Chip positioning algorithm suitable for turret type sorting machine

Abstract

The invention relates to the field of chip positioning, in particular to a chip positioning algorithm suitable for a turret type sorting machine, which comprises the following steps of: S1, obliquely arranging a visual module to obtain a plurality of frames of continuous first oblique images in real time; s2, screening the plurality of first inclined images, and marking the screened first inclined images as second inclined images; s3, analyzing the motion state of the target chip according to the second inclined image; s4, calculating the actual position of the target chip according to the motion state of the target chip and the second inclined image; according to the method, the situation that deviation still exists after the position of the chip is determined according to the movement direction and speed of the chip is considered, so that the key points with low reliability can be automatically subjected to weight descending by using angle dual constraints between the key edges, and the influence of part of key points with large errors on the whole is avoided; the calculation method provided by the invention only aims at key point optimization instead of full-graph optimization, and can reduce the calculation amount while improving the calculation precision.

Description

Technical Field

[0001] The present invention relates to the field of chip positioning, and specifically to a chip positioning algorithm for a turret type sorter. Background Art

[0002] Chips are generally fed by vibration, and then the sorter picks up the chips. When the sorter picks up the chips, it generally takes pictures of the chip positions through a camera to calculate the chip positions, and then controls the suction nozzle to pick up the chips. However, when the camera takes pictures of the chips, the vibration feeding tray is very likely to be still in a vibrating state, so there are large errors in the taken pictures, and there will also be chip gaps, which is not conducive to determining the positions of the chips. When the suction nozzle sucks the chips, it is easy to suck them obliquely or suck nothing. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides a chip positioning algorithm for a turret type sorter, which solves the technical problems in the above background art.

[0004] To achieve the above object, the present invention provides the following technical solutions: A chip positioning algorithm for a turret type sorter includes the following steps: S1. Incline the vision module to obtain several consecutive first inclined images in real time; S2. Screen several first inclined images, and mark the selected first inclined images as second inclined images; In step S2, it specifically includes the following steps: S21. Construct a real coordinate system with the center point of the picking position as the origin 0. In the real coordinate system, the east-west direction is the x-axis, the north-south direction is the y-axis, and the vertical direction is the z-axis; S22. Construct a standard chip edge map according to the chip shape. The standard chip edge map includes several standard key points and standard key edges, and mark the coordinate positions of each standard key point in the real coordinate system; S23. Use the openpose algorithm to extract the to-be-detected key points in the first inclined images, use the to-be-detected key edges to connect the to-be-detected key points to generate a to-be-detected map, and calculate the coordinate positions of each to-be-detected key point in the real coordinate system; S24. Divide several first inclined images into multiple image sets to be calculated in chronological order, and calculate the average value of the image set similarity of each image set to be calculated; S25. Select the image set to be calculated with the highest average similarity value, and mark the first inclined image therein as the second inclined image; S3. Analyze the motion state of the target chip according to the second inclined image; S4. Calculate the actual position of the target chip based on the motion state of the target chip and the second tilted image.

[0005] Further, in step S1, the first tilted images are sorted in chronological order, and the time interval between every two adjacent first tilted images is t.

[0006] Further, in step S23, it specifically includes the following steps: S231. Take the center point of the first tilted image as the origin to construct a camera coordinate system, with the horizontal and vertical directions of the first tilted image as the horizontal and vertical axes of the camera coordinate system; S232. Obtain the camera internal parameter matrix K of the vision module, and its expression is: , where and respectively represent the focal lengths of the vision module on the horizontal and vertical axes of the camera; and respectively represent the coordinates of the plane center of the first tilted image; S233. Determine the homogeneous coordinates of the key point to be measured in the camera coordinate system , and its expression is: , where and respectively represent the horizontal and vertical axis coordinates of the key point to be measured in the camera coordinate system; represents the transpose operation of the matrix; S234. Calculate the coordinate position of the key point to be measured in the real coordinate system according to the camera internal parameter matrix K and the homogeneous coordinates , and its calculation formula is: , where represents the three-dimensional coordinates of the key point to be measured in the real coordinate system, ; represents the depth value of the key point to be measured from the optical axis of the vision module; in the present invention, is detected by a depth sensor provided in the vision module.

[0007] Further, in step S24, it specifically includes the following steps: S241. Calculate the key point position difference between each key point to be measured and the corresponding standard key point , and its calculation formula is: , where and respectively represent the Euclidean distances between the i-th standard key point and the key point to be measured; S242. Calculate the average key edge difference between each key edge to be measured and the corresponding standard key edge , and its calculation formula is: , where represents the total number of standard key points; represents the Euclidean distance between the i-th and j-th standard key points in the standard chip edge map; represents the Euclidean distance between the i-th and j-th measured key points in the measured image; S243. Calculate the average difference in the angles between every two measured key edges and the corresponding two standard key edges , and its calculation formula is: , where represents the angle between the i-th and j-th standard key edges in the standard chip map; represents the angle between the i-th and j-th measured key edges in the measured image; S244. According to the key point position difference , the average key edge difference and the average angle difference calculate the similarity between the measured image and the standard chip edge map , and its calculation formula is: , where , and respectively represent the first, second, and third weight coefficients with respect to ; S245. Set a first sliding window in a number of frames of the first tilted image. The length of the first sliding window is b, and the first sliding window slides one frame each time to obtain a number of image sets to be calculated; S246. Calculate the similarity between each measured image in each image set to be calculated and the standard chip edge map , and calculate the average similarity of the image sets based on this.

[0008] Furthermore, in step S3, it specifically includes the following steps: S31. Obtain a number of consecutive historical tilted images captured by the vision module, set a second sliding window, the length of the second sliding window is b, and the second sliding window slides one frame each time to obtain b frames of historical tilted images as training samples; S32. Mark the movement direction and movement speed of the chip in the latter moment of each training sample by manual marking, and use them as sample labels; S33. Use the training samples and sample labels to train the long short-term memory network to obtain a target model; S34. Input all the second tilted images into the target model, and output the movement direction and movement speed of the target chip at this time.

[0009] Further, in step S4, it specifically includes the following steps: S41. Obtain the system delay time of the vision module ; S42. According to the system delay time and the motion state of the target chip, correct the three-dimensional coordinates of each key point to be measured in the real coordinate system to obtain the corrected key points. The calculation formula is: ; Expand it to: , In the formula, represents the three-dimensional coordinates of the corrected key point; ; represents the three-dimensional coordinates of the key point to be measured; ; represents the three-dimensional velocity vector of the chip in the real coordinate system; represents the polar angle of the chip's motion direction; represents the azimuth angle of the chip's motion direction; S43. Construct corrected key edges according to the corrected key points; S44. Calculate the comprehensive reliability of each corrected key point according to the length of the corrected key edge and the angle between adjacent corrected key edges ; S45. Select the two corrected key points with the highest comprehensive reliability as the basic key points, and construct a determined chip map according to the basic key points to determine the actual position of the target chip with the determined chip map.

[0010] Further, in step S44, it specifically includes the following steps: S441. Calculate the side length deviation of each corrected key edge , and its calculation formula is: , in the formula, represents the length of the corrected key edge ij; represents the length of the standard key edge ij; S442. Calculate the angle deviation of the angle between every two adjacent corrected key edges , and its calculation formula is: ; where, , in the formula, represents the angle between the corrected key edge ij and the corrected key edge ik; represents the angle between the standard key edge ij and the standard key edge ik; represents function; , and respectively represent the three-dimensional coordinates of the i-th, j-th, and k-th calibration key points; S443. According to the side length deviation calculate the side length reliability of each calibration key edge , and its calculation formula is: ; where , in the formula, represents the average length deviation of all adjacent calibration key edges of the i-th calibration key point; represents with respect to the attenuation control parameter; represents the number of adjacent calibration key edges of the i-th calibration key point; represents the set of all adjacent calibration key edges of the i-th calibration key point; represents function; S444. According to the included angle deviation calculate the angle reliability of each calibration key point , and its calculation formula is: ; where , in the formula, represents the average angle deviation between all adjacent calibration key edges of the i-th calibration key point; represents with respect to the attenuation control parameter; represents the number of all adjacent calibration key edges of the i-th calibration key point; represents the set of all adjacent calibration key edges of the i-th calibration key point; S445. According to the side length reliability and the angle reliability calculate the comprehensive reliability of each calibration key point.

[0011] Furthermore, in step S445, the calculation formula of the comprehensive reliability is: , in the formula, represents with respect to the weight coefficient.

[0012] Compared with the prior art, the present invention provides a sorting machine chip positioning algorithm applicable to a turret type, having the following beneficial effects: 1. After determining the position of the chip based on the movement direction and speed of the chip, there are still deviations. Therefore, the present invention uses the double constraints of the angles between the key edges, and the key points with low reliability will be automatically weighted down to avoid the influence of some key points with large errors on the whole. In addition, the calculation method of the present invention only optimizes for key points rather than the whole image, which can improve the calculation accuracy of the chip position while reducing the calculation amount, so as to avoid the situation of the nozzle sucking air in vain or sucking obliquely.

[0013] 2. When calculating the similarity between the image to be measured and the standard chip edge image, the present invention not only considers the accuracy of the position of each key point, but also considers the accuracy of the length and included angle of the key edges. In addition, the present invention also considers the factor of motion blur. Compared with the common calculation methods, the calculation accuracy of the similarity between the image to be measured and the standard chip edge image is higher, and the first inclined image in which the chip is moving rapidly can be excluded, so as to facilitate screening out the image to be measured that is closest to the standard chip edge image.

[0014] 3. The vision module of the present invention can be set obliquely to avoid the influence of the vision module on the number of nozzle components. Compared with the common vision module setting methods, nozzle components can be added, thereby improving the chip picking efficiency. Brief Description of the Drawings

[0015] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation of the present application. In the drawings: Figure 1 It is a flowchart of a chip positioning algorithm for a turret type sorter according to the present invention. Detailed Embodiments

[0016] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments. Thereby, the implementation process of how the present application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0017] Those of ordinary skill in the art can understand that all or part of the steps of implementing the following embodiment methods can be completed by instructing relevant hardware through a program. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0018] When loading the chips onto the chip, a vibrating feeder can be used to sequentially transfer the chips from a disordered state to the picking position. Then, the vision module positions the chips, and finally, the nozzle assembly is finely adjusted according to the chip position to facilitate the nozzle to pick up the chips. However, since the vibrating feeder tray feeds the chips through vibration, when the vibrating feeder tray stops, the entire device is very likely to still have vibration. If the above factors are not considered, when using the picking scheme mentioned above, problems such as the nozzle not being firmly adsorbed and the adsorption position deviation are very likely to occur. Therefore, as Figure 1 shown, the present invention proposes a chip positioning algorithm applicable to a turret sorter, including the following steps: S1. The vision module is inclined to obtain several consecutive first inclined images in real time; specifically, since the nozzle assembly is generally installed on the rotating disk, if the vision module is vertically installed, it will affect the number of rotating disks. Therefore, the vision module is inclined to facilitate increasing the number of nozzle assemblies. It should be noted that in addition, in step S1, the first inclined images are sorted in chronological order, and the time interval between every two adjacent first inclined images is t.

[0019] By inclining the vision module of the present invention, the influence of the vision module on the number of nozzle assemblies can be avoided. Compared with the common vision module setting method, the nozzle assembly can be added, thereby improving the chip picking efficiency.

[0020] S2. Screen several first inclined images, and mark the screened first inclined images as second inclined images; specifically, when the vision module takes pictures of the picking position, only the first inclined image in which the chip reaches the picking position is available. Under the action of the vibrating feeder tray, both the chip and the picking position will be in a vibrating state. Therefore, the images obtained by the vision module are very likely to be motion blurred. Therefore, in step S2, it specifically includes the following steps: S21. Construct a real coordinate system with the center point of the picking position as the origin 0. In the real coordinate system, the east-west direction is the x-axis, the north-south direction is the y-axis, and the vertical direction is the z-axis; S22. Construct a standard chip edge map according to the chip shape. The standard chip edge map includes several standard key points and standard key edges, and mark the coordinate positions of each standard key point in the real coordinate system; S23. Use the openpose algorithm to extract the key points to be measured in the first inclined image, use the key edges to be measured to connect the key points to be measured to generate a to-be-measured map, and calculate the coordinate positions of each key point to be measured in the real coordinate system; specifically, in step S23, it specifically includes the following steps: S231. Take the center point of the first inclined image as the origin Construct a camera coordinate system with the horizontal and vertical directions of the first tilted image as the horizontal and vertical axes of the camera coordinate system; S232. Obtain the camera internal parameter matrix K of the vision module, and its expression is: , where and respectively represent the focal lengths of the vision module on the horizontal and vertical axes of the camera; and respectively represent the coordinates of the center of the plane of the first tilted image; S233. Determine the homogeneous coordinates of the key point to be measured in the camera coordinate system, and its expression is: , where and respectively represent the horizontal and vertical axis coordinates of the key point to be measured in the camera coordinate system; represents the transpose operation of the matrix; S234. Calculate the coordinate position of the key point to be measured in the real coordinate system according to the camera internal parameter matrix K and the homogeneous coordinates , and its calculation formula is: , where represents the three-dimensional coordinates of the key point to be measured in the real coordinate system, ; represents the depth value of the key point to be measured from the optical axis of the vision module; in the present invention, is obtained by detecting with a depth sensor provided in the vision module.

[0021] S24. Divide several first tilted images into multiple image sets to be calculated according to the time sequence, and calculate the average similarity of each image set to be calculated; specifically, due to physical factor limitations of the vision module (including perspective distortion, lens distortion, and insufficient resolution) and dynamic factor limitations during the movement process (including motion blur and frame rate mismatch), etc., when using the openpose algorithm to extract the key points to be measured in the first tilted image and then construct the image to be measured, a certain degree of deformation will occur, and the physical factor limitations of the vision module are often approximate, while the dynamic factor limitations will exacerbate the generation of deformation. Therefore, in step S24, it specifically includes the following steps: S241. Calculate the key point position difference between each key point to be measured and the corresponding standard key point, and its calculation formula is: , where and respectively represent the Euclidean distances of the i-th standard key point and the key point to be measured; S242. Calculate the average key edge difference between each key edge to be measured and the corresponding standard key edge, and its calculation formula is: , where represents the total number of standard key points; represents the Euclidean distance between the i-th and j-th standard key points in the standard chip edge map; represents the Euclidean distance between the i-th and j-th measured key points in the measured image; S243. Calculate the average difference in the angles between every two measured key edges and the corresponding two standard key edges , and its calculation formula is: , where represents the angle between the i-th and j-th standard key edges in the standard chip map; represents the angle between the i-th and j-th measured key edges in the measured image; it should be noted that the i-th and j-th standard key edges are adjacent; S244. According to the key point position difference , the average key edge difference and the average angle difference calculate the similarity between the measured image and the standard chip edge map , and its calculation formula is: , where , and respectively represent the first, second, and third weight coefficients with respect to ; in the present invention, , and are 0.45, 0.3, and 0.25 respectively; S245. Set a first sliding window in a number of first tilted images. The length of the first sliding window is b, and the first sliding window slides one frame each time to obtain a number of image sets to be calculated; S246. Calculate the similarity between each measured image and the standard chip edge map in each image set to be calculated , and calculate the average similarity of the image sets based on this.

[0022] S25. Select the image set to be calculated with the highest average similarity, and mark the first tilted image therein as the second tilted image.

[0023] In step S2 of the present invention, when calculating the similarity between the to-be-tested image and the standard chip edge image, not only the accuracy of the position of each key point is considered, but also the accuracy of the length and included angle of the key edges is considered. In addition, the present invention also takes into account the factor of motion blur. Compared with common calculation methods, the calculation accuracy of the similarity between the to-be-tested image and the standard chip edge image is higher, and the first tilted image in which the chip is in rapid motion can be excluded, so as to facilitate screening out the to-be-tested image closest to the standard chip edge image.

[0024] S3. Analyze the motion state of the target chip according to the second tilted image; specifically, since the vibrating feeding tray may be in a continuous vibrating state, it is necessary to analyze the actual motion state of the chip in order to calculate the actual position of the chip according to it later. For this reason, in step S3, the following steps are specifically included: S31. Obtain a plurality of consecutive historical tilted images captured by the vision module, set a second sliding window, the length of the second sliding window is b, and the second sliding window slides one frame each time to obtain b-frame historical tilted images as training samples; S32. Mark the motion direction and motion speed of the chip at the later moment of each training sample by manual marking, and use it as a sample label; S33. Train the long short-term memory network with the training samples and sample labels to obtain a target model; S34. Input all the second tilted images into the target model, and output the motion direction and motion speed of the target chip at this time.

[0025] S4. Calculate the actual position of the target chip according to the motion state of the target chip and the second tilted image; specifically, in step S4, the following steps are specifically included: S41. Obtain the system delay time of the vision module ; S42. Correct the three-dimensional coordinates of each to-be-tested key point in the real coordinate system according to the system delay time and the motion state of the target chip to obtain corrected key points, and its calculation formula is: ; Expand it to: , In the formula, represents the three-dimensional coordinates of the corrected key point; ; represents the three-dimensional coordinates of the to-be-tested key point; ; represents the three-dimensional velocity vector of the chip in the real coordinate system; represents the polar angle of the motion direction of the chip; represents the azimuth angle of the motion direction of the chip; S43. Construct correction key edges based on the correction key points; S44. Calculate the comprehensive reliability of each correction key point according to the length of the correction key edge and the angle between adjacent correction key edges Specifically, since the above-mentioned calculation method is a simplified calculation model and cannot cover all the movements of the chip, there are still errors in its calculation results. Therefore, in step S44, it specifically includes the following steps: S441. Calculate the side length deviation of each correction key edge , and its calculation formula is: ; In the formula, represents the length of the correction key edge ij; represents the length of the standard key edge ij; specifically, can be directly calculated from the three-dimensional coordinates of the i-th and j-th correction key points; the correction key edge ij represents the correction key edge between the i-th and j-th correction key points; S442. Calculate the angle deviation of the angle between every two adjacent correction key edges , and its calculation formula is: ; where, , in the formula, represents the angle between the correction key edge ij and the correction key edge ik; represents the angle between the standard key edge ij and the standard key edge ik; represents function; , and respectively represent the three-dimensional coordinates of the i-th, j-th, and k-th correction key points; S443. Calculate the side length reliability of each correction key edge according to the side length deviation , and its calculation formula is: ; where, ; in the formula, represents the average length deviation of all adjacent correction key edges of the i-th correction key point; represents the attenuation control parameter with respect to ; represents the number of adjacent correction key edges of the i-th correction key point; represents the set of all adjacent correction key edges of the i-th correction key point; represents function; in the present invention, is 0.1; is 0.1; S444. According to the angle deviation Calculate the angular reliability of each calibrated key point , and its calculation formula is: ; where , in the formula, represents the average angular deviation between all adjacent calibrated key edges of the i-th calibrated key point; represents the attenuation control parameter with respect to ; represents the number of all adjacent calibrated key edges of the i-th calibrated key point; represents the set of all adjacent calibrated key edges of the i-th calibrated key point; in the present invention, is 1; S445. Calculate the comprehensive reliability of each calibrated key point according to the side length reliability and the angular reliability ; specifically, in step S445, the calculation formula of the comprehensive reliability is: , in the formula, , represents the weight coefficient with respect to ; in the present invention, is 0.7.

[0026] S45. Select the two calibrated key points with the highest comprehensive reliability as the basic key points, and construct a determined chip map according to the basic key points to determine the actual position of the target chip with the determined chip map.

[0027] In step S4 of the present invention, considering that there are still deviations after determining the position of the chip according to the moving direction and speed of the chip, so the double constraint of the angle between the key edges is used, and the key points with low reliability will be automatically down-weighted to avoid the influence of some key points with large errors on the whole. In addition, the calculation method of the present invention only optimizes the key points rather than the whole map, which can reduce the calculation amount while improving the calculation accuracy.

[0028] The above embodiments have introduced the present invention in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A chip positioning algorithm applicable to a turret type sorting machine, characterized in that, Including the following steps: S1. Set the vision module obliquely to obtain several consecutive first oblique images in real time; S2. Screen several first oblique images, and mark the selected first oblique images as second oblique images; In step S2, it specifically includes the following steps: S21. Construct a real-world coordinate system with the center point of the picking position as the origin 0. In the real-world coordinate system, the east-west direction is the x-axis, the north-south direction is the y-axis, and the vertical direction is the z-axis; S22. Construct a standard chip edge map according to the chip shape. The standard chip edge map includes several standard key points and standard key edges, and mark the coordinate positions of each standard key point in the real-world coordinate system; S23. Use the openpose algorithm to extract the key points to be measured in the first oblique image, connect the key points to be measured with the key edges to be measured to generate a to-be-measured map, and calculate the coordinate positions of each key point to be measured in the real-world coordinate system; S24. Divide several first oblique images into multiple image sets to be calculated according to the time sequence, and calculate the average similarity of each image set to be calculated; S25. Select the image set to be calculated with the highest average similarity, and mark the first oblique images in it as second oblique images; S3. Analyze the motion state of the target chip according to the second oblique image; S4. Calculate the actual position of the target chip according to the motion state of the target chip and the second oblique image.

2. The sorting machine chip positioning algorithm applicable to a turret according to claim 1, wherein In step S1, the first oblique images are sorted according to the time sequence, and the time interval between every two adjacent first oblique images is t.

3. The sorting machine chip positioning algorithm applicable to a turret according to claim 1, characterized in that, In step S23, it specifically includes the following steps: S231. Taking the center point of the first tilted image as the origin to construct a camera coordinate system, with the horizontal direction and the vertical direction of the first tilted image as the horizontal axis and the vertical axis of the camera coordinate system; S232. Obtain the camera internal parameter matrix K of the vision module, and its expression is: , where and respectively represent the focal lengths of the vision module on the horizontal axis and vertical axis of the camera; and respectively represent the coordinates of the plane center of the first tilted image; S233. Determine the homogeneous coordinates of the key point to be measured in the camera coordinate system , and its expression is: , where and respectively represent the horizontal axis coordinate and the vertical axis coordinate of the key point to be measured in the camera coordinate system; represents the transpose operation of the matrix; S234. Calculate the coordinate position of the key point to be measured in the real coordinate system according to the camera internal parameter matrix K and the homogeneous coordinates . The calculation formula is as follows: , where represents the three-dimensional coordinates of the key point to be measured in the real coordinate system, ; represents the depth value of the key point to be measured from the optical axis of the vision module; in the present invention, is obtained by detecting through the depth sensor provided in the vision module.

4. The sorting machine chip positioning algorithm applicable to a turret according to claim 1, characterized in that, In step S24, it specifically includes the following steps: S241. Calculate the key-point position difference between each key point to be measured and the corresponding standard key point , and its calculation formula is: , where in the formula, and respectively represent the Euclidean distances of the i-th standard key point and the key point to be measured; S242. Calculate the average key-edge difference between each key edge to be measured and the corresponding standard key edge , and its calculation formula is: , where in the formula, represents the total number of standard key points; represents the Euclidean distance between the i-th and j-th standard key points in the standard chip edge map; represents the Euclidean distance between the i-th and j-th key points to be measured in the map to be measured; S243. Calculate the average difference between the angles of every two key edges to be measured and the angles between the corresponding two standard key edges , and its calculation formula is: , where in the formula, represents the angle between the i-th standard key edge and the j-th standard key edge in the standard chip diagram; represents the angle between the i-th key edge to be measured and the j-th key edge to be measured in the diagram to be measured; S244. Calculate the similarity between the graph to be measured and the edge graph of the standard chip according to the position difference of key points , the average difference of key edges and the average difference of included angles . The calculation formula is as follows: , where , in the formula, , and respectively represent the first, second, and third weight coefficients with respect to ; S245. Set a first sliding window in several frames of first oblique images. The length of the first sliding window is b, and the first sliding window slides one frame each time to obtain several image sets to be calculated; S246. Calculate the similarity between each image to be calculated in each image set to be calculated and the standard chip edge image, and calculate the average value of the image set similarity based on it.

5. The sorting machine chip positioning algorithm applicable to a turret according to claim 1, characterized in that, In step S3, it specifically includes the following steps: S31. Obtain several consecutive historical oblique images taken by the vision module, set a second sliding window, the length of the second sliding window is b, and the second sliding window slides one frame each time to obtain b historical oblique images as training samples; S32. Manually mark the motion direction and motion speed of the chip in each training sample at the next moment of the training sample, and use it as a sample label; S33. Use the training samples and sample labels to train the long short-term memory network to obtain a target model; S34. Input all the second oblique images into the target model, and output the motion direction and motion speed of the target chip at this time.

6. The sorting machine chip positioning algorithm applicable to a turret according to claim 1, wherein In step S4, it specifically includes the following steps: S41. Obtain the system delay time of the vision module ; S42. According to the system delay time and the motion state of the target chip, correct the three-dimensional coordinates of each key point to be measured in the real coordinate system to obtain corrected key points. The calculation formula is as follows: ； Expand it to: , In the formula, represents the three-dimensional coordinates of the calibration key point; ; represents the three-dimensional coordinates of the key point to be measured; ; represents the three-dimensional velocity vector of the chip in the real coordinate system; represents the polar angle of the movement direction of the chip; represents the azimuth angle of the movement direction of the chip; S43. Construct a corrected key edge according to the corrected key points; S44. Calculate the comprehensive reliability of each calibration key point based on the length of the calibrated key edge and the angle between adjacent calibrated key edges ; S45. Select the comprehensive reliability Select the two calibration key points with the highest reliability as the basic key points, and construct a determined chip map based on the basic key points, and use the determined chip map as the actual position of the target chip.

7. The sorting machine chip positioning algorithm applicable to a turret according to claim 6, characterized in that, In step S44, it specifically includes the following steps: S441. Calculate the side length deviation of each calibration key edge , and its calculation formula is: , where in the formula, represents the length of the calibration key edge ij; represents the length of the standard key edge ij; S442. Calculate the angular deviation of the angle between every two adjacent calibration key edges , and its calculation formula is: ; where , in the formula, represents the angle between the calibration key edge ij and the calibration key edge ik; represents the angle between the standard key edge ij and the standard key edge ik; represents function; , and respectively represent the three-dimensional coordinates of the i-th, j-th, and k-th calibration key points; S443. Calculate the side length reliability of each calibration key side according to the side length deviation The calculation formula is as follows: , and the calculation formula is: where, , in the formula, represents the average length deviation of all adjacent calibration key sides of the i-th calibration key point; represents the attenuation control parameter with respect to ; represents the number of adjacent calibration key sides of the i-th calibration key point; represents the set of all adjacent calibration key sides of the i-th calibration key point; represents function; S444. Calculate the angular reliability of each calibration key point according to the included angle deviation The calculation formula is as follows: ; where In the formula, represents the average angular deviation between all adjacent calibration key edges of the i-th calibration key point; represents the attenuation control parameter with respect to ; represents the number of all adjacent calibration key edges of the i-th calibration key point; represents the set of all adjacent calibration key edges of the i-th calibration key point;​ S445. Calculate the comprehensive reliability of each calibration key point based on the side length reliability and the angle reliability . .

8. The sorting machine chip positioning algorithm applicable to a turret according to claim 7, characterized in that In step S445, the comprehensive reliability is calculated by the formula: , where represents the weight coefficient with respect to .

Images