Automobile wire harness precision fastener surface flaw detection system based on visual recognition

Abstract

This invention belongs to the field of automotive wiring harness precision fastener inspection technology, specifically a vision-based surface defect detection system for automotive wiring harness precision fasteners. It includes a multispectral synchronous imaging acquisition module, a dynamic threshold adaptive enhancement module, a microscopic morphology feature deconstruction module, a multi-dimensional defect mapping and classification module, an intelligent decision feedback execution module, and an operation monitoring terminal. This invention acquires raw data on the material optical properties adapted to automotive wiring harness precision fasteners, uses adaptive image enhancement based on local feature statistics to eliminate uneven lighting and noise interference, quantifies the geometric and textural attributes of defects through multi-scale feature extraction and morphology reconstruction, outputs surface defect classification results and confidence scores for automotive wiring harness precision fasteners, and drives corresponding actuators based on the surface defect detection results. This provides a high-precision, high-efficiency, and intelligent quality control solution for automotive fastener production, significantly reducing labor costs and defect rates.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of automobile wire harness precision fastener detection, in particular to a surface flaw detection system for automobile wire harness precision fasteners based on visual recognition. BACKGROUND

[0002] The automobile wire harness precision fastener is a key component for fixing, protecting and connecting the wire harness in the automobile electrical system, and with the development of new energy vehicles and automatic driving technology, the precision, material and function of the automobile wire harness precision fastener will continue to improve, promoting the evolution of the industry towards intelligence and light weight.

[0003] In the production process of the automobile wire harness precision fastener, the surface quality flaws need to be detected, and the surface quality of the automobile wire harness fastener directly affects the reliability and safety of the wire harness connection. The traditional manual detection method is not only low in efficiency, but also difficult to ensure the accuracy of the surface flaw detection result.

[0004] And the existing visual detection system mostly relies on a single light source or a fixed algorithm, and is difficult to cope with complex background interference, small flaws and material reflection characteristics differences, and is difficult to accurately analyze and judge the precision measurement result of the automobile wire harness precision fastener in the detection process and the abnormality, which is not conducive to ensuring the smoothness and high efficiency and reliability of the system detection, and the intelligent degree is low and the supervision difficulty is large.

[0005] In view of the above technical defects, a solution is proposed. SUMMARY

[0006] The purpose of the present application is to provide a surface flaw detection system for automobile wire harness precision fasteners based on visual recognition, which solves the problem that the prior art relies on a single light source or a fixed algorithm, and is difficult to cope with complex background interference, small flaws and material reflection characteristics differences, and is difficult to accurately analyze and judge the precision measurement result of the automobile wire harness precision fastener in the detection process and the abnormality, which is not conducive to ensuring the smoothness and high efficiency and reliability of the system detection, and the intelligent degree is low and the supervision difficulty is large.

[0007] To achieve the above purpose, the present application provides the following technical scheme:

[0008] The surface flaw detection system for automobile wire harness precision fasteners based on visual recognition comprises a multi-spectral synchronous imaging acquisition module, a dynamic threshold adaptive enhancement module, a micro-morphology feature deconstruction module, a multi-dimensional flaw mapping classification module, an intelligent decision feedback execution module and an operation supervision terminal; the multi-spectral synchronous imaging acquisition module acquires the original data suitable for the material optical characteristics of the automobile wire harness precision fastener through multi-band light source cooperative control and high-precision image synchronous acquisition;

[0009] The dynamic threshold adaptive enhancement module receives the multispectral image, performs adaptive image enhancement based on local feature statistics, eliminates uneven illumination and noise interference, and outputs the enhanced image to the micro-morphology feature deconstruction module; the micro-morphology feature deconstruction module quantifies the geometric and texture properties of the defects through multi-scale feature extraction and morphology reconstruction, and sends the deconstructed feature atlas to the multi-dimensional defect mapping classification module;

[0010] The multi-dimensional defect mapping classification module receives the deconstructed feature atlas, outputs the surface defect classification results and confidence scores of the automobile wire harness precision fastener to the intelligent decision feedback execution module and the operation supervision terminal; the intelligent decision feedback execution module drives the corresponding execution mechanism according to the surface defect detection results, and sends the execution information to the operation supervision terminal.

[0011] Further, the multispectral synchronous imaging acquisition module uses a ring-shaped LED array light source, integrates visible light, near-infrared and ultraviolet three-band independent control units, and based on a synchronous trigger controller of FPGA, different spectral images are completed in spatial alignment within a microsecond time sequence, and through a polarizer set, the surface specular reflection of metal parts is suppressed, and the diffuse reflection characteristic information is retained.

[0012] Further, the operation process of the dynamic threshold adaptive enhancement module includes:

[0013] The image is segmented into superpixel units, and the gray histogram distribution of each unit is calculated; the improved Otsu algorithm is used to dynamically generate a local threshold, and the low-contrast area is nonlinearly stretched; wavelet denoising and guided filtering are combined to retain the edge details of the defects while suppressing high-frequency noise.

[0014] Further, the operation process of the micro-morphology feature deconstruction module includes:

[0015] Directional texture feature extraction is performed through a Gabor filter bank; a morphological gradient reconstruction algorithm is applied to generate a surface morphology three-dimensional topological map; and defect spatial positioning across images is achieved through SIFT feature point matching.

[0016] Further, the multi-dimensional defect mapping classification module constructs a defect feature library containing several groups of samples, covering typical defects including burrs, scratches and oxidation, and in the running process, the multi-dimensional defect mapping classification module uses an improved twin network for feature similarity matching, and introduces a confidence decay mechanism to trigger a secondary verification process for low-probability results;

[0017] The intelligent decision feedback execution module designs a hierarchical rejection strategy, directly triggers a pneumatic rejection device for high-confidence defect parts, generates a visual report for edge cases and prompts manual review.

[0018] Further, the running supervision terminal communication connection result accuracy verification module, the result accuracy verification module samples and checks the automobile wire harness precision fastener after the detection is completed, analyzes based on the checking and rechecking result to generate a result high precision signal or a result low precision signal, and sends an alarm information to the running supervision terminal when the result low precision signal is generated.

[0019] Further, the specific analysis process of the result accuracy verification module is as follows:

[0020] After obtaining the checking and rechecking results of all automobile wire harness precision fasteners for sampling inspection, if the similarity of the surface defect rechecking result of the corresponding automobile wire harness precision fastener and the previous detection result is lower than the preset similarity threshold, the corresponding automobile wire harness precision fastener is marked as a different object; the number of different objects is obtained and a ratio calculation is performed between the number of different objects and the total number of automobile wire harness precision fasteners for sampling inspection to obtain a different statistical value, and the different statistical value is compared with the preset different statistical threshold value, if the different statistical value exceeds the preset different statistical threshold value, a result low precision signal is generated; if the different statistical value does not exceed the preset different statistical threshold value, a result high precision signal is generated.

[0021] Further, the result accuracy verification module is connected to the in-out monitoring judgment module, the in-out monitoring judgment module is connected to the input influence evaluation module and the output execution evaluation module, the result accuracy verification module sends the result high precision signal to the in-out monitoring judgment module, when the in-out monitoring judgment module receives the result high precision signal, the input influence evaluation module is used to monitor and evaluate the conveying condition of the conveying belt for the automobile wire harness precision fastener, and an input qualified signal or an input alarm signal is generated accordingly, and the input qualified signal or the input alarm signal is sent to the in-out monitoring judgment module;

[0022] and the output execution evaluation module is used to monitor and evaluate the execution condition of the intelligent decision feedback execution module for the automobile wire harness precision fastener, and an output qualified signal or an output alarm signal is generated accordingly, and the output qualified signal or the output alarm signal is sent to the in-out monitoring judgment module; when the in-out monitoring judgment module receives the input alarm signal or the output alarm signal, an alarm information is generated and sent to the running supervision terminal.

[0023] Further, the specific operation process of the input influence evaluation module is as follows:

[0024] During the conveying movement of the conveying belt, the real-time conveying speed of the conveying belt is obtained, the variance of all real-time conveying speeds within a unit time is calculated to obtain a conveying speed fluctuation value, the conveying speed fluctuation value is compared with a preset conveying speed fluctuation threshold value, if the conveying speed fluctuation value exceeds the preset conveying speed fluctuation threshold value, an input alarm signal is generated;

[0025] If the transmission speed fluctuation value does not exceed the preset transmission speed fluctuation threshold, the difference between the real-time transmission speed and the set standard transmission speed is calculated and the absolute value is taken to obtain the speed difference detection value. The number of times the speed difference detection value exceeds the preset speed difference detection threshold within a unit time is marked as the speed difference abnormal value. The average value of all speed difference detection values ​​within a unit time is calculated to obtain the speed difference characteristic value. The speed difference abnormal value and the speed difference characteristic value are compared with the preset speed difference abnormal threshold and the preset speed difference characteristic threshold respectively. If the speed difference abnormal value or the speed difference characteristic value exceeds the corresponding preset threshold, an input alarm signal is generated.

[0026] If the constant value of speed difference and the characteristic value of speed difference exceed the corresponding preset threshold, when the conveyor belt is in a stopped state to cooperate with the acquisition operation of the multispectral synchronous imaging acquisition module, the vibration amplitude and vibration frequency of the conveyor belt are obtained. The vibration amplitude and vibration frequency are compared with the preset vibration amplitude threshold and the preset vibration frequency threshold respectively. If the vibration amplitude or vibration frequency exceeds the corresponding preset threshold, it is determined that the conveyor belt is in a stopped unstable state.

[0027] The duration of the conveyor belt in a stopped and unstable state within a unit time is obtained and the ratio of this duration to the total duration of the conveyor belt in a stopped state within a unit time is calculated to obtain the stopped and unstable time detection value. The percentage of the duration of the conveyor belt in a stopped state within a single stopped state is also obtained. The percentage of the duration is compared with the corresponding percentage of the duration threshold. If the percentage of the duration exceeds the preset percentage of the duration threshold, the stopped state is assigned the stopped non-optimal symbol TX-1.

[0028] The number of times the non-optimal stop symbol TX-1 is assigned per unit time is obtained and the ratio is calculated with the total number of times the conveyor belt is stopped to obtain the non-optimal stop detection value. The average vibration amplitude and average vibration frequency of the conveyor belt when it is stopped per unit time are marked as amplitude characteristic value and frequency characteristic value, respectively. The stop stability assessment value is obtained by weighted summation of the non-optimal stop detection value, the non-optimal stop detection value, the amplitude characteristic value and the frequency characteristic value. The stop stability assessment value is compared with the preset stop stability assessment threshold. If the stop stability assessment value exceeds the preset stop stability assessment threshold, an input alarm signal is generated; if the stop stability assessment value does not exceed the preset stop stability assessment threshold, an input qualified signal is generated.

[0029] Furthermore, the specific analysis process of the output execution evaluation module is as follows:

[0030] All execution devices involved in the intelligent decision feedback execution module are obtained. When the intelligent decision feedback execution module issues a drive command, the operation delay duration of the corresponding execution device is collected. The operation delay duration is compared with the corresponding preset operation delay duration threshold. If the operation delay duration exceeds the corresponding preset operation delay duration threshold, the corresponding operation delay duration is marked as an operation delay abnormal value.

[0031] The system obtains the number of abnormal operation delay values ​​corresponding to the corresponding execution device within a unit time and calculates the ratio of this number to the number of executions of the corresponding execution device within the unit time to obtain the abnormal operation delay value. It also marks the average operation delay duration of all operations corresponding to the corresponding execution device within a unit time as the abnormal operation delay characteristic value. The abnormal operation delay value and the abnormal operation delay characteristic value are compared with the corresponding preset abnormal operation delay threshold and preset abnormal operation delay characteristic threshold, respectively. If the abnormal operation delay value or the abnormal operation delay characteristic value exceeds the corresponding preset threshold, the corresponding execution device is marked as a easing device. If a easing device exists within a unit time, an output alarm signal is generated; if no easing device exists within a unit time, an output pass signal is generated.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. In this invention, by acquiring the original data of the material optical properties of precision fasteners for automotive wiring harnesses, adaptive image enhancement based on local feature statistics is used to eliminate uneven lighting and noise interference. Through multi-scale feature extraction and morphology reconstruction, the geometric and texture attributes of defects are quantified, and the surface defect classification results and confidence scores of precision fasteners for automotive wiring harnesses are output. Based on the surface defect detection results, the corresponding actuator is driven, providing a high-precision, high-efficiency, and intelligent quality control solution for automotive fastener production, significantly reducing labor costs and defect rate.

[0034] 2. In this invention, the accuracy verification module samples and checks the precision fasteners of the automotive wiring harness after inspection and judges the accuracy of the results. When a low-accuracy signal is generated, the cause is investigated and analyzed and corresponding optimization and improvement measures are taken to ensure the accuracy of subsequent defect detection results for automotive wiring harness precision fasteners. When a high-accuracy signal is generated, the input monitoring and evaluation analysis results and the output monitoring and evaluation analysis results are used to comprehensively judge the abnormality. When a corresponding alarm information is generated, the cause is investigated and analyzed and reasonable improvement measures are taken to ensure the stable and efficient operation of the system detection process and improve the system's performance in detecting surface defects of automotive wiring harness precision fasteners. Attached Figure Description

[0035] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0036] Figure 1 This is a system block diagram of Embodiment 1 of the present invention;

[0037] Figure 2 This is a system block diagram of Embodiment 2 of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Example 1: As Figure 1 As shown, the visual recognition-based automotive wiring harness precision fastener surface defect detection system proposed in this invention includes a multispectral synchronous imaging acquisition module, a dynamic threshold adaptive enhancement module, a microscopic morphology feature deconstruction module, a multidimensional defect mapping and classification module, an intelligent decision feedback execution module, a result accuracy verification module, and an operation monitoring terminal.

[0040] The multispectral synchronous imaging acquisition module acquires raw data on the material optical properties of precision fasteners for automotive wiring harnesses through multi-band light source coordinated control and high-precision image synchronous acquisition. This breaks through the limitations of single-spectral imaging and improves the detection rate of micro-cracks (≥0.02mm) and coating defects.

[0041] It should be noted that the multispectral synchronous imaging acquisition module adopts a ring LED array light source, integrates three independent control units for visible light (400-700nm), near-infrared (850nm) and ultraviolet (365nm) bands, and is based on an FPGA-based synchronous trigger controller, which enables different spectral images to complete spatial alignment within microseconds, and suppresses specular reflection on the surface of metal parts through a polarizer group, thus preserving diffuse reflection characteristic information.

[0042] The dynamic threshold adaptive enhancement module receives multispectral images, performs adaptive image enhancement based on local feature statistics to eliminate uneven illumination and noise interference, and outputs the enhanced image to the microscopic morphology feature deconstruction module, providing high signal-to-noise ratio data for subsequent analysis; the operation process includes:

[0043] The image is segmented into superpixel units, and the gray-level histogram distribution of each unit is calculated. An improved Otsu algorithm is used to dynamically generate local thresholds and nonlinearly stretch low-contrast regions. Wavelet denoising and guided filtering are combined to preserve the details of flawed edges while suppressing high-frequency noise.

[0044] The microscopic morphology feature deconstruction module quantifies the geometric and textural attributes of defects through multi-scale feature extraction and morphology reconstruction. It then sends the deconstructed feature map to the multi-dimensional defect mapping and classification module, upgrading traditional two-dimensional detection to three-dimensional morphology analysis and improving the ability to identify three-dimensional defects such as depressions and protrusions. The process can be summarized as follows: directional texture feature extraction is performed using a Gabor filter bank; a morphological gradient reconstruction algorithm is applied to generate a three-dimensional topological map of the surface morphology; and cross-image defect spatial localization is achieved through SIFT feature point matching.

[0045] The multi-dimensional defect mapping and classification module receives the deconstructed feature map and outputs the surface defect classification results and confidence scores of automotive wiring harness precision fasteners to the intelligent decision feedback execution module and the operation monitoring terminal. The multi-dimensional defect mapping and classification module constructs a defect feature library containing several sets of samples (2000+ samples), covering typical defects such as burrs, scratches and oxidation. During operation, the multi-dimensional defect mapping and classification module uses an improved Siamese Network for feature similarity matching and introduces a confidence decay mechanism to trigger a secondary verification process for low-probability results.

[0046] The intelligent decision feedback execution module drives the corresponding actuators based on the surface defect detection results and sends the execution information to the operation monitoring terminal. The intelligent decision feedback execution module is designed with a hierarchical rejection strategy, which directly triggers the pneumatic rejection device for high-confidence defective parts, generates a visual report for edge cases and prompts manual re-inspection. Preferably, the intelligent decision feedback execution module also outputs process optimization suggestions to the front-end module (such as adjusting the light source angle or enhancing the threshold) as a terminal decision node, while receiving user interaction instructions and providing feedback on the execution status.

[0047] This invention integrates multimodal sensors (such as visible light, infrared, and 3D structured light) with deep learning algorithms to achieve high-precision identification and full-type coverage (geometric, surface, and material defects) of minute defects (0.02mm level) on fastener surfaces. Even under complex conditions (strong light, dark environments, and highly reflective surfaces), it maintains a 99.9% detection rate and a false detection rate of less than 0.1%, while increasing detection efficiency to 0.5 seconds per piece. It also supports end-to-end quality traceability and process optimization, significantly reducing labor costs (by more than 80%) and defect rates (by more than 30%). This provides a high-precision, high-efficiency, and intelligent quality control solution for automotive fastener production, driving the industry towards intelligent manufacturing.

[0048] The result accuracy verification module samples and inspects the completed automotive wiring harness precision fasteners. Based on the inspection and verification results, it analyzes the data to generate either a high-accuracy signal or a low-accuracy signal. When a low-accuracy signal is generated, an alarm message is sent to the operation monitoring terminal to remind supervisory personnel to investigate the cause and take corresponding optimization and improvement measures. This ensures the accuracy of subsequent defect detection results for automotive wiring harness precision fasteners, helps ensure the quality of the output automotive wiring harness precision fasteners, avoids the rejection and waste of high-quality products, and improves the system's performance in detecting surface defects in automotive wiring harness precision fasteners. The specific analysis process of the result accuracy verification module is as follows:

[0049] The inspection and verification results of all sampled automotive wiring harness precision fasteners are obtained. If the similarity between the surface defect verification result of the corresponding automotive wiring harness precision fastener and the previous inspection result is lower than the preset similarity threshold, it indicates that the system's inspection result for the corresponding automotive wiring harness precision fastener is inaccurate. In this case, the corresponding automotive wiring harness precision fastener is marked as an object of inspection. The number of objects of inspection is obtained and the ratio is calculated with the total number of sampled automotive wiring harness precision fasteners to obtain the statistical value of inspection.

[0050] The abnormality statistics are compared with the preset abnormality statistics threshold. If the abnormality statistics exceed the preset abnormality statistics threshold, it indicates that the system's detection accuracy for the surface defects of the corresponding automotive wiring harness precision fasteners is poor, and a low-precision signal is generated. If the abnormality statistics do not exceed the preset abnormality statistics threshold, it indicates that the system's detection accuracy for the surface defects of the corresponding automotive wiring harness precision fasteners is good, and a high-precision signal is generated.

[0051] Example 2: Figure 2 As shown, the difference between this embodiment and Embodiment 1 is that the result accuracy verification module is communicatively connected to the entry and exit monitoring and judgment module, and the entry and exit monitoring and judgment module is communicatively connected to the input impact assessment module and the output execution assessment module. The result accuracy verification module sends a high-precision result signal to the entry and exit monitoring and judgment module. When the entry and exit monitoring and judgment module receives the high-precision result signal, it monitors and assesses the conveyor belt's conveying status for automotive wiring harness precision fasteners through the input impact assessment module, thereby generating an input qualified signal or an input alarm signal, and sending the input qualified signal or input alarm signal to the entry and exit monitoring and judgment module.

[0052] Furthermore, the intelligent decision feedback execution module monitors and evaluates the execution status of automotive wiring harness precision fasteners through the output execution evaluation module, generating either an output pass signal or an output alarm signal, and sending the output pass signal or output alarm signal to the entry / exit monitoring and judgment module. When the entry / exit monitoring and judgment module receives an input alarm signal or an output alarm signal, it generates alarm information and sends it to the operation monitoring terminal. When the operation monitoring terminal receives the alarm information, it displays it and issues a warning to remind supervisory personnel to investigate and analyze the cause and take reasonable improvement measures, thereby ensuring the stable and efficient operation of the system's detection process and further improving the system's performance in detecting surface defects in automotive wiring harness precision fasteners.

[0053] The specific operation process of the input impact assessment module is as follows: During the conveyor belt's conveying motion, the real-time conveying speed of the conveyor belt is obtained. The variance of all real-time conveying speeds within a unit time is calculated to obtain the conveying speed fluctuation value. The conveying speed fluctuation value is compared with the preset conveying speed fluctuation threshold. If the conveying speed fluctuation value exceeds the preset conveying speed fluctuation threshold, it indicates that the speed fluctuation during the conveyor belt's movement is large, which is not conducive to the smooth arrival of the automotive wiring harness precision fasteners. In this case, an input alarm signal is generated.

[0054] If the transmission speed fluctuation value does not exceed the preset transmission speed fluctuation threshold, the difference between the real-time transmission speed and the set standard transmission speed is calculated and the absolute value is taken to obtain the speed difference detection value. The number of times the speed difference detection value exceeds the preset speed difference detection threshold within a unit time is marked as the speed difference abnormal value. The average value of all speed difference detection values ​​within a unit time is calculated to obtain the speed difference characteristic value. The speed difference abnormal value and the speed difference characteristic value are compared with the preset speed difference abnormal threshold and the preset speed difference characteristic threshold respectively.

[0055] If the speed difference abnormal value or speed difference characteristic value exceeds the corresponding preset threshold, it indicates that the speed execution accuracy of the conveyor belt movement process is not good, which is not conducive to ensuring the smooth arrival of the automotive wiring harness precision fasteners while ensuring conveying efficiency. In this case, an input alarm signal is generated.

[0056] If the speed difference constant value and speed difference characteristic value exceed the corresponding preset threshold, when the conveyor belt is in a stopped state to cooperate with the acquisition operation of the multispectral synchronous imaging acquisition module, the vibration amplitude and vibration frequency of the conveyor belt are obtained. The vibration amplitude and vibration frequency are compared with the preset vibration amplitude threshold and preset vibration frequency threshold respectively. If the vibration amplitude or vibration frequency exceeds the corresponding preset threshold, it indicates that the current vibration of the conveyor belt is obvious, and it is determined that the conveyor belt is in a stopped unstable state.

[0057] The duration of the conveyor belt in a stopped and unstable state within a unit time is obtained and the ratio of this duration to the total duration of the conveyor belt in a stopped state within a unit time is calculated to obtain the stopped and unstable time detection value. The percentage of the duration of the conveyor belt in a stopped state within a single stopped state is also obtained. The percentage of the duration is compared with the corresponding percentage of the duration threshold. If the percentage of the duration exceeds the preset percentage of the duration threshold, the stopped state is assigned the stopped non-optimal symbol TX-1.

[0058] The number of times the non-optimal stop symbol TX-1 is assigned per unit time is obtained and the ratio is calculated with the total number of times the conveyor belt is in a stopped state to obtain the non-optimal stop detection value. The average value of the vibration amplitude and the average value of the vibration frequency when the conveyor belt is in a stopped state per unit time are marked as the amplitude characteristic value and the vibration frequency characteristic value, respectively.

[0059] The stopping stability assessment value is obtained by weighted summation of the stopping non-stable detection time value, stopping non-optimal detection value, amplitude characteristic value, and frequency characteristic value. Specifically, each of the stopping non-stable detection time value, stopping non-optimal detection value, amplitude characteristic value, and frequency characteristic value is assigned a corresponding preset weight coefficient, and then each of these values ​​is multiplied by its corresponding preset weight coefficient. The sum of the four products is then marked as the stopping stability assessment value. It should be noted that the larger the stopping stability assessment value, the more unstable the conveying stopping process is per unit time, which is less conducive to ensuring the visual acquisition effect of automotive wiring harness precision fasteners.

[0060] The stop stability assessment value is compared with the preset stop stability assessment threshold. If the stop stability assessment value exceeds the preset stop stability assessment threshold, it indicates that the conveying stop process is unstable within a unit of time, which is not conducive to ensuring the visual acquisition effect of automotive wiring harness precision fasteners, and an input alarm signal is generated. If the stop stability assessment value does not exceed the preset stop stability assessment threshold, it indicates that the overall conveying performance of automotive wiring harness precision fasteners within a unit of time is better, and an input qualified signal is generated.

[0061] Furthermore, the specific analysis process of the output execution evaluation module is as follows: all execution devices involved in the intelligent decision feedback execution module (such as pneumatic rejection devices, report display devices, and voice prompt devices) are acquired. When the intelligent decision feedback execution module issues a drive command, the operation delay duration of the corresponding execution device is collected. The operation delay duration is compared with the corresponding preset operation delay duration threshold. If the operation delay duration exceeds the corresponding preset operation delay duration threshold, it indicates that the corresponding execution device is not executing the operation in a timely manner. The corresponding operation delay duration is then marked as an abnormal operation delay value.

[0062] The number of abnormal operation delay values ​​corresponding to the corresponding execution device within a unit time is obtained and the ratio of the number of executions of the corresponding execution device within a unit time is calculated to obtain the abnormal operation delay value. The average value of all operation delay durations corresponding to the corresponding execution device within a unit time is marked as the operation delay feature value. The abnormal operation delay value and the operation delay feature value are compared with the corresponding preset abnormal operation delay threshold and the preset operation delay feature threshold respectively.

[0063] If the operation delay abnormality value or operation delay characteristic value exceeds the corresponding preset threshold, it indicates that the execution timeliness of the corresponding execution device is poor within a unit time, and the corresponding execution device is marked as a easing device; if there is a easing device within a unit time, it indicates that the output side execution performance of the automotive wiring harness precision fastener detection process is poor, and an output alarm signal is generated; if there is no easing device within a unit time, it indicates that the output side execution performance of the automotive wiring harness precision fastener detection process is good, and an output qualified signal is generated.

[0064] The working principle of this invention is as follows: During use, a multispectral synchronous imaging acquisition module acquires raw data on the material optical properties of precision fasteners for automotive wiring harnesses. A dynamic threshold adaptive enhancement module uses adaptive image enhancement based on local feature statistics to eliminate uneven lighting and noise interference. A microscopic morphology feature deconstruction module quantifies the geometric and textural attributes of defects through multi-scale feature extraction and morphology reconstruction. A multi-dimensional defect mapping and classification module outputs the surface defect classification results and confidence scores for the precision fasteners for automotive wiring harnesses. An intelligent decision feedback execution module drives the corresponding actuators based on the surface defect detection results. This provides a high-precision, high-efficiency, and intelligent quality control solution for automotive fastener production. Furthermore, a result accuracy verification module samples and checks the inspected precision fasteners for automotive wiring harnesses to determine the accuracy of the results. When a low-accuracy signal is generated, the cause is investigated and analyzed, and corresponding optimization and improvement measures are taken to ensure the accuracy of subsequent defect detection results for automotive wiring harnesses. This helps ensure the quality of the output precision fasteners and avoids the rejection and waste of high-quality products.

[0065] In this invention, the threshold, preset value, or preset range settings are for result comparison and analysis to determine whether the result is good or bad. The magnitude of these values ​​is determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions. Similarly, the preset weight coefficients and influence factors are assigned specific values ​​based on the magnitude of each parameter's influence on the result, ultimately reflecting the impact on the result. These settings are also determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions.

[0066] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, enabling those skilled in the art to better understand and utilize it. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A vision-based system for detecting surface defects in precision fasteners of automotive wiring harnesses, characterized in that, It includes a multispectral synchronous imaging acquisition module, a dynamic threshold adaptive enhancement module, a microscopic morphology feature deconstruction module, a multidimensional defect mapping and classification module, an intelligent decision feedback execution module, and an operation monitoring terminal; the multispectral synchronous imaging acquisition module acquires raw data on the material optical properties of precision fasteners for automotive wiring harnesses through multi-band light source collaborative control and high-precision image synchronous acquisition. The dynamic threshold adaptive enhancement module receives multispectral images and uses adaptive image enhancement based on local feature statistics to eliminate uneven illumination and noise interference. The micro-morphology feature deconstruction module quantifies the geometric and texture attributes of defects through multi-scale feature extraction and morphology reconstruction; the multi-dimensional defect mapping and classification module receives the deconstructed feature map and outputs the surface defect classification results and confidence scores of automotive wiring harness precision fasteners; the intelligent decision feedback execution module drives the corresponding execution mechanism according to the surface defect detection results and sends the execution information to the operation monitoring terminal. The operation and supervision terminal communication connection result accuracy verification module samples and inspects the precision fasteners of the automotive wiring harness after the inspection is completed. Based on the inspection and verification results, it analyzes the results to generate a high-precision signal or a low-precision signal. When a low-precision signal is generated, it sends an alarm message to the operation and supervision terminal. The specific analysis process of the results accuracy verification module is as follows: The inspection and verification results of all sampled automotive wiring harness precision fasteners are obtained. If the similarity between the surface defect verification result of the corresponding automotive wiring harness precision fastener and the previous inspection result is lower than the preset similarity threshold, the corresponding automotive wiring harness precision fastener is marked as an object of inspection. The number of objects of inspection is obtained and the ratio is calculated with the total number of sampled automotive wiring harness precision fasteners to obtain the inspection statistics value. The inspection statistics value is compared with the preset inspection statistics threshold. If the inspection statistics value exceeds the preset inspection statistics threshold, a low-precision signal is generated; if the inspection statistics value does not exceed the preset inspection statistics threshold, a high-precision signal is generated. The result accuracy verification module is connected to the entry and exit monitoring and judgment module. The entry and exit monitoring and judgment module is connected to the input impact assessment module and the output execution assessment module. The result accuracy verification module sends a high-precision result signal to the entry and exit monitoring and judgment module. When the entry and exit monitoring and judgment module receives the high-precision result signal, it monitors and assesses the conveyor belt's conveying status for automotive wiring harness precision fasteners through the input impact assessment module. Based on this, it generates an input qualified signal or an input alarm signal and sends the input qualified signal or input alarm signal to the entry and exit monitoring and judgment module. The intelligent decision feedback execution module monitors and evaluates the execution status of automotive wiring harness precision fasteners through the output execution evaluation module, and generates an output qualified signal or an output alarm signal accordingly. The output qualified signal or output alarm signal is then sent to the entry and exit monitoring judgment module. When the entry and exit monitoring judgment module receives an input alarm signal or an output alarm signal, it generates an alarm message and sends it to the operation supervision terminal. The specific operation process of the input impact assessment module is as follows: During the conveyor belt's conveying motion, the real-time conveying speed of the conveyor belt is acquired. The variance of all real-time conveying speeds within a unit time is calculated to obtain the conveying speed fluctuation value. The conveying speed fluctuation value is compared with a preset conveying speed fluctuation threshold. If the conveying speed fluctuation value exceeds the preset conveying speed fluctuation threshold, an input alarm signal is generated. If the transmission speed fluctuation value does not exceed the preset transmission speed fluctuation threshold, the difference between the real-time transmission speed and the set standard transmission speed is calculated and the absolute value is taken to obtain the speed difference detection value. The number of times the speed difference detection value exceeds the preset speed difference detection threshold within a unit time is marked as the speed difference abnormal value. The average value of all speed difference detection values ​​within a unit time is calculated to obtain the speed difference characteristic value. The speed difference abnormal value and the speed difference characteristic value are compared with the preset speed difference abnormal threshold and the preset speed difference characteristic threshold respectively. If the speed difference abnormal value or the speed difference characteristic value exceeds the corresponding preset threshold, an input alarm signal is generated. If the speed difference constant value and speed difference characteristic value exceed the corresponding preset threshold, when the conveyor belt is in a stopped state to cooperate with the acquisition operation of the multispectral synchronous imaging acquisition module, the vibration amplitude and vibration frequency of the conveyor belt are obtained. The vibration amplitude and vibration frequency are compared with the preset vibration amplitude threshold and preset vibration frequency threshold respectively. If the vibration amplitude or vibration frequency exceeds the corresponding preset threshold, it is determined that the conveyor belt is in a stopped unstable state. The duration of the conveyor belt in the stopped unstable state per unit time is obtained and the ratio is calculated with the total duration of the conveyor belt in the stopped state per unit time to obtain the stopped unstable time detection value. The proportion of the duration of the conveyor belt in the stopped unstable state in a single stop state is obtained and the proportion of the duration is compared with the corresponding duration proportion threshold. If the duration proportion exceeds the preset duration proportion threshold, the stopped non-optimal symbol TX1 is assigned to the corresponding stop state. The number of times the non-optimal stop symbol TX1 is assigned per unit time is obtained and the ratio is calculated with the total number of times the conveyor belt is in a stopped state to obtain the non-optimal stop detection value. The average value of the vibration amplitude and the average value of the vibration frequency when the conveyor belt is in a stopped state per unit time are marked as the amplitude characteristic value and the vibration frequency characteristic value, respectively. The stopping stability assessment value is calculated by weighting and summing the stopping non-stability detection time value, stopping non-optimal detection value, amplitude characteristic value and frequency characteristic value. The stopping stability assessment value is compared with the preset stopping stability assessment threshold. If the stopping stability assessment value exceeds the preset stopping stability assessment threshold, an input alarm signal is generated; if the stopping stability assessment value does not exceed the preset stopping stability assessment threshold, an input qualified signal is generated. The specific analysis process of the output execution evaluation module is as follows: All execution devices involved in the intelligent decision feedback execution module are obtained. When the intelligent decision feedback execution module issues a drive command, the operation delay duration of the corresponding execution device is collected. The operation delay duration is compared with the corresponding preset operation delay duration threshold. If the operation delay duration exceeds the corresponding preset operation delay duration threshold, the corresponding operation delay duration is marked as an operation delay abnormal value. The number of operation delay anomalies corresponding to the corresponding execution device within a unit time is obtained and the ratio of the number of executions of the corresponding execution device within a unit time is calculated to obtain the operation delay anomaly value. The average value of all operation delay durations corresponding to the corresponding execution device within a unit time is marked as the operation delay feature value. The abnormal operation value and the abnormal operation characteristic value are compared with the corresponding preset abnormal operation threshold and the preset abnormal operation characteristic threshold respectively. If the abnormal operation value or the abnormal operation characteristic value exceeds the corresponding preset threshold, the corresponding execution device is marked as a slow-action device. If a slow-action device exists within a unit of time, an output alarm signal is generated. If no easing device is present within a unit of time, a qualified output signal is generated.

2. The vision recognition-based surface defect detection system for automotive wiring harness precision fasteners according to claim 1, characterized in that, The multispectral synchronous imaging acquisition module adopts a ring LED array light source, integrates independent control units for visible light, near-infrared and ultraviolet bands, and uses an FPGA-based synchronous trigger controller to enable spatial alignment of different spectral images within microseconds. It also suppresses specular reflection on the surface of metal parts through a polarizer group, preserving diffuse reflection characteristic information.

3. The vision recognition-based surface defect detection system for automotive wiring harness precision fasteners according to claim 1, characterized in that, The operation process of the dynamic threshold adaptive enhancement module includes: The image is segmented into superpixel units, and the gray-level histogram distribution of each unit is calculated. An improved Otsu algorithm is used to dynamically generate local thresholds and nonlinearly stretch low-contrast regions. Wavelet denoising and guided filtering are combined to preserve the details of flawed edges while suppressing high-frequency noise.

4. The vision recognition-based surface defect detection system for automotive wiring harness precision fasteners according to claim 1, characterized in that, The operation process of the microscopic morphology feature deconstruction module includes: Directional texture features are extracted using a Gabor filter bank; a morphological gradient reconstruction algorithm is applied to generate a 3D topological map of the surface morphology; and spatial localization of defects across images is achieved through SIFT feature point matching.

5. The vision recognition-based surface defect detection system for automotive wiring harness precision fasteners according to claim 1, characterized in that, The multi-dimensional defect mapping and classification module constructs a defect feature library containing several sets of samples. During operation, the multi-dimensional defect mapping and classification module uses an improved Siamese network for feature similarity matching and introduces a confidence decay mechanism to trigger a secondary verification process for low-probability results. The intelligent decision feedback execution module is designed with a tiered rejection strategy. For high-confidence defective parts, the pneumatic rejection device is directly triggered. For edge cases, a visual report is generated and a manual re-inspection is prompted.

Images