An automatic recognition and positioning method based on a machine vision binocular measurement system

Abstract

The present invention discloses an automatic recognition and positioning method based on a machine vision binocular measurement system, which relates to the technical field of automatic recognition and positioning. It solves the technical problem in the prior art that it is impossible to perform cooperative recognition and positioning through image recognition, thus reducing the recognition and monitoring accuracy of the machine vision binocular measurement system. Specifically, it performs cooperative recognition and positioning through image recognition to facilitate accurate recognition and positioning of the photographed object, improving the recognition and monitoring accuracy of the machine vision binocular measurement system. At the same time, according to the binocular measurement system, it can monitor the surface shape defect of the photographed object, and can also detect the generation of shape defects during the movement stage of the photographed object, which is beneficial to the wear traceability of the photographed object, and can also perform shape defect repair in time when wear occurs, ensuring the use quality of the photographed object to the greatest extent.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic identification and positioning, and specifically provides an automatic identification and positioning method based on a machine vision binocular measurement system. Background Art

[0002] A machine vision binocular measurement system is a measurement system based on binocular stereo vision technology. By simulating the principle of human eye vision, it uses two cameras to capture an object from different angles, obtains the three-dimensional information of the object, and realizes three-dimensional measurement. By using a high-precision calibration board and an advanced calibration algorithm, the internal and external parameters of the camera are accurately obtained, the calibration error is reduced, and the measurement and recognition accuracy are improved.

[0003] However, in the prior art, the machine vision binocular measurement system cannot perform monocular camera detection and binocular camera detection during operation. When analyzing the collected images of binocular cameras in cooperation, due to the deviation of the collected images of monocular cameras, it is also impossible to avoid the shooting deviation when the binocular cameras cooperate to shoot, resulting in abnormal shooting positioning of the binocular cameras and the inability of the collected images of the binocular cameras to cooperate with image analysis. In addition, according to the collected images of the real-time double camera group, it is impossible to perform cooperative recognition and positioning through image recognition, thus reducing the recognition and monitoring accuracy of the machine vision binocular measurement system.

[0004] In view of the above technical defects, a solution is proposed herein. Summary of the Invention

[0005] The purpose of the present invention is to solve the above-mentioned problems and provide an automatic identification and positioning method based on a machine vision binocular measurement system.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] An automatic identification and positioning method based on a machine vision binocular measurement system, the steps of the automatic identification and positioning method are as follows:

[0008] Monocular camera detection: Two cameras in the machine vision binocular measurement system capture images from different angles, and the images captured by any one of the cameras are detected.

[0009] Binocular camera detection: After completing the monocular camera detection and passing the detection, binocular camera detection is performed.

[0010] Binocular cooperative recognition and positioning: Image recognition and analysis are performed based on the collected images of the binocular cameras, and cooperative recognition and positioning are performed through image recognition.

[0011] Modeling accuracy evaluation: After the recognition and positioning by the machine vision binocular measurement system, three-dimensional modeling of the captured object is performed, and it is inferred whether there is a difference between the currently captured object and the actual object through three-dimensional modeling detection.

[0012] As a preferred embodiment of the present invention, the monocular camera detection process is as follows:

[0013] Mark the two cameras in the machine vision binocular measurement system as a dual-camera group, and uniformly mark the cameras in the dual-camera group as single cameras; collect monocular camera information, and substitute it into the calculation to obtain the monocular camera detection coefficient. If the monocular camera detection coefficient exceeds the monocular camera detection coefficient threshold, the current single camera is replaced in a timely manner and the current device is debugged. After replacement, the starting operation time is coordinated and debugged with another single camera in the dual-camera group, and the collected images are segmented with the starting operation time after replacement as the demarcation point; if the monocular camera detection coefficient does not exceed the monocular camera detection coefficient threshold, it is inferred that the current single camera is operating normally; collect and statistically store the images of all single cameras in the dual-camera group.

[0014] As a preferred embodiment of the present invention, the monocular camera information includes the number of repeated pictures of the collected images corresponding to adjacent acquisition times when the object being photographed is in a static state and the interval distance of the positions of the object being photographed shown in the collected images corresponding to adjacent acquisition times when the object being photographed is in a moving state during the operation of the single camera.

[0015] As a preferred embodiment of the present invention, the binocular camera detection process is as follows:

[0016] After ensuring that the operation detections of all single cameras in the dual-camera group are qualified, perform binocular camera detection based on the statistically stored collected images, and infer the fitting degree of the cooperative photography of the dual-camera group through image analysis; collect static cooperation data and moving deviation data. If the static cooperation data exceeds the maximum value of the area excess value threshold range, perform shooting adjustment on the dual-camera group; if the static cooperation data does not exceed the minimum value of the area excess value threshold range, perform regulation on the shooting of the dual-camera group;

[0017] If the moving deviation data exceeds the maximum floating span threshold, perform shooting control on the dual-camera group to ensure the coverage ratio of the set shooting area of all single cameras; if the static cooperation data is within the area excess value threshold range and the moving deviation data does not exceed the maximum floating span threshold, it is inferred that the binocular camera detection is qualified.

[0018] As a preferred embodiment of the present invention, the static cooperation data and the moving deviation data are respectively the sum of the actual shooting areas of the object being photographed by the dual-camera group when the object being photographed is in a static state and the corresponding area excess value of the actual area of the object being photographed, and the maximum floating span value of the overlapping area of the object being photographed corresponding to different shooting times of the dual-camera group when the object being photographed is in a moving state.

[0019] As a preferred embodiment of the present invention, the binocular cooperation recognition and positioning process is as follows:

[0020] Obtain the captured images of the dual camera group, perform cooperative recognition and analysis on the captured images, determine the contour points of the photographed object according to the captured images of the dual camera group in combination with the trajectory of the photographed object, obtain the captured pictures corresponding to the capture moments in the captured images, and perform contour point collection on the basis of all the obtained captured pictures and set them as preset contour points, and screen the preset contour points according to all the captured pictures in the captured images. If the number of continuously appearing captured pictures corresponding to the preset contour points exceeds the set number threshold, mark the preset contour points as determined contour points; otherwise, mark them as stored contour points.

[0021] As a preferred embodiment of the present invention, connect the determined contour points to construct a three-dimensional model of the photographed object. If there is an abnormal contour shape in the real-time three-dimensional model, substitute the stored contour points into the model to infer whether the contour shape is normal. If it is, keep it; otherwise, delete it.

[0022] After determining the three-dimensional model of the photographed object, collect the morphological features and visual features of each display surface of the three-dimensional model. Among them, the morphological features are represented by the flatness and the area of the deformation region of the display surface; the visual features are represented by the color values at each position of the display surface.

[0023] As a preferred embodiment of the present invention, obtain the floating span of the parameter values in the morphological features and visual features corresponding to the shooting stage of the photographed object. If the floating span of the corresponding parameter values exceeds the set floating span threshold, or the values of the corresponding type of parameters cannot be restored to the value before the floating after the numerical span floats, it is inferred that the corresponding display surface is worn, and mark the corresponding position on the display surface as the deformed position; if the floating span of the corresponding parameter values does not exceed the set floating span threshold, it is inferred that the corresponding display surface is not worn, and mark the corresponding position on the display surface as the normal position; highlight the color of the deformed position on the three-dimensional model, and keep the color of the normal position constant.

[0024] As a preferred embodiment of the present invention, the process of evaluating the modeling accuracy is as follows:

[0025] Obtain the recognition deviation data and the recognition delay data. If the recognition deviation data exceeds the area deviation threshold, or the recognition delay data does not exceed the span numerical ratio threshold, reset the dual camera group of the current machine vision binocular measurement system; if the recognition deviation data does not exceed the area deviation threshold, and the recognition delay data exceeds the span numerical ratio threshold, continue shooting according to the current dual camera group.

[0026] As a preferred embodiment of the present invention, the recognition deviation data and the recognition delay data are respectively the range area deviation value between the color highlighted position of the real-time three-dimensional model and the actual verification position of the photographed object, and the numerical deviation span ratio corresponding to the upward span of the number parameter of the color highlighted position of the real-time three-dimensional model corresponding to the verification stage of the photographed object position and the actual number parameter of the verification stage.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] 1. In the present invention, the shooting of any camera is detected to infer whether the real-time image acquisition effect during monocular shooting meets the preset effect, thereby avoiding the decrease in the accuracy of the three-dimensional model construction of the photographed object by the entire measurement system due to the deviation of the acquired image during monocular shooting when analyzing the images collected by binoculars. At the same time, according to the monocular shooting detection, the operation stability of the binocular measurement system can be ensured, the image acquisition of any camera can be ensured to be accurate, and the photographed object can be accurately and efficiently identified and positioned.

[0029] By detecting binocular shooting, it is inferred whether the real-time shooting efficiency of the two cameras shooting in cooperation at different angles meets the actual requirements, avoiding shooting deviations when the binocular cameras shoot in cooperation, resulting in abnormal shooting positioning of the binocular cameras and the inability of the acquired images of the binocular cameras to cooperate with image analysis. By detecting binocular shooting, the accuracy of binocular shooting can be ensured, ensuring that the two cameras can cooperate synchronously from multiple angles when shooting an object, and improving the accuracy of object recognition and positioning of the visual binocular measurement system.

[0030] 2. In the present invention, image recognition is used for cooperative recognition and positioning to facilitate the accurate recognition and positioning of the photographed object, improving the recognition and monitoring accuracy of the machine vision binocular measurement system. At the same time, according to the binocular measurement system, the surface shape defects of the photographed object can be monitored, and the generation of shape defects can also be detected during the movement stage of the photographed object, which is beneficial for tracing the wear of the photographed object and can also repair the shape defects in time when wear occurs, ensuring the use quality of the photographed object to the greatest extent.

[0031] After the recognition and positioning by the machine vision binocular measurement system, three-dimensional modeling of the photographed object is carried out. By detecting the three-dimensional modeling, it is inferred whether there is a difference between the current photographed object and the actual object. By inferring the recognition and positioning points of the binocular measurement system, it is determined whether the cooperative recognition efficiency of the binoculars is qualified, thereby ensuring the accuracy of the three-dimensional modeling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] For the convenience of those skilled in the art to understand, the present invention will be further described below with reference to the accompanying drawings.

[0033] Figure 1 It is a flow chart of the present invention. Detailed implementation manners

[0034] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0035] As used herein, the mention of "embodiment" means that a specific feature, structure or characteristic described in connection with the embodiment may be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0036] Please refer to Figure 1 As shown, an automatic recognition and positioning method based on a machine vision binocular measurement system, and the specific steps of the automatic recognition and positioning method are as follows:

[0037] Monocular camera detection: The two cameras in the machine vision binocular measurement system take pictures at different angles, and the pictures taken by any one camera are detected to infer whether the real-time image acquisition effect during the monocular camera shooting meets the preset effect, so as to avoid the decline in the accuracy of the three-dimensional model construction of the photographed object by the entire measurement system due to the deviation of the acquired images of the monocular camera during binocular image acquisition and analysis. At the same time, according to the monocular camera detection, the operation stability of the binocular measurement system can be ensured, the image acquisition of any camera can be ensured to be accurate, and the photographed object can be accurately and efficiently recognized and positioned;

[0038] Binocular camera detection: After the monocular camera detection is completed and the detection is qualified, the binocular camera detection is carried out. Through the binocular camera detection, it is inferred whether the real-time shooting efficiency of the two cameras shooting in cooperation at different angles meets the actual requirements, and the shooting deviation is avoided when the binocular cameras cooperate to shoot, resulting in abnormal shooting positioning of the binocular cameras and the inability of the acquired images of the binocular cameras to cooperate with image analysis. Through the binocular camera detection, the accuracy of the binocular camera can be ensured, and it is ensured that the two cameras can cooperate synchronously from multiple angles when shooting an object, improving the accuracy of object recognition and positioning of the vision binocular measurement system;

[0039] Binocular cooperative recognition and positioning. Image recognition and analysis are carried out based on the collected images of binocular cameras. Cooperative recognition and positioning are performed through image recognition to facilitate accurate recognition and positioning of the photographed object, improving the recognition and monitoring accuracy of the machine vision binocular measurement system. At the same time, the surface shape defects of the photographed object can be monitored according to the binocular measurement system, and the generation of shape defects can also be detected during the movement stage of the photographed object, which is beneficial to the wear traceability of the photographed object and can also perform shape defect repair in a timely manner when wear occurs, ensuring the use quality of the photographed object to the greatest extent;

[0040] Modeling accuracy evaluation. After the recognition and positioning of the machine vision binocular measurement system, three-dimensional modeling of the photographed object is carried out. Whether there is a difference between the current photographed object and the actual object is inferred through three-dimensional modeling detection, and whether the binocular cooperative recognition efficiency is qualified is inferred through the recognition and positioning points of the binocular measurement system, so as to ensure the accuracy of the three-dimensional modeling efficiency;

[0041] The process of monocular camera detection is as follows:

[0042] Mark the two cameras in the machine vision binocular measurement system as a double camera group, and uniformly mark the cameras in the double camera group as single cameras; when the double camera group runs, monocular camera detection is carried out, that is, camera detection is carried out on the single camera;

[0043] Obtain the number of repeated pictures of the corresponding collected images at adjacent acquisition times in the static state of the photographed object during the operation of the single camera and the interval distance of the positions of the photographed object shown in the corresponding collected images at adjacent acquisition times in the moving state of the photographed object, and mark the number of repeated pictures of the corresponding collected images at adjacent acquisition times in the static state of the photographed object during the operation of the single camera and the interval distance of the positions of the photographed object shown in the corresponding collected images at adjacent acquisition times in the moving state of the photographed object as CD and JJ respectively; among them, the acquisition time represents the corresponding time point of the set focusing buffer period during the operation of the single camera, and if the picture clarity in the collected image corresponding to the focusing buffer time of the single camera is too low, it is directly excluded;

[0044] Mark the above-mentioned acquisition information as monocular camera information, and substitute it into the formula to obtain the monocular camera detection coefficient. The formula is: , where D is the monocular camera detection coefficient, h1 and h2 are preset proportionality coefficients, both h1 and h2 are greater than 1; cd represents the set value of the number of overlapping pictures; jj represents the set allowable interval distance value, e is the natural constant, and α is the error correction factor, with a value of 0.978;

[0045] Compare the monocular camera detection coefficient with the monocular camera detection coefficient threshold:

[0046] If the monocular camera detection coefficient exceeds the monocular camera detection coefficient threshold, it is inferred that the current single camera operation detection is abnormal. The current single camera is replaced in time and the current device is debugged. After replacement, the starting operation time is coordinated and debugged with the other single camera of the dual camera group, and the collected images are segmented with the starting operation time after replacement as the demarcation point.

[0047] If the monocular camera detection coefficient does not exceed the monocular camera detection coefficient threshold, it is inferred that the current single camera operation detection is normal; the collected images of all single cameras in the dual camera group are statistically stored.

[0048] The binocular camera detection process is as follows:

[0049] After ensuring that the operation detections of all single cameras in the dual camera group are qualified, binocular camera detection is performed based on the statistically stored collected images, and the fitting degree of the dual camera group's cooperative shooting is inferred through image analysis.

[0050] Obtain the sum of the actual shooting areas of the shooting objects of the dual camera group when the shooting object is in a static state and the excess value of the corresponding area of the actual area of the shooting object, and mark the sum of the actual shooting areas of the shooting objects of the dual camera group when the shooting object is in a static state and the excess value of the corresponding area of the actual area of the shooting object as the static cooperation data.

[0051] Obtain the maximum floating span value of the overlapping area of the shooting objects corresponding to different shooting times of the dual camera group when the shooting object is in a moving state, and mark the maximum floating span value of the overlapping area of the shooting objects corresponding to different shooting times of the dual camera group when the shooting object is in a moving state as the moving deviation data.

[0052] Compare the sum of the actual shooting areas of the shooting objects of the dual camera group when the shooting object is in a static state and the excess value of the corresponding area of the actual area of the shooting object, and the maximum floating span value of the overlapping area of the shooting objects corresponding to different shooting times of the dual camera group when the shooting object is in a moving state with the excess value threshold and the maximum floating span threshold respectively:

[0053] If the sum of the actual shooting areas of the shooting objects of the dual camera group when the shooting object is in a static state and the excess value of the corresponding area of the actual area of the shooting object exceeds the maximum value of the excess value threshold range, it is inferred that the binocular camera detection of the dual camera group is inefficient, and the shooting of the dual camera group is adjusted to reduce the overlapping area of the shooting object, avoid increasing the cooperation positioning recognition intensity of the shooting object image, and perform overlapping area regulation without affecting the overall shooting effect of the shooting object, that is, by adjusting the shooting angle of the single camera.

[0054] If the excess value of the sum of the actual shooting areas of the shooting object by the dual camera group compared to the corresponding area of the actual area of the shooting object when the shooting object is in a stationary state does not exceed the minimum value of the excess area value threshold range, it is inferred that the binocular camera detection of the dual camera group is abnormal, and the shooting of the dual camera group is adjusted to increase the coverage ratio of the shooting area of the shooting object, avoiding the situation where the coverage ratio of the shooting area of the shooting object does not meet the standard, which may lead to a decrease in the point recognition efficiency of the subsequent shooting object;

[0055] If the maximum floating span value of the overlapping area of the shooting object corresponding to different shooting moments of the dual camera group when the shooting object is in a moving state exceeds the maximum floating span threshold, it is inferred that the cooperation performance of the dual camera group is abnormal, and it is unable to ensure the coverage ratio of the shooting area of the shooting object when the state of the shooting object changes. Then, shooting control is performed on the dual camera group to ensure the coverage ratio of the shooting area set by all single cameras, and the shooting angle is adjusted in cooperation according to the real-time shooting area when the shooting object is moving. When the shooting area of any single camera decreases during the moving stage, the shooting angle of the other single camera is adjusted to expand the shooting area;

[0056] If the excess value of the sum of the actual shooting areas of the shooting object by the dual camera group compared to the corresponding area of the actual area of the shooting object when the shooting object is in a stationary state is within the excess area value threshold range, and the maximum floating span value of the overlapping area of the shooting object corresponding to different shooting moments of the dual camera group when the shooting object is in a moving state does not exceed the maximum floating span threshold, it is inferred that the binocular camera detection is qualified;

[0057] The binocular cooperation recognition and positioning process is as follows:

[0058] Obtain the acquisition images of the dual camera group, and perform cooperative recognition and analysis on the acquisition images. Determine the contour points of the shooting object based on the acquisition images of the dual camera group combined with the trajectory of the shooting object. That is, according to the moving trajectory or static position of the shooting object, obtain the acquisition pictures corresponding to the acquisition moments in the acquisition images, and perform contour point acquisition on all the obtained acquisition pictures and set them as preset contour points. Then, screen the preset contour points according to all the acquisition pictures in the acquisition images. If the number of acquisition pictures corresponding to the preset contour points that continuously appear exceeds the set number threshold, mark the preset contour points as determined contour points; otherwise, mark them as stored contour points. Connect the determined contour points to construct a three-dimensional model of the shooting object. If there is an abnormal contour shape in the real-time three-dimensional model, substitute the stored contour points into the model to infer whether the contour shape is normal. If it is normal, retain it; otherwise, delete it. Among them, an abnormal contour shape means that the contour shape is obviously unreasonable, such as contour overlap when adjacent contour points are connected, and a normal contour shape means that the contour shape is reasonable and the contour shape is similar to that of the same type of shooting object;

[0059] After determining the three-dimensional model of the photographed object, morphological features and visual features of each display surface of the three-dimensional model are collected. The morphological features are represented by the flatness and the area of the deformation region of the display surface; the visual features are represented by the color values at each position of the display surface. The floating ranges of the numerical values of the morphological features and visual features corresponding to the photographing stage of the photographed object are obtained. If the floating range of the corresponding parameter value exceeds the set floating range threshold, or if the value of the corresponding type of parameter cannot be restored to the value before the floating after the floating of the numerical range, it is inferred that the corresponding display surface is worn, and the corresponding position in the display surface is marked as the deformation position;

[0060] If the floating range of the corresponding parameter value does not exceed the set floating range threshold, it is inferred that the corresponding display surface is not worn, and the corresponding position in the display surface is marked as the normal position; the deformation position is highlighted in color on the three-dimensional model, and the color of the normal position remains constant;

[0061] The process of evaluating the modeling accuracy is as follows:

[0062] After completing the cooperative recognition and positioning, the modeling accuracy of the constructed three-dimensional model is evaluated. The range area deviation value between the color-highlighted position of the real-time three-dimensional model and the actual verification position of the photographed object is obtained, and the range area deviation value between the color-highlighted position of the real-time three-dimensional model and the actual verification position of the photographed object is marked as the recognition deviation data;

[0063] The corresponding span numerical ratio between the rising span of the number parameter of the color-highlighted position of the real-time three-dimensional model and the numerical deviation span of the actual number parameter in the verification stage during the position verification stage of the photographed object is obtained, and the corresponding span numerical ratio between the rising span of the number parameter of the color-highlighted position of the real-time three-dimensional model and the numerical deviation span of the actual number parameter in the verification stage during the position verification stage of the photographed object is marked as the recognition delay data, where the number parameter is represented by numerical parameters such as the area and number of the color-highlighted positions;

[0064] And the range area deviation value between the color-highlighted position of the real-time three-dimensional model and the actual verification position of the photographed object, and the corresponding span numerical ratio between the rising span of the number parameter of the color-highlighted position of the real-time three-dimensional model and the numerical deviation span of the actual number parameter in the verification stage during the position verification stage of the photographed object are respectively compared with the area deviation threshold and the span numerical ratio threshold:

[0065] If the range area deviation value between the color highlighted position of the real-time three-dimensional model and the actual verification position of the photographed object exceeds the area deviation threshold, or the ratio of the span of the increase in the number parameter of the color highlighted position of the real-time three-dimensional model corresponding to the photographed object position verification stage to the numerical deviation span of the actual number parameter in the verification stage does not exceed the span ratio threshold, it is inferred that the three-dimensional modeling accuracy of the photographed object is poor, and the current machine vision binocular measurement system is reset for the double camera group, that is, the specifications and shooting angles are reset;

[0066] If the range area deviation value between the color highlighted position of the real-time three-dimensional model and the actual verification position of the photographed object does not exceed the area deviation threshold, and the ratio of the span of the increase in the number parameter of the color highlighted position of the real-time three-dimensional model corresponding to the photographed object position verification stage to the numerical deviation span of the actual number parameter in the verification stage exceeds the span ratio threshold, it is inferred that the three-dimensional modeling accuracy of the photographed object is high, and the shooting is continued according to the current double camera group;

[0067] The above formulas are all obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value. The coefficients in the formula are set by those skilled in the art according to the actual situation;

[0068] When the present invention is in use, single-eye camera detection is performed, and two cameras in the machine vision binocular measurement system are used to take pictures at different angles, and the pictures taken by any one camera are detected; binocular camera detection is performed after the single-eye camera detection is completed and qualified; binocular cooperation recognition and positioning are performed, and image recognition analysis is performed according to the collected images of the binocular cameras, and cooperation recognition and positioning are performed through image recognition; modeling accuracy evaluation is performed. After the recognition and positioning by the machine vision binocular measurement system, three-dimensional modeling of the photographed object is performed, and it is inferred whether there is a difference between the current photographed object and the actual object through three-dimensional modeling detection.

[0069] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to only the specific embodiments. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the present invention, so that those skilled in the art in the relevant technical field can understand and utilize the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. An automatic recognition and positioning method based on a machine vision binocular measurement system, characterized in that The steps of the automatic recognition and positioning method are as follows: Monocular camera detection: Two cameras in the machine vision binocular measurement system take pictures at different angles, and the pictures taken by any one camera are detected; Binocular camera detection: After the monocular camera detection is completed and the detection is qualified, the binocular camera detection is carried out; The process of binocular camera detection is as follows: After ensuring that the operation detections of all single cameras in the binocular camera group are qualified, binocular camera detection is carried out according to the collected images stored statistically, and the fitting degree of the binocular camera group's cooperative shooting is inferred through image analysis; Static cooperation data and moving deviation data are collected. If the static cooperation data exceeds the maximum value of the area excess value threshold range, the shooting of the binocular camera group is adjusted; If the static cooperation data does not exceed the minimum value of the area excess value threshold range, the shooting of the binocular camera group is regulated; If the moving deviation data exceeds the maximum floating span threshold, the shooting of the binocular camera group is controlled to ensure the coverage ratio of the shooting areas set by all single cameras; if the static cooperation data is within the area excess value threshold range and the moving deviation data does not exceed the maximum floating span threshold, it is inferred that the binocular camera detection is qualified; The static cooperation data and the moving deviation data are respectively the sum of the actual shooting areas of the shooting objects of the binocular camera group when the shooting object is in a static state and the corresponding area excess value of the actual area of the shooting object, and the maximum floating span value of the overlapping areas of the shooting objects corresponding to different shooting moments of the binocular camera group when the shooting object is in a moving state; Binocular cooperation recognition and positioning: Image recognition and analysis are carried out according to the collected images of the binocular camera, and cooperation recognition and positioning are carried out through image recognition; The process of binocular cooperation recognition and positioning is as follows: The collected images of the binocular camera group are obtained, and cooperation recognition analysis is carried out on the collected images. The contour points of the shooting object are determined according to the collected images of the binocular camera group combined with the trajectory of the shooting object. The collected pictures corresponding to the collected moments in the collected images are obtained, and the contour points are collected according to all the obtained collected pictures and set as the preset contour points. The preset contour points are screened according to all the collected pictures in the collected images. If the number of consecutive collected pictures corresponding to the preset contour points exceeds the set number threshold, the preset contour points are marked as determined contour points, otherwise, they are marked as stored contour points; Modeling accuracy evaluation: After the recognition and positioning by the machine vision binocular measurement system, three-dimensional modeling of the shooting object is carried out, and whether there is a difference between the current shooting object and the actual object is inferred through three-dimensional modeling detection.

2. The automatic recognition and positioning method of a machine vision binocular measurement system according to claim 1, characterized in that, The process of monocular camera detection is as follows: Mark the two cameras in the machine vision binocular measurement system as a dual-camera group, and uniformly mark the cameras in the dual-camera group as single cameras; collect monocular camera information, and substitute it into the calculation to obtain the monocular camera detection coefficient. If the monocular camera detection coefficient exceeds the monocular camera detection coefficient threshold, the current single camera will be replaced in time and the current device will be debugged. After replacement, the starting operation time will be coordinated and debugged with the other single camera in the dual-camera group, and the collected images will be segmented with the starting operation time after replacement as the demarcation point; if the monocular camera detection coefficient does not exceed the monocular camera detection coefficient threshold, it is inferred that the current single camera is operating normally. Statistically store the collected images of all single cameras in the dual-camera group.

3. An automatic recognition and positioning method based on a machine vision binocular measurement system according to claim 1, characterized in that, The monocular camera information includes the number of repeated pictures of the collected images corresponding to adjacent acquisition times when the object is in a stationary state during the operation of the single camera and the interval distance of the displayed object position in the collected images corresponding to adjacent acquisition times when the object is in a moving state.

4. An automatic recognition and positioning method for a machine vision binocular measurement system according to claim 1, characterized in that, Connect the determined contour points to construct a three-dimensional model of the object to be photographed. If there are abnormal contour shapes in the real-time three-dimensional model, substitute the stored contour points into the model to infer whether the contour shape is normal. If it is, keep it; otherwise, delete it. After determining the three-dimensional model of the object to be photographed, collect the morphological features and visual features of each display surface of the three-dimensional model. Among them, the morphological features are represented by the flatness and the area of the deformation region of the display surface; the visual features are represented by the color values of each position on the display surface.

5. An automatic recognition and positioning method for a machine vision binocular measurement system according to claim 4, characterized in that Obtain the floating span of the parameter values in the morphological features and visual features corresponding to the shooting stage of the object to be photographed. If the floating span of the corresponding parameter values exceeds the set floating span threshold, or the values of the corresponding type of parameters cannot be restored to before the floating after the numerical span floats, it is inferred that the corresponding display surface is worn, and the corresponding position on the display surface is marked as a deformed position; if the floating span of the corresponding parameter values does not exceed the set floating span threshold, it is inferred that the corresponding display surface is not worn, and the corresponding position on the display surface is marked as a normal position; highlight the color of the deformed position on the three-dimensional model, and the color of the normal position remains constant.

6. An automatic recognition and positioning method based on a machine vision binocular measurement system according to claim 1, characterized in that, The process of evaluating the modeling accuracy is as follows: Obtain the recognition deviation data and the recognition delay data. If the recognition deviation data exceeds the area deviation threshold, or the recognition delay data does not exceed the span numerical ratio threshold, reset the dual-camera group of the current machine vision binocular measurement system; if the recognition deviation data does not exceed the area deviation threshold and the recognition delay data exceeds the span numerical ratio threshold, continue shooting according to the current dual-camera group.

7. An automatic recognition and positioning method for a machine vision binocular measurement system according to claim 6, characterized in that, The recognition deviation data and the recognition delay data are respectively the range area deviation value between the color-highlighted position of the real-time three-dimensional model and the actual verification position of the object to be photographed, and the numerical deviation span ratio corresponding to the upward span of the number parameter of the color-highlighted position of the real-time three-dimensional model during the object position verification stage and the actual number parameter during the verification stage.

Images