JP2007033306A - System and method for measuring fluid flow - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system and method for measuring fluid flow allowing further reliable, accurate, and precise measurement, in flow measurement using a particle trace tracking method, especially three-dimensional flow measurement. <P>SOLUTION: The system for measuring fluid flow comprises a laser sheet forming optical system 13 for irradiating the fluid flow field 14 with laser beams oscillated from the oscillator 11, in a sheet shape, image pickup means 29L and 29R for selectively picking up a two-dimensional particle image and a two-dimensional particle trace image formed by radiation by a laser sheet 17, a first image processing means for comparing and analyzing luminance patterns of the two-dimensional particle images at two times different by a fine time and measuring the moving direction and distance of a particle group, and a second image processing means for comparing and analyzing luminance patterns of the two-dimensional particle trace images at two times different by a fine time and measuring the moving directions and distances of individual particles. The second image processing means determines the moving directions of individual particles based on the moving direction of the particle group measured by the first image processing means. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a fluid flow measurement system and a measurement method therefor, and more particularly to a fluid flow measurement system and a measurement method therefor that measure the flow velocity and flow direction of a fluid flowing in a closed space with high accuracy and precision. .

  In recent years, it has been highly desired to grasp a flow field in a closed space that is different from the outside environment, such as in a pressure vessel, from the viewpoint of improving efficiency and safety of a plant or the like. However, measurement by a particle image velocity measurement method (hereinafter referred to as PIV (Particle Imaging Velocimetry)) has been impossible in the past because of the necessity of a reference window and difficulty in installing an optical system or the like. In Non-Patent Document 1, by using a pair of fiberscopes, a stereo PIV system applicable to a closed space or the like was constructed, and a system capable of measuring the speed of a two-dimensional three-way component was constructed. However, in this case, since the image is deteriorated by passing through the fiberscope and the amount of information of particles is extremely reduced, it is necessary to improve the measurement accuracy.

  Under such circumstances, in Patent Document 1, in a PIV system using a fiberscope, an average value of luminance patterns (luminance values) is obtained for each pixel of a particle image captured through the fiberscope, and each particle image is obtained. By subtracting the average value from the luminance pattern (luminance value), the luminance pattern of the fiber array of the optical fiber is removed to remove the noise from the fiberscope image, and the flow state of the fluid, such as the flow velocity and flow direction, is accurately determined. A method for accurate measurement has been proposed.

  Furthermore, in Patent Document 2 and Non-Patent Document 2, particles that use a fiberscope in the particle tracking method (hereinafter referred to as PTV (Particle Tracking Velocimetry)) developed from PIV and record the flow of particles. A trajectory tracking method (hereinafter referred to as PTV-SS (PTV with Streak Searching)) has been proposed. At that time, the noise removal of the fiberscope image similar to Patent Document 1 is performed.

  Further, Non-Patent Document 3 proposes a fiberscope stereo PTV-SS in which a PTV-SS using a fiberscope is expanded in three dimensions.

Furthermore, Non-Patent Document 4 proposes a dynamic binarization method that individually determines a threshold for detecting the track of each particle according to its background in order to detect the track of each particle in PTV. Yes.
JP 2002-22759 A JP 2003-84005 A Proceedings of the Japan Society of Mechanical Engineers Thermal Engineering, No. 02-22 (2002), pp. 7-8 Experiments in Fluids, 33 (2002), pp. 752-758 Proceedings of the Japan Society of Mechanical Engineers Thermal Conference 2003 [2003-11.15-16, Kanazawa] pp. 273-274 Meas. Sci. Technol. , Vol. 11 (2000), pp. 603-616

  The present invention has been made in view of such a situation in the prior art, and the purpose thereof is more reliable, more accurate and precise in flow measurement using the particle trajectory tracking method, particularly in three-dimensional flow measurement. It is an object to provide a fluid flow measurement system and a measurement method thereof capable of performing accurate measurement.

  The fluid flow measurement system of the present invention that achieves the above object includes a laser oscillation device that oscillates laser light, a laser sheet forming optical system that oscillates the oscillated laser light into a fluid flow field, and a laser sheet forming optical system. An image imaging unit that selectively captures a two-dimensional particle image and a two-dimensional particle trajectory image that are formed by illumination with a laser sheet from a laser sheet forming optical system, and a small time difference 2 that is captured by the image imaging unit. The first image processing means for measuring the moving direction and the moving amount of the particle group by comparing / analyzing the luminance patterns of the two-dimensional particle images at the time, and the two-dimensional particles at the two times different from each other taken by the image imaging means. A second image processing unit that compares and analyzes the luminance pattern of the trajectory image to measure the movement direction and the movement amount of each particle, and the second image processing unit includes: Based on the moving direction of the measured particles by the first image processing means is characterized in that it is configured to define the direction of movement of the individual particles.

  In this case, it is preferable that the laser oscillation device is constituted by a semiconductor laser, and a two-dimensional particle image and a two-dimensional particle locus image are formed based on a difference in emission time of the semiconductor laser.

  The image forming apparatus further includes two image capturing units that selectively capture a two-dimensional particle image and a two-dimensional particle trajectory image simultaneously from both sides of the laser sheet from the laser sheet forming optical system. Each of the second image processing means includes the first image processing means and the second image processing means, and each of the second image processing means individually based on the moving direction of the particle group measured by each of the first image processing means. The moving direction of the particles can be determined.

  In that case, a three-dimensional reconstruction in which the movement direction and the movement amount of each particle measured by each of the second image processing means are associated with each other to reconstruct the three-dimensional movement direction and movement amount of each particle. A means for measuring the three-dimensional moving direction and moving amount of each particle.

  The image pickup means includes an image transmission means for optically transmitting a two-dimensional particle image and a two-dimensional particle trajectory image formed by illumination with a laser sheet from the laser sheet forming optical system; It can comprise as an imaging means which images a two-dimensional particle trajectory image and a two-dimensional particle image.

  In that case, the image transmission means is composed of a flexible image guide in which a large number of optical fibers are bundled and integrated, and both end faces are processed flat, and the imaging means is composed of a CCD camera. The image guide can be configured to disassemble an image formed on one end face of the fiber by the objective lens into each pixel using an optical fiber and transmit the same image to the other end face on the CCD camera side.

  In the above description, it is desirable that the laser sheet forming optical system and the image capturing unit are integrally combined.

  The first image processing means and the second image processing means preferably include fiber noise removing means for removing fiber noise based on the image guide from a two-dimensional particle image and a two-dimensional particle trajectory image, respectively.

  In addition, it is desirable to include subpixel processing means for subpixel accuracy in the movement direction and movement amount of each particle measured by the second image processing means.

  The fluid flow measuring method of the present invention is a two-dimensional particle image formed by irradiating a laser beam oscillated from a laser oscillation device into a fluid flow field in a sheet form to form a laser sheet and being illuminated by the laser sheet. And two-dimensional particle trajectory images are selectively picked up by the image pickup means, and the luminance patterns of the two-dimensional particle images at two times different from each other are compared and analyzed to measure the moving direction and moving amount of the particle group. Also, by comparing and analyzing the luminance patterns of the two-dimensional particle trajectory images at two times that are captured for a very short time, the movement direction and movement amount of each particle are measured, and the movement direction and movement amount of each individual particle are measured. In the measurement, the flow of the fluid in the fluid flow field is measured by determining the movement direction of each particle based on the movement direction of the particle group previously measured.

  According to the fluid flow measurement system and the measurement method according to the present invention, a two-dimensional particle image and a two-dimensional particle formed by irradiating the fluid flow field in a sheet shape to form a laser sheet and illuminating with the laser sheet. The trajectory image is selectively captured, the luminance patterns of the two-dimensional particle images at two times that are captured at different times are compared and analyzed, and the movement direction and amount of the particle group are measured. Compare and analyze the luminance patterns of two-dimensional particle trajectory images at two different times to measure the movement direction and movement amount of each particle, and measure the movement direction and movement amount of each individual particle first. By determining the moving direction of each particle based on the moving direction of the particles, the flow state of the fluid such as the flow rate and flow direction of the fluid is measured by measuring the flow velocity and flow direction of each particle in a complex flow field. , More reliable, higher accuracy can be more precisely measured.

  In addition, according to the present invention, the flow state of the fluid in a complex flow field can be measured by processing the particle image and the particle trajectory image with the image processing means, so the flow state of the fluid can be measured from a remote location. It is possible to measure the flow state of the fluid in the environmentally severe closed space with high reliability, high accuracy, and high precision.

  Hereinafter, a fluid flow measurement system and a measurement method thereof according to the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing an outline of one embodiment of a fluid flow measurement system of the present invention.

  The fluid flow measurement system 10 of the present invention includes a laser oscillation device 11 that oscillates a high-power laser beam as a light source. As the laser oscillation device 11, for example, a semiconductor laser that is small in size, can easily control the timing of light emission, and has high luminance and high repetition frequency is used.

  The laser light oscillated from the laser oscillation device 11 is guided to the laser sheet forming optical system 13 through the light transmission optical fiber 12. The laser sheet forming optical system 13 projects the laser beam oscillated from the laser oscillation device 11 onto the fluid flow field 14 in a sheet shape.

  The fluid flow field 14 is formed in a closed space having a different environment from the outside. Further, according to circumstances, the fluid flow field 14 may be formed in a form open to the outside. The flow field 14 is made of at least one of a gas and a liquid. Specifically, for example, the velocity field of fluid in a closed space such as a downcomer portion of a reactor pressure vessel, a core shroud, a heat exchanger or a steam generator of a thermal power plant is targeted. In addition, a fluid flow field around a structural part or equipment that comes into contact with a water flow or an air current, such as a ship, an aircraft, or an automobile, can also be targeted.

  The laser sheet forming optical system 13 forms a laser sheet 17 in the fluid flow field 14 by irradiating the fluid flow field 14 with the oscillation laser light from the laser oscillation device 11 in a sheet shape. The laser sheet 17 illuminates and visualizes fine tracer particles mixed in the fluid flow field 14.

  The laser sheet forming optical system 13 is a laser for producing the laser sheet 17 so that a cross section along the sheet surface of the laser sheet 17 is shown in FIG. 2A and a cross section perpendicular to it is shown in FIG. This is a light irradiation optical system. For example, when the semiconductor laser as described above is used as the laser oscillation device 11, the oscillation laser light from the semiconductor laser is emitted from the light transmission optical fiber 12 with a large divergence angle. The emitted laser beam is not axisymmetric light, has an astigmatism difference, and the shape conversion of the laser beam becomes complicated.

  For this reason, in the laser sheet forming optical system 13, a lens group 18 is used as a laser sheet forming optical means for forming the laser sheet 17 from the laser beam emitted from the light transmitting optical fiber 12. The lens group 18 is configured to collimate the laser beam emitted from the light transmitting optical fiber 12 with, for example, three plano-convex lenses 19a, 19b, and 19c, and to create the laser sheet 17 with the plano-concave cylindrical lens 19d. Yes. The thickness of the laser sheet 17 output by the laser sheet forming optical system 13 is, for example, 3.0 mm.

  The laser sheet 17 is irradiated from the laser sheet forming optical system 13 to the fluid flow field 14, and the fluid flow field 14 is visualized in a sheet shape. Then, the objective lenses 27a of the left and right image transmission means 22L and 22R are arranged facing the measurement range 17a of the laser sheet 17 from both sides, and the left and right particle images formed through the objective lenses 27a are Are sent to the left and right image capturing means 29L and 29R via the image transmission means 22L and 22R, and imaged there. In the flow measurement system of this embodiment, as described below, a fiberscope including an image guide 26 and an objective lens 27a is used as the image transmission means 22L and 22R.

  Here, as shown in the perspective view of FIG. 3A and the cross-sectional view of FIG. 3B, the probe portions at the tips of the left and right image transmission means 22L and 22R sandwich the laser sheet forming optical system 13 therebetween. Symmetrically, the front ends of the image guides 26L and 22R of the image transmission means 22L and 22R are integrally arranged, and the optical path is bent at the front end of each image guide 26 by 90 ° so as to face both sides of the measurement range 17a of the laser sheet 17. A 45 ° right-angle prism 27b and an objective lens 27a for capturing a particle image in the measurement range 17a disposed on the incident side are integrally disposed.

  The left and right particle images in the measurement range 17a on the laser sheet 17 transmitted by the left and right image transmission units 22L and 22R can be captured by the left and right image capturing units 29L and 29R.

  The left and right image capturing units 29L and 29R include CCD cameras 24L and 24R each having a camera lens 28 as an image capturing unit that captures the particle images transmitted by the left and right image transmitting units 22L and 22R, respectively. As shown in FIG. 4, the image transmission means 22 (L and R that distinguish left and right are omitted) is a flexible structure in which thousands to tens of thousands of optical fibers 25 are bundled and integrated (melted). An image guide 26 is used. As shown in FIG. 4, the image guides 26 are bundled side by side so that the positions on both end faces of the bundled optical fibers 25 correspond exactly. Both end surfaces of the image guide 26 are finished to be flat, and the image formed on one end surface of the optical fiber by the objective lens 27a is decomposed into optical fibers 25 as respective pixel optical fibers, and the same image is light on the CCD camera 24 side. Transmission to the end face of the fiber (in FIG. 4, the 45 ° right-angle prism 27b is omitted). The image is picked up by the CCD camera 24 through the camera lens 28 from the end face of the optical fiber. The image capturing means 29L and 29R may be configured to directly capture the particle image on the laser sheet 17 without including the image guide as described above.

  The limit (resolution) of the image of the image guide 24 is determined by the diameter of the single optical fiber 25 and the arrangement method, and the appearance of the image is determined by the total number of optical fibers 25 bundled in the image guide 26 and the arrangement. The light distribution having a light quantity smaller than the core diameter of the optical fiber 25 is averaged while traveling through the optical fiber 25, and the brightness is uniform within each core cross section. Accordingly, in the image guide 26, the core diameter of the optical fiber 25 is the minimum unit of spatial resolution as a pixel.

  In the actual image guide 26, a large number of optical fibers 25 are arranged in an ordered and dense manner in a stacked (hexagonal dense arrangement) structure. When a lattice-like object having an interval substantially equal to the optical fiber interval is observed with the image guide 26, moire interference fringes unrelated to the intensity distribution of the image of the object are generated.

  Further, in the image guide 26 in which the clad portion is fused between the optical fibers 25, when the thickness of the clad becomes thin, the light entering the core of one optical fiber 25 may leak to the adjacent core. . For this reason, the boundary of the image is blurred and the contrast of the image is lowered. For this reason, when the hexagonal close-packed structure is adopted for the image guide 26, it is a condition that the thickness of the clad is not too thin.

  On the other hand, the fluid flow measurement system 10 of the present invention includes a signal generator 30, and the TTL signal can be arbitrarily generated by controlling the signal generator 30 by software in the personal computer 31. Is transmitted to both the laser driver 32 of the laser oscillation device 11 and the CCD cameras 24L and 24R, the frames of the CCD cameras 24L and 24R are switched, and the laser irradiation time and irradiation interval of the laser oscillation device 11 can be arbitrarily adjusted. Yes. A Peltier driver 33 as a semiconductor laser cooling device of the laser oscillation device 11 is disposed between the laser driver 32 and the laser oscillation device 11, and the laser oscillation device is passed from the laser driver 32 via the Peltier driver 33. 11 is driven.

  The CCD cameras 24L and 24R are configured to transfer the left and right particle images captured via the left and right interfaces 34L and 34R to the personal computer 31, respectively. Then, the particle image captured by the personal computer 31 is immediately converted into three-dimensional velocity vector data using the particle image velocity measurement method (PIV) and the particle trajectory tracking method (PTV-SS) by image processing software described below. It is possible to convert to.

  In the flow measurement system as described above, the three-dimensional velocity vector information of the fluid flow field 14 is acquired with high reliability, high accuracy, and high precision by image processing software as shown in FIG. In the following, the process flow of FIG. 5 will be described in order.

  The following processing is performed in parallel for each of the left and right camera images 41L and 41R captured by the left and right CCD cameras 24L and 24R at the same time and sent to the personal computer 31 through the interfaces 34L and 34R. In FIG. 5, the PIV processing 42L, the PTV-SS processing 43L, and the subpixel processing 44L for image degradation due to the optical fiber bundle are illustrated only for the left camera image 41L, and the right camera image 41R is not illustrated. The same applies to the right camera image 41R.

  First, the PIV process 42L is a process for performing a normal PIV process after removing the arrangement pattern of the optical fibers 25 of the image guide 26 of the fiberscope, as in the method proposed in Patent Document 1. The vector calculated by this process becomes auxiliary vector information at the time of the process in the PTV-SS process 43L.

  The PTV-SS process 43L is a process after the PIV process 42L, and uses the information on the flow velocity vector (PIV vector) calculated by the PIV process 42L performed as a pre-process in addition to the conventional PTV-SS method. This makes it possible to associate particles with high accuracy.

  First, the PIV process 42L will be described. In the PIV processing, the obtained particle image is regarded as a luminance distribution, and the amount of movement of the luminance pattern of the particle group within a minute time is quantitatively analyzed. For this reason, a cross-correlation method is often used for analysis of the amount of particle movement (cross-correlation PIV processing 422). However, since the camera image 41L acquired by the CCD camera 24L is captured through the image guide 26 of the fiberscope, not only the luminance pattern of the particle group but also the luminance pattern of the array of the optical fibers 25 exists. For this reason, the arrangement of the optical fibers is detrimental in detecting the peak of the correlation value of the luminance pattern of the particle group, and as a result, erroneous correspondence (error vector) of the particle group is caused. An example of the erroneous correspondence is shown in FIG. FIG. 6 clearly shows that the arrangement pattern of the optical fiber is detected as the correlation value peak, not the particle group.

  As described above, the arrangement of the optical fibers 25 recorded in the camera image 41L becomes noise with respect to the luminance pattern of the particle group and causes an error vector. Therefore, in the PIV process 42L, the fiber noise removal process 421 is performed before the cross-correlation PIV process 422. The fiber noise removal process 421 is performed as follows.

The camera image 41L captured by the CCD camera 24L through the image guide 26 of the fiberscope is processed, and the particle image obtained in time series is obtained for each pixel to obtain an average value of luminance patterns (luminance values), and each particle is obtained. The luminance pattern of the arrangement of the optical fibers 25 is removed by subtracting the addition average value from the luminance pattern (luminance value) of the image. The method will be described. If the measurement data (luminance value in the particle image (camera image) 41L) is I,
I = s + n (1)
It is represented by Here, s means a signal component (luminance pattern of optical fiber array), and n means a noise component (particle luminance pattern).

  That is, it is considered that the signal (luminance pattern of the optical fiber array) and noise (particle luminance pattern) are measured. Addition averaging is an effective technique when the SN ratio is small and there is no significant difference between the signal and noise constituent frequencies, and when measurement can be repeated many times under the same conditions. When Expression (1) is expressed as a time series, it can be expressed as follows.

I ⁱ (k) = s ⁱ (k) + n ⁱ (k) (2)
Here, i represents the i-th image, and k represents the k-th pixel luminance value of the i-th image.

  In the capturing of particle images, if the number of captured particle images is M, the addition average I (k) for the M particle images can be expressed as follows.

_M
I (k) = (1 / M) Σ I ⁱ (k)
^{i = 1}
_{M M}
= (1 / M) Σs ⁱ (k) + Σ n ⁱ (k) (3)
^{i = 1 i = 1}
The luminance pattern s of the optical fiber array has the same condition (light transmission system / light reception system), and the same pattern appears in successive particle images. Is averaged (divided by M), the original luminance value remains unchanged. That is, an image obtained by taking only the optical fiber array in which no particles exist can be obtained by the averaging operation, and the luminance value (I (k)) of the optical fiber array is obtained for each pixel from the particle image 41L in which the optical fiber array is present. If the luminance value for each pixel is subtracted, the luminance pattern of the optical fiber array can be removed.

  FIG. 7 shows a flow velocity vector (PIV vector) distribution obtained by performing cross-correlation calculation in the next PIV process 422 on the measurement data subjected to the fiber noise removal process 421 in the same example as FIG. Compared with FIG. 6, it can be clearly seen that the error vector is well removed by performing the fiber noise removal processing 421.

  FIG. 8 is a diagram showing the flow of image processing in which the luminance pattern of the optical fiber array is removed by the fiber noise removal processing 421 and only the particle image remains, as a luminance value distribution of an arbitrary line.

  The particle image from which the luminance pattern of the optical fiber array has been removed by the fiber noise removal processing 421 is subjected to particle image flow velocity measurement processing based on the cross-correlation method in the next cross-correlation PIV processing 422. A particle image obtained by digitally processing a two-dimensional particle image (fiber noise is removed in process 421) taken at the first time taken by the CCD camera 24L is a “reference image”, and the first time is a minute time. A digitally processed particle image of a two-dimensional particle image at a different second time (similarly, fiber noise is removed in the process 421) is a “search image”, and a limited rectangular area in the reference image is displayed. A “reference window image” and a limited rectangular area in the search image are defined as a “search window image”. A value of one point in the digitized particle image is defined as “luminance value”, and the luminance value distributed in a rectangular area is referred to as “luminance pattern”.

  The “reference window image” and the “search window image” represent a rectangular distribution of luminance values and indicate luminance patterns.

  The algorithm for estimating the movement amount of the particle group in the fluid from the particle image is determined to be a change in the luminance distribution of the obtained particle image, and the luminance pattern of the particle group on the laser sheet 17 moves within a minute time. The quantity is analyzed quantitatively.

  A cross-correlation method is used for the analysis of the movement amount of the particle group, and the analysis method of the cross-correlation method is also used in this processing 422.

  It is known that the movement amount estimation of the particle group by the cross-correlation method has a correlation value R expressed by the following equation between the “search window image” and the “reference window image”. This is a technique for obtaining the degree of similarity between patterns and obtaining a moving amount of a particle group between two images by comparison and examination.

As shown in the equations (4) and (5), the correlation value R is divided by the mean square of the luminance values of the search window image and the reference window image (normalized), thereby obtaining values from −1 to 1. Take. When the correlation value R is 1, the two images are completely matched, and the larger the correlation value R, the greater the similarity between the window images.

In Expressions (4) and (5), I ₁ and I ₂ represent the luminance of each pixel of the reference window image and the search window image, and ξ and η are represented by relative positions of the search window image and the reference window image. . The positions ξ and η at which the correlation value R is maximum correspond to the movement amounts ΔX ′ and ΔY ′ of the particle group in the image. The size of the reference window image and the relative position with respect to the reference window image are determined from the predicted minimum and maximum velocities. The size nxm of the reference window image is smaller than the smallest vortex existing in the flow field in the real coordinate system of the reference window image, and 6 to 8 or more tracer particles in the reference window image It is desirable to include.

  As described above, the flow velocity vector (PIV vector) distribution information as illustrated in FIG. 7 in the measurement range 17a of the laser sheet 17 is obtained through the fiber noise removal process 421 and the cross-correlation PIV process 422. .

  Then, an erroneous vector removal and spatial averaging process 423 is performed as the last process of the PIV process 42L. In this process, in order to remove a vector considered to be an erroneous vector from the vector calculated through the fiber noise elimination process 421 and the cross-correlation PIV process 422, a variation error based on the normal distribution is rejected. Assuming that the vector interval calculated by the cross-correlation PIV processing is N pixels, for all the calculated vectors (u, v), the vector components in the 2N pixels surrounding each vector are, for example, an average value ± 3σ (σ : If it does not fall within the standard deviation), it is rejected as an erroneous vector. Of course, this reference value is not limited to ± 3σ, and other values may be used. Then, spatial average processing is performed by extracting and averaging the remaining vectors having different positions in the vicinity of the spatially regular intersection point coordinate positions to obtain representative values on the intersection point coordinates.

  A vector resulting from these processes is defined as a vector by PIV processing, and this is used as auxiliary vector information when performing the PTV-SS method in the next PTV-SS process 43L.

  In addition, the outline information of the vector of the fluid flow field 14 can be acquired by the above PIV process 42L. However, in the PIV method for estimating the amount of movement of the particle group, it is desirable that there should be at least 6 to 8 particles in the reference window image, so the reference window image must be relatively large. The amount of calculation increases enormously, and the algorithm becomes unrealistic (when the area to be searched increases, the amount of calculation increases in proportion to the fourth power). That is, the acquired image parameters (particle image diameter, number density) are very limited, and it becomes difficult to deal with various actual flow fields only by PIV processing. Therefore, in the PIV process 42L described above, it is necessary to perform analysis in a stance that the reference window size is limited to a small size even if the number of particles is reduced, the accuracy is reduced, and an outline of the vector is acquired.

  In the present invention, the PTV-SS process 43L for improving the measurement accuracy is applied as the next stage process of the PIV process 42L, and one of the major problems that arises at that time is the problem of erroneous correspondence of particles. There is. As shown in FIG. 9, the erroneous correspondence between particles means that the correspondence between particles between images is caused by a decrease in the amount of particle information due to the reduction in resolution of the image by the optical fiber bundle. FIG. 9 is a diagram for explaining erroneous correspondence of particles. In this figure, for easy understanding, a reference image and a search image are displayed on the same image. The correct association result is a position indicated by a large arrow as shown in FIG. 9A. However, due to a lack of information amount due to the above reason, erroneous correspondence of particles as shown by an arrow in FIG. It is likely to happen.

  Therefore, in the present invention, as the left and right camera images 41L and 41R of the PTV-SS process 43L, the emission time of the semiconductor laser used in the laser oscillation device 11 that illuminates the fluid flow field 14 with the laser sheet 17 is relatively long. Using particle streak tracking velocimetry (PTV-SS) (Patent Document 2, Non-Patent Document 2) that estimates the amount of movement of particles individually instead of particles using an image of particle tracks recorded. By applying this, vector information is calculated from an image passing through the image guide 26 of the fiberscope.

  The PTV-SS process 43L is divided into three stages: a fiber noise removal process 431 upon particle detection, a subsequent particle detection process 432, and a PTV-SS method 433 using a PIV vector. These processes will be described below.

  The fiber noise removal processing 431 will be described. In the PIV processing 42L, in the fiber noise removal processing 421, a method of subtracting the addition average image I (k) of the particle image that is the luminance value of the optical fiber array for each pixel from the particle image 41L in which the optical fiber array exists. Was used to remove noise. The same method may be adopted in the fiber noise removal processing 431.

  However, in this process, the information on the change in luminance of the particles, which is a necessary component, is inevitably lost due to the use of the averaging. For this reason, it is conceivable that the number of detected particles is reduced by the particle detection processing 432. In the PIV process 42L, the biggest problem caused by fiber noise is to detect the arrangement pattern of the optical fibers as a correlation value peak in the particle detection, and this problem occurs in the particle detection in the PIV process. Does not occur.

Therefore, in the present embodiment, the fiber noise removal process 431 of the particle image 41L in the PTV process is a pattern of the optical fiber 25 with respect to the particle image 41L so as not to drop luminance component information as much as possible. Spatial averaging is performed as a high frequency filter that removes high frequency components of the image. In performing this processing, the spatial frequency of the image guide 26 is obtained, and the number of pixels of the CCD array of the CCD camera 24L corresponding to the core diameter of the optical fiber 25 is calculated. FIGS. 10A and 10B are examples of luminance value distribution and frequency analysis of an arbitrary line of an image of the image guide 26 actually taken. It is considered that a change in luminance value due to the brightness of the core and clad of the optical fiber 25 is detected as a dominant spatial frequency. In this example, the image guide 26 in which 0.320 pixels- ¹ is photographed by the CCD camera 24L. Spatial frequency. Therefore, in this example, in order to remove the pattern component of the optical fiber 25, a spatial averaging process is performed in a range of 3.124 pixels or less, thereby removing a spatial frequency of 0.320 pixels ^-1 or more and removing the fiber noise. It is carried out. In order to remove the fiber noise, a spatial frequency filtering process for removing a high-frequency spatial frequency by electrical signal processing may be used instead of the spatial averaging process as described above.

  The particle detection process is performed in the next particle detection process 432 from the image obtained by performing the fiber noise removal process from the particle image 41L in which the optical fiber array exists as described above. One of the important problems when calculating vector information using PTV processing is that the particles are specified accurately. Since the captured particle image 41L is represented by brightness gradation, a particle specifying method includes a simple binarization method using a certain luminance threshold and a particle extraction method using a mask operator. However, the simple binarization method has a problem that the number of detected particles is greatly reduced or the partially overlapping particles are recognized as the same particle when the luminance fluctuation between particles is large due to factors such as image noise. There is a point. In addition, the particle extraction method using the mask operator has a problem that the particle shape must match the estimated pattern.

Therefore, in the present invention, a dynamic binarization method (Non-patent Document 4) that can be applied to a case where luminance fluctuation is large is applied by determining the threshold value of the binarization method for each particle. . In the dynamic binarization method, by setting different threshold values for individual particles, the problem that the particle detection accuracy decreases due to variations in the luminance distribution in the entire image is solved. An outline of this dynamic binarization method is shown in FIG. As an initial threshold value, an image is divided (FIG. 12), and an average luminance value I _avg in the region is used. If binarization is performed simply using the average luminance value of the image as the initial threshold value (x, y), there is a problem in that particles are erroneously extracted at the boundary of the region when the threshold values of adjacent regions are different. Set the initial threshold smoothly. As shown in FIG. 12, the initial threshold value in the center coordinates (x ₁ , y ₁ ), (x ₁ , y ₂ ), (x ₂ , y ₂ ), (x ₂ , y ₁ ) of the divided area is the average luminance. _{I avg1, I avg2, I avg3} , as _I avg4, x ₁ <x <in x _2, y ₁ <y <range of y _2, the coordinates (x, y) the initial threshold (x, y) around four points Interpolation is performed from the average luminance. In two dimensions, bilinear interpolation per grid eye, the equation is
i≡ (x−x ₁ ) / (x ₂ −x ₁ )
j≡ (y−y ₁ ) / (y ₂ −y ₁ )
... (6)
And using the parameters represented by i and j,

Θ (x, y) = (1-i) (1-j) I _avg1 + i (1-j) I _avg2
+ IjI _avg3 + (1-i) jI _avg4 (7)
Bilinear interpolation improves the accuracy of particle detection because the value of the interpolation function (x, y) changes continuously when the interpolation point moves from eye to eye of the grid. Thereafter, labeling is performed to determine the area S of the region recognized as a particle of the binarized image. If the value is within the set range (S ₁ <S <S ₂ ), it is regarded as a particle, and the threshold value (x, y) Does not change. If it is smaller than S _{1, the} binarized region is regarded as noise rather than particles, and the threshold (x, y) of the region is increased to the maximum value. Conversely, if it is larger than S ₂ , binarization is performed. The threshold value (x, y) is increased on the assumption that the region thus obtained includes a plurality of particles. The cycle of threshold setting, binarization, labeling, and determination is repeated until the threshold is not changed in the entire screen area.

As will be described later, since the PTV-SS method is applied in the present invention, the particles are imaged in a streak shape, so that the area of each particle is sufficiently larger than the luminance distribution generated by one optical fiber. Therefore, by setting the minimum particle area S ₁ , which is a parameter of the above-described dynamic binarization process, to be larger than the luminance distribution by one optical fiber, it is possible to prevent the optical fiber pattern from being recognized as particles. .

  Through the fiber noise removal processing 431 and the particle detection processing 432 described above, fiber noise, which is an optical fiber pattern, is removed, and binarization is performed by setting different threshold values for individual particles by a dynamic binarization method. Using the particle images (“reference image” and “search image”) 41L at two different times, the movement amount of each particle is obtained in the PTV-SS method 433 using the PIV vector. However, as described above, the particle image 41L used in this case is a particle locus image in which the emission time of the semiconductor laser is relatively long.

  Hereinafter, first, the particle trajectory tracking method (PTV-SS) will be described. At this time, two PTV-SS methods detected by the PTV-SS method using the PIV vector obtained by the PIV processing 42L as auxiliary vector information are described. The point of limiting to one of the possible vectors will be described.

  As described above, the PTV-SS method is a method in which the emission time of the semiconductor laser used as the laser oscillation device 11 of the illumination light source is relatively long and the track of particles is recorded. This is a method for estimating the amount of movement. This is because, by recording a trace, it is possible to provide sufficient particle image information for the single optical fiber 25 having a minimum spatial resolution without taking a particle image diameter larger than necessary. Because. The PTV-SS method is made possible by using a semiconductor laser that can easily control the light emission time of the laser oscillation device 11.

  Here, FIG. 13 shows a comparison between the PTV-SS method used in the present invention and the conventional PTV method. In FIG. 13, (a1) and (a2) are diagrams showing reference images and search images of the respective methods, and (b1) and (b2) are diagrams showing oscillation timings of the semiconductor lasers of the respective methods. Further, FIG. 14 shows a diagram relating only to the particle trajectory portion of the PTV-SS method.

PTV-SS method, as shown in FIG. 13 (a1), by setting the longer the laser irradiation time t _d the particle migration velocity, in the reference image, the length as shown in FIG. 14 in the search image A movement trajectory with an axis and a short axis is recorded. On the other hand, in the PTV method, as shown in FIG. 13 (a2), the laser irradiation time is shorter than the particle moving speed, so that the moving locus is not recorded only by the particle image.

Here, assuming that the laser irradiation time is t _d , the laser irradiation interval is Δt, and the major axis and minor axis length of the movement trajectory are L _p1 and L _p , respectively, the condition is that the particle velocity is constant until the next captured image. Below, the next particle position can be estimated by the following equation (8).

L = (L _p1 −L _p ) Δt / t _d (8)
Here, L is an estimated moving distance to the next photographed image. Since this value can be estimated from the particle trajectory image, the search window image can be set more efficiently than the conventional PTV method using only the cross-correlation as shown in FIGS. 13 (a2) and (b2). The calculation amount is reduced, that is, the calculation speed of the PTV is improved. In addition, as compared with the case of FIG. 13A2 in which a wide search window image is opened, the erroneous correspondence in particle detection is greatly reduced by estimating the movement position of particles. In addition, by calculating the cross-correlation between the “search window image” and the “reference window image” thus set based on the equations (4) and (5), the amount of movement of each particle Is obtained.

  As described above, in the present invention, by applying the PTV-SS method, it is possible to reduce particle miscorrespondence by estimating the particle position of the next time image from the particle trajectory length. . However, with the information only from the particle trajectory, the destination of movement is not uniquely determined, and two vectors that are 180 ° different from each other are candidates for movement, as shown in FIG. Conventionally, the direction information has been processed so as to become a single vector by artificially giving a set value by the flow field, but this is insufficient to cope with various flow fields.

  Therefore, in the present invention, in order to determine this direction information, using the direction of the PIV vector obtained by the PIV process 42L, the information only exists from the particle trajectory as shown in FIG. Limit the possibility of one vector to one vector pointing in the direction of the PIV vector. Note that the white arrow group in the background of FIG. 15B is the PIV vector, the vector marked with a cross is a negated vector, and the vector marked with a circle is affirmed the possibility. The remaining vector.

  In FIG. 16, as an example, a PTV-SS method 433 using the above-described PIV vector is used from a particle trajectory image 41L obtained by imaging a pseudo flow formed by placing a glycerin solution mixed with tracer particles on a rotating disk. The vector map which shows the calculated movement amount detection result is shown.

  As described above, in the PTV-SS method 433 using the PIV vector, the PIV vector obtained by the PIV process 42L is used as an auxiliary, and the particle is automatically set in various flow fields without setting artificial parameters. The moving direction and amount of movement can be measured.

  By the way, the measurement accuracy of the movement amount by the cross-correlation method in the PTV-SS processing 43L is that the images obtained by the CCD cameras 24L and 24R are spatially discretized, and thus the movement amount obtained is an integer value. is there. Therefore, the amount of movement below the integer value cannot be derived by the cross correlation method. Therefore, in such an velocimeter, it is necessary to evaluate the particle movement amount (subpixel movement amount) of one pixel or less, and it is necessary to fully consider this characteristic. In general, a normal distribution function is often used as an approximation function to make the measurement result subpixel accurate. This is because the particle images used for the cross-correlation are the same as the normal distribution function, and therefore their cross-correlation value maps similarly take the form of a Gaussian distribution function. However, in the present invention, since the particle image is captured through the fiberscope, the particle image is deteriorated, and the luminance distribution of the particles does not form the normal distribution function. FIG. 18 is a diagram showing an example of a cross-correlation map in a particle trajectory image when passing through a fiberscope. In the case of a particle image passing through a fiberscope, the correlation value map is greatly influenced by the core and clad of the optical fiber 25, and the frequency component of the optical fiber 25 significantly protrudes as shown in FIG. I understand. For this reason, it is not appropriate to perform approximation using the normal distribution function when calculating the sub-pixel movement amount with respect to the correlation value map of FIG.

  Therefore, in the present invention, the autocorrelation map of the reference image is used as an approximation function for subpixel accuracy of the measurement result in the subpixel processing 44L for image degradation due to the optical fiber bundle following the PTV-SS processing 43L. An example of the autocorrelation map of the reference image is shown in FIG. Then, a quadratic curve approximation is performed on the calculated cross-correlation value map (FIG. 19) to calculate the final subpixel movement amount.

  Here, the sub-pixel movement amount calculation procedure for one particle is summarized as follows.

  1. An autocorrelation value map in which the autocorrelation values of the reference images are arranged is created (FIG. 17).

  2. A cross-correlation value map in which the cross-correlation values of the reference image and the search image are arranged is created (FIG. 18).

  3. In order to obtain the maximum point with sub-pixel accuracy in the cross-correlation value map (FIG. 18), the autocorrelation value map (FIG. 17) is used as an approximation function.

  4). A cross-correlation value map (FIG. 19) in which the cross-correlation values of the autocorrelation value map (FIG. 17) and the cross-correlation value map (FIG. 18) are arranged is created.

  5. A quadratic curve approximation is performed on the calculated cross-correlation value map (FIG. 19) to calculate the final subpixel movement amount.

  FIG. 20 shows an example in which quadratic curve approximation is performed on a cross-correlation value map obtained from the cross-correlation values of the autocorrelation value map and the cross-correlation value map (however, this example is a quadratic curve approximation in a one-dimensional direction). .)

  By following the above procedure in the sub-pixel processing 44L for image degradation due to optical fiber bundles, it is possible to set an approximate function corresponding to each particle even when there is a difference in particle image due to particle diameter or particle velocity. become. In addition, this makes it possible to significantly reduce the influence of the resolution reduction by the optical fiber and the image loss due to the clad on the subpixel movement amount calculation.

  The above PIV processing 42L (fiber noise removal processing 421, cross-correlation PIV processing 422, erroneous vector removal and spatial averaging processing 423), PTV-SS processing 43L (fiber noise removal processing 431, particle detection processing 432, PIV vector Using the PTV-SS method 433) and the sub-pixel processing 44L for image degradation due to the optical fiber bundle, the next final vector data processing 45L is calculated from the left camera image 41L captured by the left CCD camera 24L. Two-dimensional velocity vector data information is obtained.

  Also for the right camera image 41R captured by the right CCD camera 24R, PIV processing 42L (fiber noise removal processing 421, cross-correlation PIV processing 422, erroneous vector removal and spatial averaging processing 423), PTV-SS processing 43L ( Fiber noise removal processing 431, particle detection processing 432, PTV-SS method 433 using PIV vectors, and processing similar to sub-pixel processing 44L for image degradation due to optical fiber bundles are performed, and the following final vector data processing Two-dimensional velocity vector data information is obtained at 45R.

  Each of the two-dimensional velocity vector data information obtained for the left and right camera images 41L and 41R may be obtained as a flow measurement result based on the present invention. In the above embodiment, a pair of two-dimensional images having left and right parallaxes are used. Since velocity vector data information is obtained, a flow field having three-dimensionality is obtained by obtaining three-dimensional velocity vector data information from the pair of two-dimensional velocity vector data information by the next three-dimensional reconstruction processing 46. 3D flow measurement becomes possible. Hereinafter, the three-dimensional reconstruction process 46 will be described.

  What is used here is a method called stereo PTV, which uses two imaging systems (CCD cameras 24L, 24R and their optical systems 27a, 27b, 28, etc.) and three-dimensional information by calibration, thereby speeding up. It is possible to measure the third direction component. That is, this is a method of measuring the direction component existing in the direction penetrating the region illuminated by the laser sheet 17 and simultaneously measuring the two-dimensional distribution of the three-direction component velocity vectors. Then, by associating the actual coordinates by the mapping function with the coordinates in the camera, it is possible to ignore the influence of refraction due to the optical window and the working fluid, and therefore, the curved optical window can be measured. It is also possible to measure a system such as a flow channel that has been difficult to measure.

  As shown in FIG. 21, the stereo PTV performs the calculation of the velocity in the real coordinate system by associating the real coordinates with the in-camera coordinates of the two cameras (hereinafter referred to as pixel coordinates) by calibration. Is what you do. Since the correspondence between the real coordinates and the pixel coordinates is accompanied by lens aberration, refraction due to fluid and optical windows, and perspective effects, it is difficult to determine the function precisely. In particular, the objective lens 27a of the fiberscope has a small lens diameter and a short focal length. In a lens system with a short focal length, distortion is not negligible. In addition, since the laser sheet 17 is inclined with respect to the imaging surfaces of the CCD cameras 24L and 24R, it is important to associate the coordinates in the camera with the actual coordinates. In addition, the data obtained by the calibration becomes very important in calculating the three-component position of the particle. However, since the association between the real coordinates and the pixel coordinates involves lens aberration, refraction by the fluid and the optical window, and perspective effects, it is difficult to determine the function precisely. Therefore, in this embodiment, a method using a function that associates real coordinates and pixel coordinates called a mapping function (Non-Patent Document 3) is adopted.

  That is, the mapping function approximates the actual coordinates x and y by a polynomial of pixel coordinate values X and Y. In this embodiment, an approximation to a cubic polynomial as shown in the following equation (9) is adopted.

In Expression (9), X _i and Y _i are pixel coordinates obtained by calibration, and x _i and y _i are corresponding coordinate values (known) in real coordinates. Further, a _i and b _i are coefficients and need to be calculated by calibration. Here, n is the number of points used for calibration. When n = 16, the number coincides with the number of coefficients a _i and b _i , and the coefficient can be calculated from the calculation of the inverse matrix. However, since the actual positions of X _i and Y _i have uncertainties that depend on the resolution of the camera, in order to suppress the mapping error caused by this, the optimum coefficients from more calibration points. Need to be determined. Therefore, in this embodiment, the least square method is used as the coefficient determination method. Coefficient determination by the method of least squares is obtained from a normal equation shown in Expression (10).

For x = XA,
X ^T x = X ^T XA (10)
Here, X corresponds to the pixel coordinate value matrix of Equation (9), x corresponds to the real coordinate value matrix of Equation (9), and A corresponds to the coefficient matrix of Equation (9). The product X ^T X with the transposed matrix becomes a square matrix by the normal equation, and the inverse matrix (X ^T X) ⁻¹ can be easily obtained. By calculating the coefficient matrix A, it is possible to convert the in-camera coordinates to the real coordinates via the equation (9).

  Conversely, conversion from real coordinates to in-camera coordinates can be easily performed by calculating an inverse matrix of a coefficient matrix. Further, it can be performed by a method similar to associating the coordinates in the two cameras, and the coordinates in the two cameras and the actual coordinates are associated with each other.

  A specific calibration procedure in the present embodiment will be described below.

First, a calibration sheet having 16 lattice points as shown in FIG. 22 (white portion transmits light) is created by printing. In the present embodiment, a sample having a lattice point spacing of 2.5 mm was prepared. As shown in FIG. 23, the calibration sheet is moved to each of three positions (z = a, 0, b), and images of the calibration sheet are taken by the two cameras A and B on the left and right. The positions of 16 grid points are detected based on the captured pixel image. In the detection, the particle detection algorithm described in the particle detection process 432 is used. After calculating the pixel coordinates, a coefficient matrix is obtained for a total of six pixel images captured by the left and right cameras at three locations from Equations (9) and (10). Here, the coefficient matrix at the position of z = a in the camera A is A _a , the coefficient matrix at z = b is A _b , the coefficient matrix at z = 0 is A _c , and the coefficient at the position of z = a in the camera B is The matrix is B _a , the coefficient matrix at z = b is B _b , and the coefficient matrix at z = 0 is B _c .

  Based on the conversion matrix thus calculated, the actual three-dimensional coordinates are calculated. The method will be described next.

A method of calculating the three-dimensional real coordinates of the particles using the above calibration method will be described. FIG. 24 is a conceptual diagram of the calculation of the actual coordinates of the particles, and is a view of the area illuminated by the laser sheet as viewed from above. First, for one camera, two calibrations on the left and right sides of z = a and b are performed. Real coordinates are calculated for the matrix by equation (9). As a result, the two real space coordinates (x _a , y _a , a), (x _b , y _b , b) are obtained. The formula of the straight line connecting these is (x−x _a ) / (x _a −x _b ) = (y−y _a ) / (y _a −y _b )
= (Za) / (ab) (11)
It becomes. Similarly, a straight line equation is obtained for the other camera, and the position where these two straight lines intersect is the actual particle position. However, in reality, the two straight lines do not intersect exactly due to an error caused by the limited resolution of the camera, an error inherent in the two-dimensional vector, the influence of noise, and the like. In the present embodiment, the process is performed by regarding the midpoint of the point on the two straight lines that minimizes the distance between the two straight lines as an intersection.

However, in performing this processing, first, the particles must be associated with each other on the left and right cameras. Therefore, in actuality, first, in order to search for a corresponding candidate for a particle, it is assumed that all particles exist at z = 0, and conversion into real coordinates is performed using coefficient matrices A _c and B _c at z = 0. Since the converted real coordinate value is a position where the particle is assumed to be z = 0, the real coordinate values of the left and right camera particles do not necessarily match. Accordingly, the right camera particle candidates existing around a certain particle of the left camera are obtained, and the two straight lines in FIG. 24 are associated with each of the candidates. This process is performed, the case where the distance between the straight lines is the shortest is regarded as the same particle, and the particle matching process is performed.

  In the three-dimensional reconstruction process 46, in the final vector data processes 45L and 45R, images of the individual particles captured in synchronization with the left and right CCD cameras 24L and 24R and correlated in the PTV-SS process 433 are exchanged. Three-dimensional velocity vector data information can be obtained by obtaining the actual coordinate values at the first time and the second time that are slightly different from each other by the above method.

  In order to confirm the applicability of the flow measurement method of the present invention to an actual flow field, the height 2D of the axially symmetric jet (outlet flow velocity 214 mm / s, Re = 5300) is arranged as shown in FIG. A speed measurement experiment was conducted at D = nozzle diameter. In order to verify whether or not the third direction (z direction) component is detected, the spatial average vector obtained by measuring the probe by tilting about 15 ° with respect to the flow field is as shown in FIGS. An average slope of 13.3 ° was detected.

  Table 1 below shows the relationship between the number of three-dimensional vectors detected as a result of processing 50 sets and 100 identical images using the conventional measurement method (for example, Non-Patent Document 1) and the method of the present invention. . In the conventional method, when the laser sheet is rotated around the x-axis, that is, when a flow field is generated in a direction penetrating the surface of the laser sheet, the number of vector calculations is reduced to about 1/3. However, such a decrease was not observed when the method of the present invention was used. In addition, the number of vector detections increased by about 25 times on average.

    Table 1 ┌────┬──────┬──────┬───────┐ │Tilt angle │Old method (this) │ This method (this) │ Detection rate (times) ) │ ├────┼──────┼──────┼───────┤ │ 0 ° │ 289 │ 3583 │ 12.40 │ ├────┼── ────┼──────┼───────┤ │ 15 ° │ 99 │ 3844 │ 38.83 │ ├────┼──────┼───── ─┼───────┤ │ 30 ° │ 127 │ 3967 │ 31.24 │ └────┴──────┴──────┴──────┘.

  As described above, according to the present invention, it is possible to achieve a fluid flow measurement system and a measurement method thereof that are more reliable, more accurate, and capable of precise measurement.

It is a figure which shows the outline | summary of 1 Example of the flow measurement system of the fluid of this invention. It is a figure which shows the cross section along the sheet surface of the laser sheet of the optical system for laser sheet formation, and a cross section orthogonal to it. It is the perspective view and sectional drawing of the probe part of the front-end | tip of a right and left image transmission means. It is a figure for demonstrating the structure of an image transmission means. It is a figure for demonstrating the image processing software of the fluid flow measurement system of one Example of this invention. It is a figure which shows the example of the miscorrespondence of the particle group in a cross correlation PIV process. It is a figure which shows the example of the flow velocity vector (PIV vector) distribution obtained by performing cross correlation calculation by the PIV process of this invention. It is the figure which showed the flow of the image processing by which the luminance pattern of the optical fiber arrangement | sequence was removed by the removal process of fiber noise, and only a particle image remains with the luminance value distribution of arbitrary lines. It is a figure for demonstrating the false correspondence of the particle | grains by the resolution reduction of the image by an optical fiber bundle. It is a figure which shows the example of the luminance value distribution of the arbitrary lines of the image of the image guide image | photographed actually, and frequency analysis. It is a figure which shows the outline | summary of the dynamic binarization method. It is a figure for demonstrating the division | segmentation and interpolation of an image in a dynamic binarization method. It is a figure which shows the comparison of the PTV-SS method used by this invention, and the conventional PTV method. It is a figure regarding only the part of the locus of the particle of the PTV-SS method. It is a figure which shows the possibility of the vector in the PTV-SS method using the PIV vector of this invention, and its selection. It is a vector map which shows the example of the movement amount detection result computed by the method of this invention from the particle locus image which image | photographed the pseudo | simulation flow. It is a figure which shows the example of the autocorrelation value map which arranged the autocorrelation value of the reference image. It is a figure which shows the example of the cross correlation value map which arranged the cross correlation value of a reference image and a search image. FIG. 19 is a diagram illustrating an example of a cross-correlation value map in which cross-correlation values of the autocorrelation value map of FIG. 17 and the cross-correlation value map of FIG. 18 are arranged. It is a figure which shows an example which performed the quadratic curve approximation to the cross correlation value map obtained from the cross correlation value of an autocorrelation value map and a cross correlation value map. It is a figure which shows typically the process of calculating the speed in a real coordinate system by matching a real coordinate and the in-camera coordinate of two cameras by calibration in stereo PTV. It is a figure which shows the example of a calibration sheet | seat. It is a figure which shows a mode that the calibration sheet | seat of FIG. 22 is moved to a position, and the image of the calibration sheet | seat is imaged by two right and left cameras. It is a figure for demonstrating the method of calculating the three-dimensional real coordinate of particle | grains using the calibration method of FIG. It is a figure which shows the example of arrangement | positioning for confirming the applicability to the actual flow field of the flow measurement method of this invention. It is a figure which shows the space average vector distribution of the result measured by the example of arrangement | positioning of FIG. It is another figure which shows the space average vector distribution as a result measured by the example of arrangement | positioning of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Flow measurement system 11 ... Laser oscillator 12 ... Optical fiber 13 for light transmission ... Optical system 14 for laser sheet formation ... Fluid flow field 17 ... Laser sheet 17a ... Measurement range 18 ... Lens group 19a, 19b, 19c ... Plano-convex lens 19d ... Plano-concave cylindrical lenses 22, 22L, 22R ... Image transmission means 24, 24L, 24R ... CCD camera 25 ... Optical fiber 26 ... Image guide 27a ... Objective lens 27b ... 45 ° right angle prism 28 ... Camera lenses 29L, 29R ... Images Imaging means 30 ... Signal generator 31 ... PC 32 ... Laser driver 33 ... Peltier driver 34L, 34R ... Interface 41L, 41R ... Camera image 42L ... PIV processing 421 ... Fiber noise removal processing 422 ... Correlation PIV processing 423 ... Error vector removal And space flat Processing 43L ... PTV-SS processing 431 ... Fiber noise removal processing 432 ... Particle detection processing 433 ... PTV-SS method 44L using PIV vector ... Subpixel processing 45L, 45R of image degradation due to optical fiber bundles ... Final vector Data processing 46 ... 3D reconstruction processing

Claims (10)

Formed by a laser oscillation device that oscillates laser light, a laser sheet forming optical system that allows the oscillated laser light to be injected into a fluid flow field in a sheet shape, and a laser sheet from the laser sheet forming optical system. The image capturing means for selectively capturing the two-dimensional particle image and the two-dimensional particle trajectory image, and the brightness pattern of the two-dimensional particle image captured at two times different for a very short time by the image capturing means The first image processing means for measuring the moving direction and the moving amount of the particle group, and the luminance pattern of the two-dimensional particle trajectory images at two times different for a minute time taken by the image pickup means are compared and analyzed, and each particle A second image processing means for measuring a moving direction and a moving amount, wherein the second image processing means is based on the moving direction of the particle group measured by the first image processing means. Flow measurement system of the fluid, characterized by being configured to determine the direction of movement of the individual particles. 2. The fluid flow measurement system according to claim 1, wherein the laser oscillation device is constituted by a semiconductor laser, and forms a two-dimensional particle image and a two-dimensional particle locus image according to a difference in emission time of the semiconductor laser. Two image capturing means for selectively capturing a two-dimensional particle image and a two-dimensional particle trajectory image simultaneously from both sides of the laser sheet from the laser sheet forming optical system, and for each of the image capturing means, A first image processing unit; and a second image processing unit. Each of the second image processing units includes individual particles based on the moving direction of the particle group measured by each of the first image processing units. The fluid flow measurement system according to claim 1, wherein the movement direction of the fluid is determined. There is a three-dimensional reconstruction means for reconstructing the three-dimensional movement direction and movement amount of each particle by associating the movement direction and movement amount of each particle measured by each of the second image processing means with each other. 4. The fluid flow measurement system according to claim 3, wherein the three-dimensional movement direction and movement amount of each particle are measured. The image pickup means includes an image transmission means for optically transmitting a two-dimensional particle image and a two-dimensional particle trajectory image formed by illumination with a laser sheet from the laser sheet forming optical system, and an optically transmitted two-dimensional particle. The fluid flow measurement system according to any one of claims 1 to 4, further comprising an image pickup unit configured to pick up an image and a two-dimensional particle trajectory image. The image transmission means is composed of a flexible image guide in which a large number of optical fibers are bundled and integrated, and both end faces are flattened. The imaging means is composed of a CCD camera, and the image The guide is configured to disassemble an image formed on an end face of one fiber by an objective lens into each pixel by an optical fiber and transmit the same image to the other end face on the CCD camera side. Item 6. The fluid flow measurement system according to Item 5. The fluid flow measurement system according to any one of claims 1 to 6, wherein the laser sheet forming optical system and the image capturing unit are integrally combined. The first image processing means and the second image processing means each include fiber noise removal means for removing fiber noise based on the image guide from a two-dimensional particle image and a two-dimensional particle trajectory image. Item 8. The fluid flow measurement system according to Item 6 or 7. 9. The apparatus according to claim 1, further comprising sub-pixel processing means for making sub-pixel accuracy the moving direction and moving amount of each particle measured by the second image processing means. The fluid flow measurement system described. The laser beam oscillated from the laser oscillation device is irradiated onto the fluid flow field in the form of a sheet to form a laser sheet, and a two-dimensional particle image and a two-dimensional particle trajectory image formed by illumination with the laser sheet are selectively used. The image is picked up by the image pickup means, the luminance patterns of the two-dimensional particle images at two times that are taken at different times are compared and analyzed, and the moving direction and amount of movement of the particle group are measured. The luminance patterns of the two-dimensional particle trajectory images at the time were compared and analyzed to measure the movement direction and movement amount of each particle, and were measured first when measuring the movement direction and movement amount of each particle. A fluid flow measuring method, comprising: measuring a fluid flow in a fluid flow field by determining a moving direction of each particle based on a moving direction of a particle group.