assuming that the X-ray absorption rate of the substance is α and the thickness of the substance is L, the intensity S ′ after the X-ray of the intensity S is transmitted through the substance is theoretically S ′ = S · exp (−α · L ). If transformation is performed by taking the logarithm of both sides, it can be written as α · L = log (S) −log (S ′). The X-ray transmission image data corresponds to a two-dimensional distribution of S ′, and can be converted into X-ray absorption image data indicating a two-dimensional distribution of the absorption amount α · L due to a substance by modifying the logarithm as described above. . Both the X-ray transmission image data and the X-ray absorption image data contain physical property information of the object to be inspected, that is, the amount of absorption α · L by the substance, but the luminance value shows the amount of absorption by the substance in a straight line. Image data is more advantageous for emphasis and detection of foreign matter having a higher X-ray absorption rate. In this case, for example, in the X-ray absorption image data of food, the likelihood of a foreign object can be evaluated by treating a part that exhibits high brightness locally or a sharp edge part as a foreign object candidate point.
[0017]
The image input means 5 logarithmically converts the X-ray transmission image data to X-ray absorption image data as necessary, and outputs the image data to the image processing means 6.
[0018]
The image processing means 6 is a CPU or the like having a parameter setting function, which is implemented with the foreign matter detection method of the present invention described with reference to FIG. 1, and receives the input of the image data output from the image input means 5 and sets parameters in advance. The presence / absence of a foreign object is detected by the processed image processing. That is, generating the reduced image data to the image data including the property information of the specimen 3 that is input in step S1 of FIG. 1, the pixel value of evaluation values of the foreign matter likeliness of each small area obtained by equally dividing in step S2 In this way, the image reducing means 61 that reduces noise and concentrates the likelihood of foreign matter, and the foreign matter enhancement means 62 that generates foreign matter-emphasized image data by applying an image filter that emphasizes the likelihood of foreign matter in step S3 to the reduced image data. And threshold value processing means 63 for detecting the presence or absence of mixed foreign matter by applying threshold value processing to the foreign matter emphasized image data in step S4. Here, the foreign matter emphasizing means 62 in step S3 can be omitted if not necessary. In this case, in step S4, threshold processing means 63 for detecting the presence or absence of mixed foreign matters by applying threshold processing to the reduced image data in step S2. become. Further, by changing the small regions of a size equal division in step S2 if desired may be configured to repeat the steps S2 to S4. When the image display means 7 is added, the image data input in step S1 is output to the image display means 7 together with the reduced image data subjected to the threshold processing in step S4. Further, the logarithmic conversion from the X-ray transmission image data to the X-ray absorption image data performed by the image input means 5 may be substituted for Step S1.
[0019]
Here, an embodiment of the image processing means 6 in the case where the likelihood of a foreign object is evaluated by luminance will be described. For example, when detecting the presence or absence of a foreign object smaller than approximately 8 × 8 pixels, such as bone fragments experimentally mixed with sliced ham or stainless steel scraps, the thickness of the inspected object 3 measured by the amount of X-ray absorption as shown in FIG. X-ray absorption image data that is a saddle shape is input in step S1 of FIG. In step S2, this is equally divided into small areas of 8 × 8 pixels, and reduced image data is generated at a reduced scale of 1/8 with the maximum luminance of each small area as the pixel value, thereby reducing noise and concentrating the likelihood of foreign matter. To do. This reduced image data is shown in FIG. In step S3, a top hat filter having a kernel size of 3 × 3 pixels is applied to the reduced image data in FIG. 4 to generate foreign matter emphasized image data. This foreign matter emphasized image data is shown in FIG. In step S4, threshold value processing is applied to the foreign substance emphasized image data in FIG. 5 to detect the presence or absence of mixed foreign substances. The result of the threshold processing is shown in FIG.
[0020]
The threshold value process in step S4 is not only a process for determining a pixel having a luminance equal to or higher than the threshold value as a foreign object, but for example, if a luminance difference between adjacent 8 pixels is equal to or greater than a threshold value, it is determined as a foreign object. Various processes can be selected by parameter setting. Therefore, highly sensitive foreign matter detection can be performed on the reduced image data in step S2 in which step S3 is omitted.
[0021]
Further, an embodiment of the image processing means 6 in the case of evaluating the likelihood of a foreign object with an edge will be described using the slice ham of the previous example. The X-ray absorption image data shown in FIG. 3 is equally divided into small areas of 3 × 3 pixels in step S2 of FIG. 1, and reduced image data A at a reduced scale having the maximum luminance for each small area as a pixel value. (First reduced image data) and reduced image data B (second reduced image data) at a scale of 1/3 having the luminance dispersion value for each small region as a pixel value are generated in the middle. Here, the minimum value may be used instead of the maximum value, and the luminance range may be used instead of the luminance dispersion value. As shown in FIG. 7, the reduced image data B has emphasized edge portions such as slice ham, bone fragments, and stainless scrapes, and can function as a foreign substance candidate point extraction mask. That is, a reduced image in which the pixel position showing high brightness in the reduced image data B is set as a foreign object candidate point, and all the pixels except for the pixel at the position of the foreign object candidate point in the reduced image data A are replaced with low brightness to reduce the likelihood of the foreign object. Data C (third reduced image data) can be generated. FIG. 8 shows the reduced image data C finally generated in step S2. As shown in FIG. 8, edge portions such as sliced ham, bone fragments, and stainless steel scraps are extracted as foreign object candidate points in the reduced image data. The brightness is higher than that of its own edge, and it is possible to easily determine a foreign object without emphasizing the likelihood of the foreign object. For this reason, step S3 can be omitted, and the presence or absence of mixed foreign matter is detected by directly applying the threshold processing in step S4.
[0022]
The image display means 7 generates foreign object image data obtained by enlarging the reduced image data subjected to the threshold processing in step S4 of the image processing means 6 and returning it to the original size, and in this step S1 of the image processing means 6 The input image data is ORed and displayed on a CRT or the like. As a result, as shown in FIG. 9, the pixels determined to be a foreign object are greatly enlarged and displayed, and visual confirmation on the CRT is easy.
[0023]
【The invention's effect】
As described above, according to the present invention, the degree to which the size of foreign matter that can be detected is limited by the image processing capability is reduced, and it is relatively easy to widen the measurement range of foreign matter detection and increase the spatial resolution. become. Further, a large foreign object can be detected in a very short time, and the remaining time can be reduced to a processing time necessary for realizing foreign object detection with higher sensitivity and robustness than in the past.
[Brief description of the drawings]
FIG. 1 is a flowchart for explaining a foreign object detection method of the present invention.
FIG. 2 is a diagram for explaining an embodiment of the foreign object detection device of the present invention.
FIG. 3 is a diagram illustrating X-ray absorption image data.
FIG. 4 is a diagram illustrating reduced image data.
FIG. 5 is a diagram illustrating foreign object emphasized image data.
FIG. 6 is a diagram illustrating a result of threshold processing.
FIG. 7 is a diagram illustrating reduced image data B with edge enhancement.
FIG. 8 is a diagram illustrating reduced image data C subjected to edge concentration.
FIG. 9 is a diagram illustrating a display screen of foreign object image data.
FIG. 10 is a flowchart for explaining a conventional foreign object detection method;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Conveyance means 2 X-ray source 3 Measured object 4 X-ray detector 5 Image input means 6 Image processing means 7 Image display means 61 Image reduction means 62 Foreign matter emphasis means 63 Threshold processing means

Claims (4)

In the foreign matter detection method for detecting the presence or absence of mixed foreign matter by processing image data including physical property information of the inspection object,
The image data is equally divided into small regions Do that a plurality of pixels to evaluate the local high luminance portion or edge portions in the foreign matter likelihood for each said small region, likeness foreign substances obtained for each the small region of the evaluation value A step (S2) of newly generating reduced image data having a pixel value as a pixel value, and an image that emphasizes a local high-intensity portion or an edge portion of the reduced data for evaluating the likelihood of a foreign object with respect to the reduced image data A foreign matter detection method comprising a step (S3, S4) of detecting the presence or absence of mixed foreign matter by applying a filter and threshold processing. Detecting the presence or absence of the mixed foreign substance and step (S2) for generating a pre-Symbol reduced image data (S3), according to claim 1, wherein the executing a plurality of times while changing the size of the small region Foreign object detection method. The step of generating the reduced image data includes: first reduced image data having a pixel value as an evaluation value of the first foreign object likelihood obtained for each small area by evaluating the first foreign object likelihood for each small area; Generating a second reduced image data having a pixel value as an evaluation value of the second foreign substance likelihood obtained for each small area by evaluating the second foreign substance likelihood for each small area; The pixel of the first reduced image data excluding the pixel position indicating the high luminance of the reduced image data is replaced with a low luminance so as to function as a foreign substance candidate point extraction mask, and the first reduced image data is converted into the first reduced image data. Generating a third reduced image data masked with two reduced image data, wherein the third reduced image data is used as the reduced image data. 2. The foreign matter detection method according to 2. X-ray irradiation means (2) for irradiating the inspection object (3) with X-rays, an X-ray detector (4) for receiving X-rays transmitted through the inspection object and converting them into digital data, and the X-ray detection An image input means (5) for taking digital data output from the container as image data, and detecting the presence or absence of foreign matter mixed in the inspection object by performing image processing on the image data fetched by the image input means In the foreign object detection device having the image processing means (6) for
Wherein the image processing means, the image data is equally divided into small regions Do that a plurality of pixels to evaluate the local high luminance portion or edge portions in the foreign matter likelihood for each said small region, obtained for each small-area and the reduced image data of the evaluation value as the pixel value of the foreign substance likeliness comprise image reduction means for newly generated (61), a local of said reduced data to evaluate the foreign substances likelihood for the reduced image data A foreign matter detection apparatus for detecting the presence or absence of the foreign matter by applying an image filter for emphasizing a high brightness portion or an edge portion and threshold processing .

Images