Intelligent wire harness terminal clamping system and method based on machine vision

Abstract

The invention relates to the technical field of wire harness terminal intelligent clamping, and discloses a wire harness terminal intelligent clamping system and method based on machine vision, and the system comprises a data collection module, an image analysis module, an image optimization module, a pose analysis module, a pose optimization module, a path analysis module, a path optimization module, and a feedback module. Compared with a traditional wire harness clamping system, the intelligent clamping system based on machine vision achieves multi-dimensional information fusion and collaborative optimization in key links such as image recognition, pose judgment and path planning. Through layer-by-layer processing and feedback of the eight modules, the system achieves higher recognition precision, lower error rate and higher operation flexibility, and effectively solves the problems that a traditional method is poor in adaptability to complex environments, low in repeated positioning precision and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent clamping of wiring harness terminals, in particular to a wiring harness terminal intelligent clamping system and method based on machine vision. BACKGROUND

[0002] In the fields of modern automobile manufacturing, home appliance assembly and industrial control system, wiring harness as a key electrical connection unit, its processing quality directly affects the performance and reliability of the whole machine. In the production process of wiring harness, the grasping and positioning operation of the terminal is usually completed by manual or semi-automatic mode, which has problems of low efficiency, large error, high repetitive labor intensity and the like.

[0003] Although some terminal recognition and clamping systems based on machine vision have appeared at present, there are still several technical bottlenecks in actual application, which are difficult to meet the production requirements under complex conditions. On the one hand, limited by the depth perception ability of the traditional two-dimensional vision system, the existing technology is prone to recognition error when facing the rotation, tilt or feature blur of the terminal posture caused by the reflection of the metal surface, thereby causing clamping deviation or failure. On the other hand, in the scene of terminal stacking or partial occlusion, the system is difficult to accurately segment and extract the position of multiple terminals, resulting in the inability to realize effective single terminal recognition and clamping control.

[0004] Therefore, we propose a wiring harness terminal intelligent clamping system and method based on machine vision to solve the above problems. SUMMARY

[0005] The present application aims to provide a wiring harness terminal intelligent clamping system and method based on machine vision to solve the above problems in the background art. On the one hand, limited by the depth perception ability of the traditional two-dimensional vision system, the existing technology is prone to recognition error when facing the rotation, tilt or feature blur of the terminal posture caused by the reflection of the metal surface, thereby causing clamping deviation or failure. On the other hand, in the scene of terminal stacking or partial occlusion, the system is difficult to accurately segment and extract the position of multiple terminals, resulting in the inability to realize effective single terminal recognition and clamping control.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a wiring harness terminal intelligent clamping system based on machine vision, comprising a data acquisition module, an image analysis module, an image optimization module, a pose analysis module, a pose optimization module, a path analysis module, a path optimization module and a feedback module;

[0007] The data acquisition module is used for data detection and acquisition of the region, and pre-processing of the collected data to generate a first data set, a second data set and a third data set;

[0008] The image analysis module is used for data coupling of the first data set, so as to generate an image recognition coefficient TXB, data analysis is performed on the image recognition coefficient TXB, and an image optimization data set is generated;

[0009] The image optimization module is used for data coupling of the image optimization data set, so as to generate an image optimization coefficient TXY, and analysis is performed on the image optimization coefficient TXY;

[0010] The pose analysis module is used for data coupling of the second data set, so as to generate a pose recognition coefficient WZB, analysis is performed on the pose recognition coefficient WZB, and a pose optimization data set is generated;

[0011] The pose optimization module is used for data coupling of the pose optimization data set, so as to generate a pose optimization coefficient WZY, and analysis is performed on the pose optimization coefficient WZY;

[0012] The path analysis module is used for data coupling of the third data set, so as to generate a route recognition coefficient LJB, analysis is performed on the route recognition coefficient LJB, and a pose optimization data set is generated;

[0013] The path optimization module is used for data coupling of the pose optimization data set, so as to generate a path optimization coefficient LJY, and analysis is performed on the path optimization coefficient LJY;

[0014] The feedback module is used for feeding back the results to a visual terminal.

[0015] Preferably, the data acquisition module comprises an information acquisition unit and an information preprocessing unit;

[0016] The information acquisition unit is used for data detection and acquisition of an equipment region, and the data comprises an image gray value matrix A, a blue channel average value B, a gray co-occurrence matrix contrast C, and a region of interest area D, a terminal center point position E in a camera coordinate system, a terminal main shaft and horizontal axis included angle F, a distance from the camera depth G, and a terminal pitch angle H, a starting pose I, a target pose J, an interpolation trajectory function K, and a motion speed limit L;

[0017] The information preprocessing unit is used for preprocessing and dimensionless of the acquired data, and the processed data is arranged, so as to generate a first data set, a second data set, and a third data set;

[0018] The first data set comprises the image gray value matrix A, the blue channel average value B, the gray co-occurrence matrix contrast C, and the region of interest area D;

[0019] The second data set comprises the terminal center point position E in the camera coordinate system, the terminal main shaft and horizontal axis included angle F, the distance from the camera depth G, and the terminal pitch angle H;

[0020] The third data set includes the starting pose I, the target pose J, the interpolation trajectory function K, and the motion speed limit L.

[0021] Preferably, the image analysis module includes an image analysis calculation unit and an image analysis processing unit;

[0022] The image analysis and calculation unit is used to extract data from the first data set, including the image gray value matrix A, the blue channel average value B, the gray level co-occurrence matrix contrast C, and the area of ​​the region of interest D, and input the extracted multiple data into the pre-trained deep learning framework, perform feature fusion through a multi-layer neural network, and then calculate the image recognition coefficient TXB. The specific calculation formula is as follows:

[0023]

[0024] Where: image gray value matrix A, blue channel average value B, gray level co-occurrence matrix contrast C and area of ​​the region of interest D;

[0025] The image analysis processing unit is used to perform data analysis on the image recognition coefficient TXB and determine whether the current image needs to be optimized based on the analysis results. The specific analysis method is as follows:

[0026] When TXB≤0.56, it means that the current image does not need to be optimized;

[0027] When TXB>0.56, it means that the current image needs to be optimized and an optimized image data set is generated, including the optimized image gray value matrix YA, the optimized blue channel average value YB, the optimized gray co-occurrence matrix contrast YC, and the optimized region of interest area YD.

[0028] Preferably, the image optimization module includes an image optimization calculation unit and an image optimization processing unit;

[0029] The image optimization calculation unit is used to extract data from the optimized image data group, including optimizing the image gray value matrix YA, optimizing the blue channel average value YB, optimizing the grayscale co-occurrence matrix contrast YC, and optimizing the area of ​​the region of interest YD, and inputting the extracted multiple data into a pre-trained deep learning framework, performing feature fusion through a multi-layer neural network, and then calculating the image optimization coefficient TXY. The specific calculation formula is as follows;

[0030]

[0031] Where: YA is the optimized image gray value matrix, YB is the optimized blue channel average, YC is the optimized gray level co-occurrence matrix contrast, and YD is the optimized region of interest area;

[0032] The image optimization processing unit is used to perform data analysis on the image optimization coefficient TXY obtained by calculation, and judge whether the image optimization effect is qualified according to the analysis result. The specific steps are as follows:

[0033] When TXY≤0.65, it means that the current image optimization process is qualified and the next adjustment is continued;

[0034] When TXY>0.65, it means that the current image optimization process is unqualified and needs to be optimized again.

[0035] Preferably, the posture analysis module includes a posture analysis calculation unit and a posture analysis processing unit;

[0036] The posture analysis and calculation unit is used to extract data from the second data group, including the position E of the terminal center point in the camera coordinate system, the angle F between the terminal main axis and the horizontal axis, the depth G from the camera, and the pitch angle H of the terminal, and input the extracted multiple data into the pre-trained deep learning framework, perform feature fusion through a multi-layer neural network, and then calculate the posture recognition coefficient WZB. The specific calculation formula is as follows:

[0037]

[0038] Where: E is the position of the terminal center point in the camera coordinate system, F is the angle between the terminal main axis and the horizontal axis, G is the depth from the camera, and H is the terminal pitch angle;

[0039] The posture analysis processing unit is used to perform data analysis on the posture identification coefficient WZB, and based on the analysis results, determine whether the current posture needs to be optimized. The specific analysis steps are as follows;

[0040] When WZB≤0.55, it means that the current posture does not need to be optimized;

[0041] When WZB>0.55, it means that the current posture needs to be optimized and an optimized posture data group is generated, including optimizing the position YE of the terminal center point in the camera coordinate system, optimizing the angle YF between the terminal main axis and the horizontal axis, optimizing the depth YG from the camera, and optimizing the terminal pitch angle YH.

[0042] Preferably, the posture optimization module includes a posture optimization calculation unit and a posture optimization processing unit;

[0043] The posture optimization calculation unit is used to extract data from the posture optimization data group, including optimizing the position YE of the terminal center point in the camera coordinate system, optimizing the angle YF between the terminal main axis and the horizontal axis, optimizing the depth YG from the camera, and optimizing the terminal pitch angle YH. The extracted multiple data are input into the pre-trained deep learning framework, and feature fusion is performed through a multi-layer neural network to calculate the posture optimization coefficient WZY. The specific calculation formula is as follows:

[0044]

[0045] Where: YE is the position of the center point of the optimized terminal in the camera coordinate system, YF is the angle between the main axis of the optimized terminal and the horizontal axis, YG is the depth of the optimized distance from the camera, and YH is the pitch angle of the optimized terminal;

[0046] The posture optimization processing unit is used to perform data analysis on the posture optimization coefficient WZY obtained by calculation, and judge whether the current posture optimization is qualified according to the analysis results. The specific steps are as follows:

[0047] When WZY≤0.65, it means that the current posture optimization process is qualified and the next adjustment is continued;

[0048] When WZY>0.65, it means that the current pose optimization process is unqualified and needs to be re-optimized.

[0049] Preferably, the path analysis module includes a path analysis calculation unit and a path analysis processing unit;

[0050] The path analysis calculation unit is used to extract data from the third data set, including the starting posture I, the target posture J, the interpolation trajectory function K, and the movement speed limit L, and input the extracted multiple data into the pre-trained deep learning framework, perform feature fusion through a multi-layer neural network, and thus calculate the route identification coefficient LJB. The specific calculation formula is as follows:

[0051]

[0052] Where: I is the starting pose, J is the target pose, K is the interpolation trajectory function, and L is the motion speed limit;

[0053] The path analysis processing unit is used to perform data analysis on the calculated route identification coefficient LJB and determine whether the current path needs to be optimized based on the analysis results. The specific steps are as follows:

[0054] When LJB≤0.55, it means that the current path does not need to be optimized;

[0055] When LJB>0.55, it means that the current path needs to be optimized and an optimized path data set is generated, including the optimized starting pose YI, the optimized target pose YJ, the optimized interpolation trajectory function YK, and the optimized motion speed limit YL.

[0056] Preferably, the path optimization module includes a path optimization calculation unit and a path optimization processing unit;

[0057] The path optimization calculation unit is used to extract data from the path optimization data group, including optimizing the starting posture YI, optimizing the target posture YJ, optimizing the interpolation trajectory function YK, and optimizing the movement speed limit YL. The extracted multiple data are input into the pre-trained deep learning framework, and the path optimization coefficient LJY is obtained by feature fusion calculation through a multi-layer neural network. The specific calculation method is as follows:

[0058]

[0059] Where: YI is the optimized starting pose, YJ is the optimized target pose, YK is the optimized interpolation trajectory function, and YL is the optimized motion speed limit;

[0060] The path optimization processing unit is used to perform data analysis on the calculated path optimization coefficient LJY and generate analysis results. According to the analysis results, it is determined whether the current path optimization is qualified. The specific steps are as follows:

[0061] When LJY≤0.65, it means that the current path optimization process is qualified and the next adjustment is continued;

[0062] When LJY>0.65, it means that the current path optimization process is unqualified and needs to be re-optimized.

[0063] This application also includes a method for intelligently clamping wiring harness terminals based on machine vision, the specific steps of which are as follows:

[0064] S1. Detect and collect data in the region through a data collection module, and pre-process the collected data to generate a first data group, a second data group, and a third data group;

[0065] S2. Performing data coupling on the first data set using the image analysis module to generate an image recognition coefficient TXB, performing data analysis on the image recognition coefficient TXB, and generating an image optimization data set. Performing data coupling on the image optimization data set using the image optimization module to generate an image optimization coefficient TXY, and performing analysis on the image optimization coefficient TXY.

[0066] S3. Perform data coupling on the second data set through the posture analysis module to generate a posture identification coefficient WZB, analyze the coefficient, and generate a posture optimization data set. Perform data coupling on the posture optimization data set through the posture optimization module to generate a posture optimization coefficient WZY, and analyze the coefficient.

[0067] S4. Data coupling is performed on the third data set by the path analysis module to generate a route identification coefficient LJB, which is analyzed to generate a posture optimization data set. Data coupling is performed on the posture optimization data set by the path optimization module to generate a path optimization coefficient LJY, which is also analyzed.

[0068] S5. Feedback each result to the visualization terminal through the feedback module.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] 1. Compared to traditional wire harness clamping systems, this machine vision-based intelligent clamping system achieves multi-dimensional information fusion and collaborative optimization in key aspects such as image recognition, posture judgment, and path planning. Through layered processing and feedback from eight modules, the system achieves higher recognition accuracy, lower error rates, and greater operational flexibility, effectively addressing issues such as poor adaptability to complex environments and low repeatability of positioning methods.

[0071] 2. When an image requires optimization, the system automatically generates an optimized image dataset consisting of an optimized image grayscale value matrix YA, an optimized blue channel average YB, an optimized grayscale co-occurrence matrix contrast YC, and an optimized region of interest area YD. This provides the image optimization module with a clear and precise processing target and parameter basis. This mechanism ensures that subsequent image optimization operations are targeted and accurate, improving optimization efficiency and image usability. By combining deep learning with mathematical modeling to calculate the image recognition coefficient TXB, this approach not only achieves deep coupling of visual features but also offers strong interpretability and computational simplicity. While ensuring performance, this method is also suitable for deployment in resource-constrained embedded devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a system flow chart of the present invention.

[0073] Figure 2 A diagram showing the steps of the method of the present invention.

[0074] In the figure: 1. Data acquisition module; 11. Information acquisition unit; 12. Information preprocessing unit; 2. Image analysis module; 21. Image analysis calculation unit; 22. Image analysis processing unit; 3. Image optimization module; 31. Image optimization calculation unit; 32. Image optimization processing unit; 4. Posture analysis module; 41. Posture analysis calculation unit; 42. Posture analysis processing unit; 5. Posture optimization module; 51. Posture optimization calculation unit; 52. Posture optimization processing unit; 6. Path analysis module; 61. Path analysis calculation unit; 62. Path analysis processing unit; 7. Path optimization module; 71. Path optimization calculation unit; 72. Path optimization processing unit; 8. Feedback module. DETAILED DESCRIPTION

[0075] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0076] Example 1: Please refer to Figure 1 , a wire harness terminal intelligent clamping system based on machine vision, including a data acquisition module 1, an image analysis module 2, an image optimization module 3, a posture analysis module 4, a posture optimization module 5, a path analysis module 6, a path optimization module 7 and a feedback module 8;

[0077] The data acquisition module 1 is used to detect and collect data in the area, and pre-process the collected data to generate the first data group, the second data group and the third data group;

[0078] The image analysis module 2 performs data coupling on the first data set to generate an image recognition coefficient TXB, performs data analysis on the first data set, and generates an image optimization data set;

[0079] The image optimization module 3 is used to perform data coupling on the image optimization data group, thereby generating an image optimization coefficient TXY and analyzing it;

[0080] The posture analysis module 4 is used to perform data coupling on the second data set to generate a posture identification coefficient WZB, analyze it, and generate a posture optimization data set;

[0081] The posture optimization module 5 is used to perform data coupling on the posture optimization data group, thereby generating the posture optimization coefficient WZY and analyzing it;

[0082] The path analysis module 6 is used to perform data coupling on the third data set to generate a route identification coefficient LJB, analyze the coefficient, and generate a posture optimization data set;

[0083] The path optimization module 7 is used to perform data coupling on the posture optimization data group to generate the path optimization coefficient LJY and analyze it;

[0084] The feedback module 8 is used to feed back various results to the visualization terminal.

[0085] In this embodiment, Data Acquisition Module 1: This module serves as the starting point for the system's information flow, primarily utilizing visual sensing equipment to perform real-time monitoring and data collection of the target operating area. The collected raw images and pose data undergo preliminary preprocessing and are then divided into three types of structured data: the first data set is used for image processing, the second for pose analysis, and the third for path planning. This module's beneficial effect lies in its efficient acquisition and organized organization of raw data, significantly improving the targeted processing efficiency of subsequent modules and enhancing the overall system's responsiveness and data utilization.

[0086] Image Analysis Module 2: This module performs in-depth image analysis on the first data set, extracting key image features of the wiring harness terminals using an image recognition algorithm and generating an image recognition coefficient TXB. These extracted features form the image optimization data set, serving as the basis for image optimization. This module effectively separates terminal image information from complex backgrounds, improving feature extraction accuracy and providing reliable data support for subsequent precise clamping operations.

[0087] Image Optimization Module 3: This module further enhances and reconstructs the image optimization data set, calculating and generating the image optimization coefficients TXY. Through techniques such as image enhancement, denoising, and reconstruction, this module significantly improves image quality, eliminates redundant features, and sharpens key boundaries. This module enhances the robustness and adaptability of the image recognition system, maintaining high recognition accuracy even under complex lighting and occlusion conditions.

[0088] Posture Analysis Module 4: This module performs 3D posture analysis on the second data set, extracts the terminal's spatial position information, generates 3D posture characteristic coefficients (WZB), and constructs a posture optimization data set. Its core technologies, including point cloud registration and depth estimation algorithms, accurately determine the target's position and orientation in space. This module ensures accurate recognition even in complex and changing postures, providing a key guarantee for optimal clamping path planning.

[0089] Posture Optimization Module 5: Based on posture analysis, this module further optimizes the terminal's spatial position and generates optimized posture coefficients WZY. Combining deep learning and model fitting techniques, it corrects deviations and abnormal postures in the initial posture, enhancing the accuracy and stability of command execution. This module significantly improves the reliability and fault tolerance of the system's clamping action.

[0090] Path Analysis Module 6: Based on the third data set, this module extracts path features and generates a path identification coefficient (LJB) for evaluating candidate clamping trajectories. By applying path planning algorithms such as A*, DWA dynamic window method, and RRT, the clamping path is modeled and output as a path optimization data set. This module primarily provides a reasonable path starting point, obstacle avoidance mechanism, and clamping sequence planning to ensure efficient and safe path execution.

[0091] Path Optimization Module 7: This module performs advanced analysis and processing on the path optimization data set to generate the path optimization coefficient LJY. This module combines trajectory smoothing, dynamic simulation, and path reconstruction technologies to optimize path continuity and stability, effectively reducing errors, jitter, and uncertainty during path execution. This module enables smoother gripping paths and significantly reduces motion completion time.

[0092] Feedback Module 8: This module integrates the key data and coefficients generated by each module and provides real-time feedback to the terminal display interface. Users can monitor the results of image processing, pose recognition, and path planning through this interface, adjusting parameters or performing operational interventions as necessary. The introduction of this module establishes a closed-loop control mechanism for the system, improving the efficiency of human-computer interaction and the safety of system operation.

[0093] Compared to traditional harness clamping systems, this machine vision-based intelligent clamping system achieves multi-dimensional information fusion and collaborative optimization in key areas such as image recognition, posture determination, and path planning. Through layered processing and feedback from eight modules, the system achieves higher recognition accuracy, lower error rates, and greater operational flexibility, effectively addressing issues such as poor adaptability to complex environments and low repeatability of positioning methods.

[0094] Example 2: Please refer to Figure 1 , the data acquisition module 1 includes an information acquisition unit 11 and an information preprocessing unit 12;

[0095] The information acquisition unit 11 is used to detect and collect data on the device area, including the image gray value matrix A, the blue channel average B, the gray level co-occurrence matrix contrast C, and the area of ​​the region of interest D, the position E of the terminal center point in the camera coordinate system, the angle F between the terminal main axis and the horizontal axis, the depth G from the camera, and the terminal pitch angle H, the starting posture I, the target posture J, the interpolation trajectory function K, and the movement speed limit L;

[0096] The information preprocessing unit 12 is used to preprocess and dimensionlessly process the collected data, and to organize the processed data to generate a first data group, a second data group, and a third data group;

[0097] The first data set includes the image gray value matrix A, the blue channel average value B, the gray level co-occurrence matrix contrast C, and the area of ​​the region of interest D;

[0098] The second data set includes the position E of the terminal center point in the camera coordinate system, the angle F between the terminal main axis and the horizontal axis, the depth G from the camera, and the pitch angle H of the terminal;

[0099] The third data set includes the starting pose I, the target pose J, the interpolation trajectory function K, and the motion speed limit L.

[0100] In this embodiment, the system uses information acquisition unit 11 to not only collect basic visual features such as the image grayscale matrix A and the blue channel average B, but also incorporates texture and target saliency metrics such as grayscale co-occurrence matrix contrast C and region of interest area D, significantly enhancing the multidimensional characterization of target features. In the pose dimension, not only spatial position E and depth G are collected, but also posture information such as the main axis angle F and pitch angle H are combined to further refine the 3D modeling parameters.

[0101] Unlike the traditional method that relies solely on real-time path calculation, this module introduces parameters such as the starting posture I, the target posture J, the interpolation trajectory function K, and the motion speed limit L to form a third data group. This provides sufficient prior constraints and simulation basis for the path analysis and optimization module in advance, enhancing the foresight and controllability of the system path planning.

[0102] The information preprocessing unit 12 uniformly preprocesses and dimensionlessly transforms multi-source heterogeneous data, significantly improving data consistency and comparability. Subsequently, the data is divided into the first, second, and third data groups according to task type. This allows subsequent modules in the system to efficiently access the required data, reducing redundant processing and cross-interference, and improving overall parallel processing efficiency and module decoupling. The three constructed data groups cover the three core functional dimensions of visual recognition, spatial positioning, and trajectory control, providing stable, accurate, and structured data support for subsequent processing modules. This fundamentally improves the accuracy of the system's image recognition, the precision of posture judgment, and the smoothness and robustness of path execution.

[0103] Example 3: Please refer to Figure 1 , the image analysis module 2 includes an image analysis calculation unit 21 and an image analysis processing unit 22;

[0104] The image analysis and calculation unit 21 is used to extract data from the first data set, including the image gray value matrix A, the blue channel average value B, the gray level co-occurrence matrix contrast C, and the area of ​​the region of interest D. The extracted data are input into the pre-trained deep learning framework, and feature fusion is performed through a multi-layer neural network to calculate the image recognition coefficient TXB. The specific calculation formula is as follows:

[0105]

[0106] Where: image gray value matrix A, blue channel average value B, gray level co-occurrence matrix contrast C and area of ​​the region of interest D;

[0107] The image analysis processing unit 22 is used to perform data analysis on the image recognition coefficient TXB and determine whether the current image needs to be optimized based on the analysis results. The specific analysis method is as follows:

[0108] When TXB≤0.56, it means that the current image does not need to be optimized;

[0109] When TXB>0.56, it means that the current image needs to be optimized and an optimized image data set is generated, including the optimized image gray value matrix YA, the optimized blue channel average value YB, the optimized gray co-occurrence matrix contrast YC, and the optimized region of interest area YD.

[0110] In the embodiment: the image analysis calculation unit 21 extracts four types of features, including the image gray value matrix A, the blue channel average value B, the gray co-occurrence matrix contrast C, and the area of the region of interest D, and fuses these features by means of the pre-trained deep learning framework and the multi-layer neural network, thereby calculating the image recognition coefficient TXB which is comprehensive and has high discrimination. This feature fusion strategy effectively improves the recognition accuracy of the wire harness terminal target under complex background, low contrast, or image noise conditions.

[0111] By setting the threshold judgment mechanism, the image analysis processing unit 22 can automatically determine whether the current image needs to be optimized according to the TXB value. This method avoids the subjectivity and uncertainty of relying on human experience judgment in traditional image systems, and improves the intelligence and consistency of the system in batch operations or continuous recognition.

[0112] When the image needs to be optimized, the system automatically generates an optimized image data set containing the optimized image gray value matrix YA, the optimized blue channel average value YB, the optimized gray co-occurrence matrix contrast YC, and the optimized area of the region of interest YD, providing a clear and explicit processing object and parameter basis for the image optimization module 3. This mechanism ensures that the subsequent image optimization operation can be targeted and precise, improving optimization efficiency and image usability. The image recognition coefficient TXB is calculated by combining deep learning with mathematical formula modeling, which not only realizes deep coupling of visual features, but also has strong explainability and computational simplicity. This method not only guarantees performance, but also is beneficial to deployment in resource-constrained embedded devices.

[0113] Embodiment four: please refer to Figure 1 , the image optimization module 3 includes an image optimization calculation unit 31 and an image optimization processing unit 32;

[0114] The image optimization calculation unit 31 is used for data extraction of the optimized image data set, including the optimized image gray value matrix YA, the optimized blue channel average value YB, the optimized gray co-occurrence matrix contrast YC, and the optimized area of the region of interest YD, and inputting the extracted multiple data into the pre-trained deep learning framework, and fusing the features by the multi-layer neural network, and then calculating and obtaining the image optimization coefficient TXY, the specific calculation formula is as follows:

[0115]

[0116] In the formula: YA is the optimized image gray value matrix, YB is the optimized blue channel average value, YC is the optimized gray co-occurrence matrix contrast, and YD is the optimized area of the region of interest;

[0117] The image optimization processing unit 32 is used to perform data analysis on the image optimization coefficient TXY obtained by calculation, and judge whether the image optimization effect is qualified according to the analysis result. The specific steps are as follows:

[0118] When TXY≤0.65, it means that the current image optimization process is qualified and the next adjustment is continued;

[0119] When TXY>0.65, it means that the current image optimization process is unqualified and needs to be optimized again.

[0120] In this embodiment, image optimization calculation unit 31 constructs and calculates the image optimization coefficient TXY by performing a deep fusion analysis on the optimized image grayscale value matrix YA, the optimized blue channel average value YB, the optimized grayscale co-occurrence matrix contrast YC, and the optimized region of interest area YD. The proposed TXY calculation formula fully considers image contrast, color channel characteristics, and the nonlinear response of salient areas. Compared with traditional single-metric evaluation methods, it can more comprehensively and accurately assess the visual quality and effective information intensity of an image.

[0121] Image optimization processing unit 32 incorporates an automatic judgment rule based on the image optimization coefficient TXY, setting a threshold TXY ≤ 0.65 as the optimization acceptance criterion. This ensures that only images that meet the optimization criteria are sent to subsequent processing. This mechanism effectively avoids feature extraction errors caused by insufficient optimization, improving system stability and image processing quality consistency in complex environments.

[0122] For images with a TXY greater than 0.65, the system automatically determines that the optimization process has failed and re-enters the optimization process, implementing a closed-loop control mechanism of "rollback if quality fails to meet standards." This mechanism ensures that all image inputs meet a uniform quality standard, fundamentally improving the robustness and usability of image feature extraction in subsequent modules of the system, and preventing path planning errors or clamping failures caused by poor-quality images.

[0123] This module optimizes key image metrics through reprocessing, resulting in sharper image contours and more defined boundaries. This provides a higher-quality visual foundation for subsequent terminal posture recognition, path planning, and motion control. This module demonstrates significant advantages in environments with fluctuating lighting, complex backgrounds, or partial terminal occlusion, enhancing the system's overall adaptability and environmental robustness.

[0124] Example 5: Please refer to Figure 1 , the posture analysis module 4 includes a posture analysis calculation unit 41 and a posture analysis processing unit 42;

[0125] The pose analysis calculation unit 41 is configured to perform data extraction on the second data set, including the position E of the terminal center point in the camera coordinate system, the angle F between the terminal main shaft and the horizontal axis, the depth G from the camera, and the terminal pitch angle H, and input the extracted data into the pre-trained deep learning framework, perform feature fusion through a multi-layer neural network, and then calculate and obtain the pose recognition coefficient WZB. The specific calculation formula is as follows:

[0126]

[0127] In the formula, E is the position of the terminal center point in the camera coordinate system, F is the angle between the terminal main shaft and the horizontal axis, G is the depth from the camera, and H is the terminal pitch angle.

[0128] The pose analysis processing unit 42 is configured to perform data analysis on the pose recognition coefficient WZB, and determine whether the current pose needs to be optimized according to the analysis result. The specific analysis steps are as follows:

[0129] When WZB≤0.55, it represents that the current pose does not need to be optimized.

[0130] When WZB>0.55, it represents that the current pose needs to be optimized, and an optimized pose data set is generated, including the position YE of the optimized terminal center point in the camera coordinate system, the angle YF between the optimized terminal main shaft and the horizontal axis, the depth YG from the camera, and the optimized terminal pitch angle YH.

[0131] In the embodiment, the pose analysis calculation unit 41 extracts a plurality of key pose parameters in the second data set, and realizes multi-layer neural network fusion through a pre-trained deep learning model, and then calculates the pose recognition coefficient WZB. This fusion method not only improves the accuracy of pose estimation, but also enhances the adaptability of the system to complex spatial placement situations, which is significantly better than the traditional method based on single angle or coordinate judgment.

[0132] Through mathematical formula, a pose evaluation standard based on multi-variable fusion is constructed, and the system can scientifically judge the spatial placement state of the current target object. The pose analysis processing unit 42 further sets a threshold value judgment logic according to the WZB value: when WZB≤0.55, it represents that the pose is qualified and does not need to be optimized; when WZB>0.55, it represents that there is obvious error and needs to be optimized. This standardized judgment method improves the objectivity of system decision and the automation degree of processing.

[0133] For the target determined to be optimized, the system will automatically generate a structured optimized pose data set, providing accurate input information for the subsequent pose optimization module 5. This mechanism can effectively reduce the path planning error or clamping failure risk caused by the deviation of the pose information, and enhance the robustness and action reliability of the overall system.

[0134] This module combines four-dimensional modeling and analysis of terminal posture information to quickly identify unfavorable posture conditions such as tilt, offset, and occlusion, effectively guiding subsequent adjustments to the path and clamping strategy. It is particularly suitable for dynamic environments or diverse placement scenarios, significantly improving the system's responsiveness and success rate in real-world industrial scenarios.

[0135] Example 6: Please refer to Figure 1 , the posture optimization module 5 includes a posture optimization calculation unit 51 and a posture optimization processing unit 52;

[0136] The posture optimization calculation unit 51 is used to extract data from the posture optimization data group, including optimizing the position YE of the terminal center point in the camera coordinate system, optimizing the angle YF between the terminal main axis and the horizontal axis, optimizing the depth YG from the camera, and optimizing the pitch angle YH of the terminal. The extracted multiple data are input into the pre-trained deep learning framework, and feature fusion is performed through a multi-layer neural network to calculate the posture optimization coefficient WZY. The specific calculation formula is as follows:

[0137]

[0138] Where: YE is the position of the center point of the optimized terminal in the camera coordinate system, YF is the angle between the main axis of the optimized terminal and the horizontal axis, YG is the depth of the optimized distance from the camera, and YH is the pitch angle of the optimized terminal;

[0139] The posture optimization processing unit 52 is used to perform data analysis on the posture optimization coefficient WZY obtained by calculation, and judge whether the current posture optimization is qualified according to the analysis results. The specific steps are as follows:

[0140] When WZY≤0.65, it means that the current posture optimization process is qualified and the next adjustment is continued;

[0141] When WZY>0.65, it means that the current pose optimization process is unqualified and needs to be re-optimized.

[0142] In this embodiment: the posture optimization calculation unit 51 performs deep learning fusion modeling on the optimization data obtained from the second stage analysis, and calculates the optimization coefficient WZY representing the quality of the spatial posture. This quantitative evaluation index takes into account the spatial position, posture angle and depth information, and constructs an evaluation model that can sensitively perceive clamping deviations, providing a scientific basis for posture optimization.

[0143] The pose optimization processing unit 52 sets a judgment threshold based on the numerical distribution of WZY: when WZY ≤ 0.65, the optimization is effective; when WZY > 0.65, the optimization fails and requires further correction. This judgment mechanism effectively suppresses pose deviations caused by initial estimation errors, sensory interference, or irregular object morphology, significantly improving the system's stability and reliability in grasping tasks. During the processing process, this module supports recalculation and iterative optimization of unqualified results. Its closed-loop feedback feature enables the system to adjust gripping parameters in real time based on pose changes, effectively preventing gripping errors or mechanical impact, and ensuring the continuity and safety of system operation. The WZY coefficient incorporates key geometric variables and performs nonlinear pose fitting through a deep learning model. This effectively adapts to diverse operating environments such as irregular placement, complex lighting, and visual occlusion, enabling the robot arm to maintain high recognition accuracy and path precision under varying working conditions. Through the module's automatic judgment and secondary correction capabilities, the system can automatically identify and adjust potential pose issues, significantly reducing the need for manual intervention and improving overall operational efficiency, providing technical support for building a truly efficient intelligent assembly system.

[0144] Example 7: Please refer to Figure 1 , the path analysis module 6 includes a path analysis calculation unit 61 and a path analysis processing unit 62;

[0145] The path analysis calculation unit 61 is used to extract data from the third data set, including the starting position I, the target position J, the interpolation trajectory function K, and the movement speed limit L, and input the extracted multiple data into the pre-trained deep learning framework. The multi-layer neural network is used to perform feature fusion to calculate the route identification coefficient LJB. The specific calculation formula is as follows:

[0146]

[0147] Where: I is the starting pose, J is the target pose, K is the interpolation trajectory function, and L is the motion speed limit;

[0148] The path analysis processing unit 62 is used to perform data analysis on the calculated route identification coefficient LJB and determine whether the current path needs to be optimized based on the analysis results. The specific steps are as follows:

[0149] When LJB≤0.55, it means that the current path does not need to be optimized;

[0150] When LJB>0.55, it means that the current path needs to be optimized and an optimized path data set is generated, including the optimized starting pose YI, the optimized target pose YJ, the optimized interpolation trajectory function YK, and the optimized motion speed limit YL.

[0151] In this embodiment: The path analysis calculation unit 61 extracts the starting posture I, target posture J, interpolation trajectory function K and motion speed limit L in the third data group, and integrates them into the pre-trained multi-layer neural network. Through the nonlinear feature cross-extraction method, the path identification coefficient LJB is calculated. This model establishes a comprehensive evaluation index for path rationality by integrating the degree of spatial posture change, trajectory function complexity and speed constraint parameters, breaking through the problem of traditional path analysis relying on a single factor and improving the accuracy of judgment before path planning.

[0152] The path analysis processing unit 62 sets a reasonable threshold based on the LJB value. When LJB ≤ 0.55, the path planning is considered qualified and can be directly executed. When LJB > 0.55, the path is considered abnormal or uncontrollable and requires optimization. This mechanism effectively identifies problems in the path that may cause interference, speeding, or spatial conflicts, providing the system with early intervention capabilities and enhancing safety and robustness. The neural network feature fusion mechanism can generate high-dimensional modeling of nonlinear trajectory morphology and multi-segment speed change curves. It is particularly suitable for complex operating environments that require obstacle avoidance and dynamic path adjustment, significantly improving the adaptability of traditional interpolation algorithms under changing working conditions.

[0153] Example 8: Please refer to Figure 1 , the path optimization module 7 includes a path optimization calculation unit 71 and a path optimization processing unit 72;

[0154] The path optimization calculation unit 71 is used to extract data from the path optimization data set, including optimizing the starting posture YI, optimizing the target posture YJ, optimizing the interpolation trajectory function YK, and optimizing the movement speed limit YL. The extracted multiple data are input into the pre-trained deep learning framework, and the path optimization coefficient LJY is obtained by feature fusion calculation through a multi-layer neural network. The specific calculation method is as follows:

[0155]

[0156] Where: YI is the optimized starting pose, YJ is the optimized target pose, YK is the optimized interpolation trajectory function, and YL is the optimized motion speed limit;

[0157] The path optimization processing unit 72 is used to perform data analysis on the calculated path optimization coefficient LJY and generate analysis results. Based on the analysis results, it is determined whether the current path optimization is qualified. The specific steps are as follows:

[0158] When LJY≤0.65, it means that the current path optimization process is qualified and the next adjustment is continued;

[0159] When LJY>0.65, it means that the current path optimization process is unqualified and needs to be re-optimized.

[0160] In this embodiment: Compared with the traditional path planning method that relies on rules or heuristic algorithms, this system inputs the optimized starting posture YI, target posture YJ, interpolation trajectory function YK and speed limit YL into a pre-trained multi-layer neural network, and utilizes the feature fusion and nonlinear mapping capabilities of the deep learning model to effectively improve the accuracy and adaptability of path optimization, and can dynamically respond to complex posture changes and path constraints in the process of wiring harness terminal clamping.

[0161] The proposed path optimization coefficient, LJY, integrates spatial pose distance, trajectory function complexity, and speed limit effects to form a unified and quantifiable path evaluation system. By setting a clear judgment threshold, it addresses the existing problem of inconsistent evaluation standards and reliance on human experience in path optimization, thereby improving system processing efficiency and standardization.

[0162] The path optimization processing unit analyzes the path optimization coefficient LJY value and uses this to determine whether the current path meets the clamping requirements. This system has path evaluation and feedback capabilities. This mechanism significantly enhances the intelligent closed-loop capability of path adjustment, avoids clamping failures caused by path quality differences, and improves the stability and success rate of the entire clamping system under complex working conditions.

[0163] Because the path optimization mechanism has the ability to "judge unqualified → automatically re-plan", it can quickly respond and adjust the path under variable working conditions such as wiring harness posture deviation and terminal position slight offset, thereby improving the system's robustness to interference from the actual production site and meeting the needs of automated assembly production lines with high cycle time and high reliability requirements.

[0164] This application also includes a method for intelligently clamping wiring harness terminals based on machine vision, see Figure 2 , the specific steps are as follows:

[0165] S1. Detect and collect data in the region through data collection module 1, and pre-process the collected data to generate a first data group, a second data group, and a third data group;

[0166] S2. The image analysis module 2 performs data coupling on the first data set to generate an image recognition coefficient TXB, performs data analysis on the first data set, and generates an image optimization data set. The image optimization module 3 performs data coupling on the image optimization data set to generate an image optimization coefficient TXY, and performs analysis on the image optimization coefficient TXY.

[0167] S3. Perform data coupling on the second data set through the posture analysis module 4 to generate a posture identification coefficient WZB, analyze the coefficient, and generate a posture optimization data set. Perform data coupling on the posture optimization data set through the posture optimization module 5 to generate a posture optimization coefficient WZY, and analyze the coefficient.

[0168] S4. The path analysis module 6 performs data coupling on the third data set to generate a route identification coefficient LJB, analyzes the coefficient, and generates a posture optimization data set. The path optimization module 7 performs data coupling on the posture optimization data set to generate a path optimization coefficient LJY, and analyzes the coefficient.

[0169] S5. Feedback each result to the visualization terminal through the feedback module 8.

[0170] The contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0171] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A machine vision-based intelligent clamping system for wiring harness terminals, characterized by: It includes a data acquisition module (1), an image analysis module (2), an image optimization module (3), a posture analysis module (4), a posture optimization module (5), a path analysis module (6), a path optimization module (7) and a feedback module (8); The data acquisition module (1) is used to detect and acquire data in the region, and pre-process the acquired data to generate a first data group, a second data group, and a third data group; The image analysis module (2) performs data coupling on the first data set to generate an image recognition coefficient TXB, performs data analysis on the first data set, and generates an image optimization data set; The image optimization module (3) is used to perform data coupling on the image optimization data group, thereby generating an image optimization coefficient TXY and analyzing the same; The posture analysis module (4) is used to perform data coupling on the second data group, thereby generating a posture identification coefficient WZB, analyzing the coefficient, and generating a posture optimization data group; The posture optimization module (5) is used to perform data coupling on the posture optimization data group, thereby generating a posture optimization coefficient WZY and analyzing it; The path analysis module (6) is used to perform data coupling on the third data group, thereby generating a route identification coefficient LJB, analyzing the third data group, and generating a posture optimization data group; The path optimization module (7) is used to perform data coupling on the posture optimization data group, thereby generating a path optimization coefficient LJY and analyzing the same; The feedback module (8) is used to feed back various results to the visualization terminal.

2. The machine vision-based intelligent clamping system for wiring harness terminals according to claim 1, characterized in that: The data acquisition module (1) includes an information acquisition unit (11) and an information preprocessing unit (12); The information acquisition unit (11) is used to detect and collect data on the device area, including the image gray value matrix A, the blue channel average value B, the gray co-occurrence matrix contrast C and the area of ​​the region of interest D, the position E of the terminal center point in the camera coordinate system, the angle F between the terminal main axis and the horizontal axis, the depth G from the camera and the terminal pitch angle H, the starting posture I, the target posture J, the interpolation trajectory function K and the movement speed limit L; The information preprocessing unit (12) is used to preprocess and dimensionlessly process the collected data, and to sort the processed data, thereby generating a first data group, a second data group, and a third data group; The first data set includes the image gray value matrix A, the blue channel average value B, the gray level co-occurrence matrix contrast C and the area of ​​the region of interest D; The second data set includes the position E of the terminal center point in the camera coordinate system, the angle F between the terminal main axis and the horizontal axis, the depth G from the camera, and the pitch angle H of the terminal; The third data set includes the starting pose I, the target pose J, the interpolation trajectory function K, and the motion speed limit L.

3. The machine vision-based intelligent clamping system for wiring harness terminals according to claim 2, characterized in that: The image analysis module (2) includes an image analysis calculation unit (21) and an image analysis processing unit (22); The image analysis calculation unit (21) is used to extract data from the first data group, including the image gray value matrix A, the blue channel average value B, the gray co-occurrence matrix contrast C and the area of ​​the region of interest D, and input the extracted multiple data into the pre-trained deep learning framework, perform feature fusion through the multi-layer neural network, and then calculate and obtain the image recognition coefficient TXB. The specific calculation formula is as follows: Where: image gray value matrix A, blue channel average value B, gray level co-occurrence matrix contrast C and area of ​​the region of interest D; The image analysis processing unit (22) is used to perform data analysis on the image recognition coefficient TXB and determine whether the current image needs to be optimized based on the analysis result. The specific analysis method is as follows: When TXB≤0.56, it means that the current image does not need to be optimized; When TXB>0.56, it means that the current image needs to be optimized and an optimized image data set is generated, including the optimized image gray value matrix YA, the optimized blue channel average value YB, the optimized gray co-occurrence matrix contrast YC, and the optimized region of interest area YD.

4. The machine vision-based intelligent clamping system for wiring harness terminals according to claim 3, characterized in that: The image optimization module (3) includes an image optimization calculation unit (31) and an image optimization processing unit (32); The image optimization calculation unit (31) is used to extract data from the optimized image data group, including optimizing the image gray value matrix YA, optimizing the blue channel average value YB, optimizing the gray co-occurrence matrix contrast YC, and optimizing the area of ​​the region of interest YD, and inputting the extracted multiple data into a pre-trained deep learning framework, performing feature fusion through a multi-layer neural network, and then calculating and obtaining the image optimization coefficient TXY. The specific calculation formula is as follows: Where: YA is the optimized image gray value matrix, YB is the optimized blue channel average, YC is the optimized gray level co-occurrence matrix contrast, and YD is the optimized region of interest area; The image optimization processing unit (32) is used to perform data analysis on the image optimization coefficient TXY obtained by calculation, and judge whether the image optimization effect is qualified according to the analysis result. The specific steps are as follows: When TXY≤0.65, it means that the current image optimization process is qualified and the next adjustment is continued; When TXY>0.65, it means that the current image optimization process is unqualified and needs to be optimized again.

5. The intelligent clamping system for wire harness terminals based on machine vision according to claim 4, characterized in that: The posture analysis module (4) includes a posture analysis calculation unit (41) and a posture analysis processing unit (42); The posture analysis and calculation unit (41) is used to extract data from the second data group, including the position E of the terminal center point in the camera coordinate system, the angle F between the terminal main axis and the horizontal axis, the depth G from the camera, and the terminal pitch angle H, and input the extracted multiple data into a pre-trained deep learning framework, perform feature fusion through a multi-layer neural network, and then calculate and obtain the posture recognition coefficient WZB. The specific calculation formula is as follows: Where: E is the position of the terminal center point in the camera coordinate system, F is the angle between the terminal main axis and the horizontal axis, G is the depth from the camera, and H is the pitch angle of the terminal; The posture analysis processing unit (42) is used to perform data analysis on the posture identification coefficient WZB, and judge whether the current posture needs to be optimized based on the analysis results. The specific analysis steps are as follows: When WZB≤0.55, it means that the current posture does not need to be optimized; When WZB>0.55, it means that the current posture needs to be optimized and an optimized posture data group is generated, including optimizing the position YE of the terminal center point in the camera coordinate system, optimizing the angle YF between the terminal main axis and the horizontal axis, optimizing the depth YG from the camera, and optimizing the terminal pitch angle YH.

6. The machine vision-based intelligent clamping system for wiring harness terminals according to claim 5, characterized in that: The posture optimization module (5) includes a posture optimization calculation unit (51) and a posture optimization processing unit (52); The posture optimization calculation unit (51) is used to extract data from the posture optimization data group, including optimizing the position YE of the terminal center point in the camera coordinate system, optimizing the angle YF between the terminal main axis and the horizontal axis, optimizing the depth YG from the camera, and optimizing the terminal pitch angle YH, and inputting the extracted multiple data into the pre-trained deep learning framework, performing feature fusion through a multi-layer neural network, and thus calculating and obtaining the posture optimization coefficient WZY. The specific calculation formula is as follows: Where: YE is the position of the center point of the optimized terminal in the camera coordinate system, YF is the angle between the main axis of the optimized terminal and the horizontal axis, YG is the depth of the optimized distance from the camera, and YH is the pitch angle of the optimized terminal; The posture optimization processing unit (52) is used to perform data analysis on the posture optimization coefficient WZY obtained by calculation, and judge whether the current posture optimization is qualified according to the analysis result. The specific steps are as follows: When WZY≤0.65, it means that the current posture optimization process is qualified and the next adjustment is continued; When WZY>0.65, it means that the current pose optimization process is unqualified and needs to be re-optimized.

7. The machine vision-based intelligent clamping system for wiring harness terminals according to claim 6, characterized in that: The path analysis module (6) includes a path analysis calculation unit (61) and a path analysis processing unit (62); The path analysis calculation unit (61) is used to extract data from the third data group, including the starting posture I, the target posture J, the interpolation trajectory function K and the movement speed limit L, and input the extracted multiple data into the pre-trained deep learning framework, perform feature fusion through the multi-layer neural network, and thus calculate the route identification coefficient LJB. The specific calculation formula is as follows: Where: I is the starting pose, J is the target pose, K is the interpolation trajectory function, and L is the motion speed limit; The path analysis processing unit (62) is used to perform data analysis on the calculated route identification coefficient LJB, and determine whether the current path needs to be optimized based on the analysis result. The specific steps are as follows: When LJB≤0.55, it means that the current path does not need to be optimized; When LJB>0.55, it means that the current path needs to be optimized and an optimized path data set is generated, including the optimized starting pose YI, the optimized target pose YJ, the optimized interpolation trajectory function YK, and the optimized motion speed limit YL.

8. The machine vision-based intelligent clamping system for wiring harness terminals according to claim 7, characterized in that: The path optimization module (7) includes a path optimization calculation unit (71) and a path optimization processing unit (72); The path optimization calculation unit (71) is used to extract data from the path optimization data group, including optimizing the starting posture YI, optimizing the target posture YJ, optimizing the interpolation trajectory function YK and optimizing the movement speed limit YL, and inputting the extracted multiple data into the pre-trained deep learning framework, and performing feature fusion calculation through a multi-layer neural network to obtain the path optimization coefficient LJY. The specific calculation method is as follows: Where: YI is the optimized starting pose, YJ is the optimized target pose, YK is the optimized interpolation trajectory function, and YL is the optimized motion speed limit; The path optimization processing unit (72) is used to perform data analysis on the calculated path optimization coefficient LJY and generate analysis results. According to the analysis results, it is judged whether the current path optimization is qualified. The specific steps are as follows: When LJY≤0.65, it means that the current path optimization process is qualified and the next adjustment is continued; When LJY>0.65, it means that the current path optimization process is unqualified and needs to be re-optimized.

9. A machine vision-based intelligent clamping method for wiring harness terminals, characterized by: The machine vision-based intelligent clamping method for wire harness terminals is performed by the machine vision-based intelligent clamping system for wire harness terminals described in any one of claims 1 to 8, and the specific steps are as follows: S1, performing data detection and collection on the region through the data collection module (1), and pre-processing the collected data to generate a first data group, a second data group, and a third data group; S2, performing data coupling on the first data set by the image analysis module (2) to generate an image recognition coefficient TXB, performing data analysis on the first data set, and generating an image optimization data set; performing data coupling on the image optimization data set by the image optimization module (3) to generate an image optimization coefficient TXY, and performing analysis on the image optimization coefficient TXY; S3, performing data coupling on the second data set through the posture analysis module (4) to generate a posture identification coefficient WZB, analyzing the coefficient and generating a posture optimization data set, performing data coupling on the posture optimization data set through the posture optimization module (5) to generate a posture optimization coefficient WZY, and analyzing the coefficient; S4, performing data coupling on the third data set through the path analysis module (6) to generate a route identification coefficient LJB, analyzing the route identification coefficient LJB, and generating a posture optimization data set; performing data coupling on the posture optimization data set through the path optimization module (7) to generate a path optimization coefficient LJY, and analyzing the route identification coefficient LJB; S5. Feedback each result to the visualization terminal through the feedback module (8).

Images