Defect detection method, device and equipment for automobile parts

Abstract

The invention relates to the technical field of automatic detection, and discloses a defect detection method, device and equipment for automobile parts, and the method comprises the steps: fixing a detection object through a clamping structure, collecting image detection data, carrying out digital modeling based on the image detection data, and generating a detection information feedback model; carrying out same-batch defect detection and evaluation on the feedback model by utilizing a machine learning algorithm, formulating a subsequent process detection scheme, transmitting a detection object to a necessary and selected detection position, carrying out defect detection through a detection unit, obtaining supplementary detection data, substituting the supplementary detection data into the feedback model, carrying out evaluation, and generating a detection and evaluation label; according to the method, by combining image recognition, digital modeling and machine learning, part defect detection is automatically carried out in a multi-level mode, the detection precision and efficiency are improved, whole-process quality control is achieved, and the problem that in the prior art, automatic detection is difficult to consider cost and accuracy at the same time is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic detection, and in particular to a defect detection method, device and equipment for automobile parts. BACKGROUND

[0002] In the automobile manufacturing industry, the quality of parts directly affects the performance, safety and service life of vehicles, so efficient and accurate defect detection of automobile parts is an important link to ensure product quality. With the development of automation technology, more and more automobile manufacturing enterprises have begun to use automated detection equipment, such as automatic vision detection systems and machine vision systems. These systems use cameras and image processing software to automatically identify and detect parts. In order to detect parts more accurately, more types of detection units are needed, which increases costs and reduces efficiency. SUMMARY

[0003] The present application relates to the technical field of automatic detection, and in particular to a defect detection method, device and equipment for automobile parts.

[0004] The present application is implemented as follows: in a first aspect, the present application provides a defect detection method for automobile parts, comprising: The image recognition unit is used to identify the basic information of the automobile parts to be detected, and the automobile parts are marked as detection objects according to the identified basic information; The clamping structure is used to clamp the detection objects in a specified manner, and the image recognition unit is used to collect images of the detection objects to obtain image detection data of the detection objects; The detection information feedback model of the detection objects is obtained by digital modeling according to the image detection data, and the detection information feedback model is evaluated for defects in the same production batch by a pre-trained machine learning algorithm to obtain a subsequent process detection scheme; The detection objects are sequentially transmitted to necessary detection positions and selected detection positions by a conveying structure according to the subsequent process detection scheme, and the detection objects are detected for defects by detection units pre-deployed at the necessary detection positions and selected detection positions to obtain supplementary detection data; The detection objects are evaluated for defects in the same production batch by substituting the supplementary detection data into the detection information feedback model to obtain a detection evaluation label of the detection objects, and the detection result of the detection objects is described by the detection evaluation label.

[0005] In a second aspect, the present application provides a defect detection device for automobile parts, which is used to implement the defect detection method for automobile parts according to any one of the first aspect. An object marking module is configured to identify the basic information of the automobile parts to be detected by a pre-deployed image recognition unit, and mark the automobile parts as detection objects according to the identified basic information. An image detection module is configured to perform a specified form of clamping operation on the detection objects by a pre-deployed clamping structure, and perform image acquisition on the detection objects by the image recognition unit to obtain image detection data of the detection objects. A preliminary evaluation module is configured to perform digital modeling on the detection objects according to the image detection data to obtain a detection information feedback model of the detection objects, and perform defect detection evaluation of the same production batch on the detection information feedback model by a pre-trained machine learning algorithm to obtain a subsequent process detection scheme. A supplementary detection module is configured to sequentially transmit the detection objects to necessary detection positions and selection detection positions by a conveying structure according to the subsequent process detection scheme, and perform defect detection on the detection objects by detection units pre-deployed at the necessary detection positions and the selection detection positions to obtain supplementary detection data. A final evaluation module is configured to substitute the supplementary detection data into the detection information feedback model to perform defect detection evaluation of the same production batch on the detection objects to obtain a detection evaluation label of the detection objects, and describe the detection result of the detection objects by the detection evaluation label.

[0006] In a third aspect, the present application provides a computer device, which comprises a memory and a processor, the memory stores a computer program capable of running on the processor, and the processor implements the defect detection method for automobile parts according to any one of the first aspect when executing the computer program.

[0007] The present application provides a defect detection method for automobile parts, which has the following beneficial effects: The present invention fixes the inspection object through a clamping structure, collects image inspection data, performs digital modeling based on the image inspection data, generates an inspection information feedback model, uses a machine learning algorithm to evaluate the defect detection of the same batch on the feedback model, formulates a subsequent process inspection plan, transmits the inspection object to the necessary and selected inspection position, performs defect detection through the inspection unit, obtains supplementary inspection data, substitutes the supplementary inspection data into the feedback model, performs evaluation, generates inspection evaluation labels, and describes the inspection results. This method combines image recognition, digital modeling and machine learning to automatically and multi-levelly perform component defect detection, improves inspection accuracy and efficiency, realizes quality control of the entire process, and solves the problem that automatic inspection in the existing technology is difficult to strike a balance between cost and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a schematic diagram of the steps of a defect detection method for automotive parts provided by an embodiment of the present invention; Figure 2 The figure is a schematic structural diagram of a defect detection device for automobile parts provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0009] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0010] The implementation of the present invention is described in detail below with reference to specific embodiments.

[0011] Reference Figure 1 、 Figure 2 As shown, a preferred embodiment of the present invention is provided.

[0012] In a first aspect, the present invention provides a method for detecting defects in automotive parts, comprising: S1: Using a pre-deployed image recognition unit to identify basic information of the automobile parts to be inspected, and marking the automobile parts as inspection objects based on the identified basic information; S2: performing a specified clamping operation on the detection object by using a pre-deployed clamping structure, and performing image acquisition on the detection object by using the image recognition unit to obtain image detection data of the detection object; S3: Digitally modeling the inspection object based on the image inspection data to obtain an inspection information feedback model of the inspection object, and performing defect inspection evaluation of the same production batch on the inspection information feedback model using a pre-trained machine learning algorithm to obtain a subsequent process inspection plan; S4: sequentially transporting the detection object to the necessary detection position and the selective detection position through the conveying structure according to the subsequent flow detection scheme, and performing defect detection on the detection object by the detection unit pre-deployed at the necessary detection position and the selective detection position to obtain supplementary detection data; S5: substituting the supplementary detection data into the detection information feedback model to perform defect detection evaluation on the detection object of the same production batch to obtain a detection evaluation label of the detection object, and describing the detection result of the detection object through the detection evaluation label.

[0013] Specifically, in step S1 of the embodiments provided in the present application, image recognition units are pre-deployed on the production line or detection equipment to ensure that they can capture image information of automobile parts. These image recognition units usually include high-resolution cameras and lighting equipment to ensure that the images of the parts can be clearly obtained.

[0014] More specifically, the image recognition unit captures the image of the automobile part, performs image preprocessing, including denoising, enhancement, correction, etc., to improve the image quality and facilitate subsequent recognition work. Through a pre-trained image recognition algorithm (such as a convolutional neural network CNN), the preprocessed image is analyzed to extract the features of the part and identify the basic information of the part (such as the model, specification, production batch, serial number, etc.). According to the identified basic information, the automobile part is marked as a detection object. This process may include recording the identification information of the part in a database and generating a unique identification code (such as a two-dimensional code or an RFID tag) for subsequent tracking and management.

[0015] More specifically, the parts marked as detection objects are classified into different detection processes, and according to the identified basic information, it is determined which specific detection items the part needs to perform and is assigned to the corresponding detection unit.

[0016] It can be understood that the automatic recognition and marking process reduces the need for manual work, improves the recognition and marking efficiency, and reduces the possibility of human error. Through high-precision image recognition technology, the basic information of the part can be accurately identified to ensure that each part can be accurately detected and tracked. Automatic basic information recognition and marking make data management of the parts more convenient. All recognition and detection data can be recorded and stored in real time for subsequent data analysis and quality control. The pre-deployed image recognition unit can be adjusted according to different parts and production needs, has high flexibility and adaptability, and can adapt to the detection needs of various parts. Through accurate marking and classification of the detection object, quality control can be better performed. The recognition and detection data can be fed back to the production and detection system in real time to help timely discover and solve quality problems.

[0017] Specifically, in step S2 of the embodiments provided by the present application, a clamping structure suitable for different automobile parts is designed to ensure that it can firmly clamp the parts without damaging their surfaces. The clamping structure can take various forms such as a mechanical arm, a vacuum chuck, a pneumatic clamp, etc. The clamping structure is deployed on the production line or in the detection equipment. According to the shape and size of the parts, the parameters of the clamping structure (such as clamping force, position, etc.) are adjusted. The detection object (automobile parts) is subjected to a specified form of clamping operation by the clamping structure to ensure that the parts remain stable during detection and avoid movement or vibration. The clamping structure fixes the parts at a predetermined position so that the image recognition unit can capture images from the best angle. The parameters of the light source and camera are adjusted to ensure uniform light and clear images. The image recognition unit (usually including a high-resolution camera and an illumination system) captures images of the clamped detection object. The captured images may include multiple angles and different parts to fully capture the appearance features of the parts.

[0018] More specifically, the captured images are preprocessed, such as denoising, enhancement, correction, etc. to improve image quality. Image recognition algorithms (such as convolutional neural networks CNN) are applied to analyze the preprocessed images, extract relevant features, and generate detection data. The acquired image detection data is recorded in the database and analyzed to determine the quality status of the parts and identify possible defects or problems.

[0019] It can be understood that the clamping structure firmly clamps the detection object, reducing the image blurring phenomenon caused by movement or vibration of the parts, improving the stability and accuracy of image acquisition, and adjusting the posture of the parts to enable the image recognition unit to capture images of the parts from multiple angles, ensuring comprehensive detection and coverage of the parts. The automated clamping and image acquisition process reduces manual intervention, improves detection efficiency, and reduces the complexity and error rate of manual operation. The clamping structure design is flexible and can adapt to parts of different shapes and sizes to meet various detection needs. By adjusting the clamping parameters and image acquisition settings, it can quickly adapt to the requirements of different detection tasks. The detection data generated by the automated image acquisition and processing process can be recorded in the database in real time for subsequent analysis and quality control. Through analysis of the detection data, quality problems existing in the production process can be discovered and fed back in a timely manner to improve the overall quality management level.

[0020] Specifically, in step S3 of the embodiments provided by the present application, high-quality image detection data of the detection object (including images of multiple angles and multiple parts) is obtained from the image recognition unit, the collected image data is preprocessed, such as denoising, enhancement, image correction, etc., to improve the data quality, the image is segmented and feature extracted, etc., to extract key feature points and regions, the preprocessed image data is used to apply three-dimensional reconstruction technology (such as structured light scanning, stereo vision, deep learning algorithm, etc.) for digital modeling, and a three-dimensional model of the detection object is generated to ensure that the model accurately reflects the shape, size and surface features of the detection object.

[0021] More specifically, the digital modeling result is compared with the detection standard to mark the possible defects or abnormal areas of the detection object, and a detection information feedback model is generated, including the three-dimensional model of the part and the defect marking and detection result.

[0022] More specifically, a large amount of historical detection data and feedback model are used to pre-train machine learning algorithms (such as convolutional neural network CNN, support vector machine SVM, etc.) to have the ability to identify and evaluate defects, and the detection information feedback model is input into the pre-trained machine learning algorithm for defect detection and evaluation of the same production batch.

[0023] More specifically, the machine learning algorithm analyzes the feedback model to evaluate the defect type, severity and distribution of the detection object, and performs batch evaluation on the detection objects of the same production batch to find common defects and characteristic defects, and according to the defect detection and evaluation result, a detection scheme for the subsequent process is formulated, including key detection areas, detection frequency, detection methods, etc., to provide improvement suggestions, optimize production process and detection process, and reduce defect occurrence rate.

[0024] It can be understood that through the combination of digital modeling and machine learning algorithm, the defects of the detection object can be accurately identified and evaluated, the detection accuracy and reliability are improved, the three-dimensional model generated by digital modeling can fully reflect the appearance and structural features of the detection object, combined with the evaluation of the machine learning algorithm, the defect type and distribution can be fully analyzed, the pre-trained machine learning algorithm is used for defect evaluation, which reduces the manual intervention and improves the automation degree and efficiency of the detection process, the detection information feedback model and the defect detection and evaluation result provide data support for production and quality control, and help to make data-based optimization decisions.

[0025] Specifically, in step S4 of the embodiments provided by the present application, a conveying structure (such as a conveyor belt, a mechanical arm, a conveying track, etc.) is designed and deployed to ensure that the detection object can be stably and continuously transmitted to each detection position, the necessary detection position and the selected detection position are determined, the conveying path is reasonably arranged to ensure the efficiency and smoothness of the detection process.

[0026] More specifically, detection units (such as visual inspection systems, laser scanners, X-ray equipment, etc.) are deployed at necessary detection positions and selective detection positions, ensuring that they can cover the key parts and important features of the detection object. According to the detection requirements, the parameters and positions of the detection units are adjusted to optimize the detection effect.

[0027] More specifically, the conveying structure is started to transport the detection object to the predetermined necessary detection positions and selective detection positions one by one. At each detection position, the detection object is ensured to be stably stopped and accurately positioned for precise detection. At the necessary detection position, the detection object is comprehensively detected by the detection unit to obtain preliminary defect data. At the selective detection position, further fine detection is carried out for certain specific parts or suspicious areas to obtain supplementary detection data.

[0028] More specifically, each detection unit collects detection data (images, scanning results, etc.) and transmits the data to the central processing system for processing and analysis. Defect information is extracted and compared with the previous detection information feedback model to ensure the integrity and accuracy of the data.

[0029] More specifically, the detection process is monitored in real time to ensure the normal operation of the detection unit and the conveying structure, and the detection results are fed back in time. When defects are found, they are immediately recorded and marked for subsequent processing and improvement.

[0030] More specifically, the supplementary detection data is integrated with the preliminary detection data to form a complete detection report. Through data analysis, the overall quality status of the detection object is evaluated, common defects and individual defects are identified, and improvement suggestions are put forward.

[0031] It can be understood that through the reasonable arrangement of necessary detection positions and selective detection positions, comprehensive coverage and precise detection of the detection object are realized, ensuring that each key part can be fully detected. The design and deployment of the conveying structure improve the transmission efficiency and positioning accuracy of the detection object, reduce the stagnation and error in the detection process, and the real-time monitoring and feedback mechanism ensures the normal operation of the detection unit and the conveying structure. It can timely discover and handle abnormal situations in the detection process, improve the detection efficiency and accuracy, and the collection and integration of supplementary detection data make the detection results more comprehensive and accurate, which is helpful for comprehensive analysis and evaluation of the quality status of the detection object.

[0032] It should be noted that for the same production batch of automobile parts, the failure of the production equipment is the cause of the failure of the automobile parts, and these possible failures can be divided into high probability and low probability. Generally, important defect detection is performed for high probability defects, and in this step, the necessary detection unit of the necessary detection position is used to detect the necessity of the high probability defects, and the selective detection unit of the selected detection position is used to selectively detect the low probability defects. The specific implementation scheme of selective detection is that each component in the same production batch is allocated a different detection device for parallel detection to improve the detection effect.

[0033] Specifically, in step S5 of the embodiment provided by the present application, in the detection process, the detection unit comprehensively detects the detection object, collects supplementary detection data (such as images, measurement data, etc.), integrates the supplementary detection data with the preliminary detection data to ensure the integrity and accuracy of the data, pre-processes the collected supplementary detection data, including data cleaning, denoising, standardization and other processing steps to improve data quality and consistency, and inputs the pre-processed supplementary detection data into the detection information feedback model. The detection information feedback model analyzes and processes the input data through a pre-trained algorithm (such as a machine learning model, a deep learning network, etc.).

[0034] More specifically, the detection information feedback model processes the input data to identify defect features in the detection object, performs defect detection evaluation on the detection object according to the defect features, determines the type, severity and distribution of the defects, generates a detection evaluation label according to the detection evaluation result to describe the detection result of the detection object, and the detection evaluation label includes defect type, number, location, severity and other information to facilitate subsequent processing and decision-making. The detection evaluation label is integrated with other information of the detection object to form a complete detection report, and the detection evaluation label is fed back to the production management system for quality control and production decision-making reference.

[0035] It can be understood that through the combination of supplementary detection data and detection information feedback model, the defects of the detection object can be more accurately identified and evaluated, the detection precision and accuracy are improved, the defects of the detection object of the same production batch are evaluated, the quality status of the production batch is comprehensively evaluated, the batch problems and common defects are identified, the generation and feedback mechanism of the detection evaluation label ensures that the detection result can be fed back to the production management system in real time, supports rapid response and decision-making, and timely handles the problems in production. The detection information feedback model analyzes historical data and real-time data to continuously optimize and improve the detection process and quality control strategy, promotes the continuous improvement of quality, the detection evaluation label provides detailed defect information, and the defect data is displayed through a visualization tool for easy tracing and analysis, supporting quality management and improvement measures.

[0036] The application provides a defect detection method for automobile parts, which has the following beneficial effects: The application fixes the detection object through the clamping structure, collects image detection data, performs digital modeling based on the image detection data, generates a detection information feedback model, evaluates the feedback model for same-batch defect detection by using a machine learning algorithm, formulates a subsequent process detection scheme, transmits the detection object to necessary and selective detection positions, performs defect detection through a detection unit, obtains supplementary detection data, substitutes the supplementary detection data into the feedback model, performs evaluation, generates a detection evaluation label, and describes the detection result. The method automatically and multi-levelly detects the defects of the parts by combining image recognition, digital modeling and machine learning, improves the detection precision and efficiency, realizes full-process quality control, and solves the problem that automatic detection in the prior art cannot balance cost and accuracy.

[0037] Preferably, the step of marking the automobile part to be detected as a detection object according to the basic information obtained by identification includes: S11: conveying the automobile part to be detected to the first detection position through the conveying structure, and collecting the image of the part appearance of the automobile part at the first detection position through the pre-deployed image recognition unit to obtain the appearance image data of the automobile part at the first detection position; S12: matching the appearance image data according to the detection database, and analyzing the part type and part specification of the automobile part according to the result of information matching to obtain the basic information composed of the type information and specification information of the automobile part; S13: registering the basic information through the specified data processing platform to mark the automobile part as a detection object.

[0038] Specifically, the automobile part to be detected is conveyed to the first detection position through a conveying belt, a mechanical arm or other conveying structure. At the first detection position, the appearance image of the automobile part is collected by using the pre-deployed image recognition unit (such as a high-definition camera, a machine vision system, etc.). The collected appearance image data includes the appearance, shape, size and other information of the automobile part.

[0039] More specifically, the collected topographic image data is matched with the standard data in the detection database, which contains the standard topographic image of various automobile parts and the corresponding type and specification information, and the image recognition algorithm (such as pattern recognition, machine learning algorithm, etc.) is used for information matching to identify the type and specification of the automobile parts. According to the information matching result, the type and specification of the automobile parts are analyzed to obtain the type information and specification information of the automobile parts, and the obtained type information and specification information are integrated to form the basic information of the automobile parts.

[0040] More specifically, the obtained basic information is registered through a designated data processing platform (such as an MES system, a quality management system, etc.), and the automobile parts that have completed information registration are marked as detection objects and enter the subsequent detection process.

[0041] It can be understood that the use of the conveying structure and the automatic operation of the image recognition unit can quickly and continuously identify and mark a large number of automobile parts, improve the overall efficiency of the detection process, and improve the accuracy and consistency of the identification result based on the information matching of the detection database. The errors and deviations that may be caused by manual identification are avoided, and the basic information obtained by identification is registered in real time to the data processing platform, ensuring that the basic information of the detection object can be recorded in time and accurately, supporting subsequent detection and quality control.

[0042] Preferably, the step of performing a specified form of clamping operation on the detection object through a pre-deployed clamping structure and performing image acquisition on the detection object through the image recognition unit to obtain image detection data of the detection object comprises: S21: analyzing the image acquisition form of the detection object according to the pre-set automobile part image acquisition detection standard on the data processing platform to obtain a set of image acquisition forms of the detection object; wherein the set of image acquisition forms includes a plurality of preliminary acquisition forms and image acquisition effects corresponding to each preliminary acquisition form; S22: multiple combination simulation of image acquisition effects of each preliminary acquisition form in the set of image acquisition forms is performed to obtain a sequence of image detection forms composed of each preliminary acquisition form and an overall detection effect corresponding to each image detection form sequence, and a weight evaluation of the execution feasibility of the image detection form sequence is performed according to the overall detection effect to obtain an evaluation reference selection weight of each image detection form sequence; S23: retrieve the image detection form sequence used by the detection object with the same type and specification as the detection object from the historical database as historical reference information, analyze the execution times and execution frequencies of each image detection form sequence of the detection object according to the historical reference information, and generate selection weights for each image detection form of the detection object based on the analysis results to obtain historical reference selection weights of each image detection form; S24: weighting fusion and interactive evaluation of the evaluation reference selection weights and the historical reference selection weights of each image detection form sequence are performed to determine the final executed image detection form sequence, and the detection object is sequentially subjected to clamping operation conforming to the image detection form sequence through the pre-deployed clamping structure, so that the detection object is sequentially in the position form corresponding to the image detection form in front of the image recognition unit; S25: image acquisition of the detection object in various position forms is performed through the image recognition unit to obtain form image data of the detection object in various position forms, and the form image data of the detection object in various position forms collectively constitute the image detection data of the detection object.

[0043] Specifically, according to the image acquisition detection standard pre-set on the data processing platform, the characteristics of the detection object and the applicable image acquisition form are analyzed, the image recognition unit is used to preliminarily scan the detection object, the shape and key parts are identified, different image acquisition forms may be applicable to different parts, and the best acquisition method can be determined through analysis to ensure the comprehensiveness and accuracy of the image data.

[0044] More specifically, an image acquisition form set is established, which contains a plurality of pre-acquisition forms and respective image acquisition effects, and each acquisition form is provided with specific operation parameters such as angle, lighting condition, etc. By pre-setting a plurality of acquisition forms, various choices can be provided for subsequent image acquisition, and the data quality can be improved.

[0045] More specifically, the pre-acquisition forms in the set are combined and simulated to generate a plurality of image detection form sequences. During the simulation process, the overall detection effect of each combination is evaluated, and through simulation of different combinations, the best image detection form sequence can be identified to optimize the detection process.

[0046] More specifically, according to the overall detection effect of the simulation, the feasibility of each image detection form sequence is evaluated to generate evaluation reference selection weights. Through weight evaluation, the detection form sequence with the best execution effect can be screened out to improve the detection efficiency and accuracy.

[0047] More specifically, the image detection form sequence of the same type and specification of the parts in the history database is called, the execution times and frequencies are analyzed, the history reference selection weight is generated, the history data can provide reference basis, repeated test is avoided, and the reliability of the detection process is improved.

[0048] More specifically, the evaluation reference selection weight is weighted and fused with the history reference selection weight, and the final executed image detection form sequence is obtained, the current evaluation and the history data are comprehensively considered, and it is ensured that the selected image detection form sequence is optimal.

[0049] More specifically, through the pre-deployed clamping structure, the detection object is clamped according to the determined image detection form sequence, so that it is in each preset position form in turn, the accurate clamping operation ensures that the detection object is in the best image acquisition position, the image quality is improved, the image recognition unit is used to acquire the form image data of the detection object in various position forms, the image acquisition in multiple angles and multiple positions ensures that the image data of the detection object is comprehensive and accurate.

[0050] More specifically, the form image data of the detection object in various position forms is integrated to generate complete image detection data, which ensures that the image detection data comprehensively covers each detail of the detection object and provides reliable data for subsequent analysis.

[0051] It can be understood that through the image acquisition in multiple angles and multiple positions, it is ensured that the image data of the detection object is comprehensive, the preset standard and the history data are used for multiple evaluations, the optimal image detection form sequence is selected, the detection accuracy is improved, the automatic clamping and image acquisition operation is improved, the detection efficiency is improved, the human intervention is reduced, the history data and real-time simulation evaluation are used to ensure the reliability and repeatability of the detection process.

[0052] Preferably, the detection object is digitally modeled according to the image detection data to obtain a detection information feedback model of the detection object, and a same production batch defect detection evaluation is performed on the detection information feedback model through a pre-trained machine learning algorithm to obtain a subsequent process detection scheme, and the steps include: S31: The image detection data is substituted into a data processing platform, the data processing platform obtains an image detection form sequence of a clamping operation of the clamping structure on the detection object, the data processing platform positions an information feedback mode of the image detection data according to the image detection form sequence, and an image data feedback mode of each form image data in the image detection data is obtained. S32: retrieve the standard digital model of the detection object from the prepared model database, and perform spatio-temporal distribution analysis on each of the modality image data according to the image data feedback mode corresponding to each of the modality image data, to obtain an image detection feature matrix of each of the modality image data; wherein the image detection feature matrix comprises a plurality of image detection feature vectors having a spatio-temporal distribution relationship; S33: perform spatio-temporal mapping of the image detection feature vectors on the standard digital model according to each of the image detection feature matrix, to assign detection information feedback parameters corresponding to each of the image detection feature matrix on the standard digital model, and perform multi-dimensional confidence analysis on the detection information feedback parameters corresponding to each specific position of the standard digital model, to obtain final feedback parameters of each specific position of the standard digital model; S34: perform detailed reconstruction of the model parameters of the standard digital model according to the final feedback parameters of each specific position of the standard digital model, to obtain a detection information feedback model for digital feedback of the detection object; S35: perform detection information analysis of surface defects and spatio-temporal correlation prediction analysis of other form defects on the detection information feedback model through a pre-trained machine learning algorithm, to obtain surface defect analysis feature distribution and potential defect analysis feature distribution of the detection object; S36: upload the surface defect analysis feature distribution and potential defect analysis feature distribution of the detection object, to perform time sequence integration of the surface defect analysis feature distribution and potential defect analysis feature distribution of the detection object through a data processing platform, to obtain an analysis sequence composed of surface defect distribution features and potential defect analysis feature distribution between the detection object and a plurality of detection objects of the same production batch; S37: perform analysis of production defect elements on the production equipment of the detection object according to the analysis sequence, to obtain a possible production defect element of the production equipment of the detection object in the current production batch, and perform usage demand analysis of a specific detection unit on the detection object according to the possible production defect element, to obtain a specific detection unit demand feature corresponding to the possible production defect element of the detection object; S38: retrieve the specific detection unit usage records of the detection objects of the same production batch that have been executed, and perform demand implementation analysis on the specific detection unit demand feature according to the specific detection unit usage records, to obtain a specific detection unit that needs to be used in the subsequent detection process of the detection object; wherein the specific detection unit is deployed at a selected detection position; S39: determining necessary detection units required for the detection object in the subsequent detection process according to the preset standard, and arranging the detection object to the subsequent detection instrument according to the specific detection unit and the necessary detection unit to obtain a subsequent process detection scheme; wherein the necessary detection unit is arranged at a necessary detection position.

[0053] Specifically, the image detection data is input into the data processing platform, the data processing platform obtains an image detection form sequence of the clamping operation of the clamping structure on the detection object, and the image detection data is positioned according to an information feedback mode of the image detection form sequence to obtain an image data feedback mode of each morphological image data.

[0054] More specifically, a standard digital model of the detection object is called from a preliminary model database, the morphological image data is analyzed for the spatio-temporal distribution of feedback information according to the image data feedback mode corresponding to each morphological image data to obtain an image detection feature matrix, and the image detection feature matrix includes a plurality of image detection feature vectors having a spatio-temporal distribution relationship.

[0055] More specifically, the standard digital model is subjected to spatio-temporal mapping of the image detection feature vectors according to each image detection feature matrix, the detection information feedback parameters corresponding to each image detection feature matrix are assigned to the standard digital model, the feedback parameters corresponding to specific positions of the standard digital model are subjected to multi-dimensional certainty analysis to obtain final feedback parameters, the model parameters of the standard digital model are subjected to detailed reconstruction according to the final feedback parameters, and a detection information feedback model is generated.

[0056] More specifically, the detection information feedback model is subjected to detection information analysis of surface defects and spatio-temporal correlation prediction analysis of other forms of defects using a pre-trained machine learning algorithm to obtain a surface defect analysis feature distribution and a potential defect analysis feature distribution of the detection object, the surface defect analysis feature distribution and the potential defect analysis feature distribution are uploaded to the data processing platform for time sequence integration to obtain an analysis sequence composed of the surface defect distribution feature and the potential defect analysis feature distribution between the current detection object and a plurality of detection objects of the same production batch.

[0057] More specifically, the production equipment of the detection object is analyzed for production defect elements according to the analysis sequence to obtain possible production defect elements, the use demand of the specific detection unit is analyzed according to the possible production defect elements to obtain specific detection unit demand characteristics, the specific detection unit use records of the detection objects of the same production batch that have been executed are called, the demand implementation analysis of the specific detection unit demand characteristics is performed according to the specific detection unit use records, and the specific detection unit required in the subsequent detection process is determined.

[0058] More specifically, the necessary detection units required for the subsequent detection process are determined according to the preset standard, the arrangement of the detection object by the specific detection unit and the necessary detection unit is arranged according to the subsequent detection instrument, and the subsequent process detection scheme is obtained.

[0059] It can be understood that the defects of the detection object are accurately positioned through the digital model and the feature matrix analysis, the detection accuracy is improved, the machine learning algorithm is used for defect analysis and prediction, the detection efficiency is improved, the comprehensive analysis of the defect distribution of the detection object and the same batch object is realized through the space-time distribution analysis and time sequence integration, the production defect elements are analyzed, the defect possibility of the production equipment under the current batch is identified, which is helpful for the optimization of the production process, the reliability and effectiveness of the detection scheme are ensured through the multi-dimensional confidence analysis and demand implementation analysis.

[0060] Preferably, the step of transmitting the detection object to the necessary detection position and the selection detection position in sequence according to the subsequent process detection scheme and detecting the defects of the detection object by the detection unit deployed in advance at the necessary detection position and the selection detection position to obtain the supplementary detection data comprises: S41: transmitting the detection object to the necessary detection position through the conveying structure according to the subsequent process detection scheme, and detecting the detection object by the necessary detection unit deployed in advance at the necessary detection position to obtain the first supplementary detection data of the detection object; S42: transmitting the detection object to the selection detection position through the conveying structure according to the subsequent process detection scheme, and detecting the detection object by the selection detection instrument deployed in advance at the selection detection position to obtain the second supplementary detection data of the detection object; S43: the first supplementary detection data and the second supplementary detection data are combined to obtain the supplementary detection data of the detection object.

[0061] Specifically, the conveying structure is initialized to ensure that it can accurately and reliably transmit the detection object to each set position, the detection object is transmitted to the necessary detection position by the conveying structure according to the subsequent process detection scheme, the stability and accurate positioning of the detection object during the conveying process are ensured, the detection object is comprehensively detected by the necessary detection unit deployed in advance at the necessary detection position, and the first supplementary detection data of the detection object is recorded and stored.

[0062] More specifically, according to the subsequent process detection scheme, the detection object is transmitted to the selected detection position through the conveying structure, ensuring the stability and accurate positioning of the detection object during the conveying process, and further detection of the detection object is carried out by the pre-deployed selected detection unit (selected detection instrument) at the selected detection position, the second supplementary detection data of the detection object is recorded and stored, the first supplementary detection data and the second supplementary detection data are summarized and combined to form complete supplementary detection data, ensuring the integrity and accuracy of the data.

[0063] It can be understood that through multiple, multi-angle detection of the detection object by the necessary detection unit and the selected detection unit, the comprehensiveness and accuracy of the detection data are ensured, the combination of the first supplementary detection data and the second supplementary detection data makes the defect information of the detection object more complete, the use of the conveying structure enables the detection object to be quickly and accurately transmitted to each detection position, improving the efficiency of the overall detection process, and the automatic operation of the conveying structure reduces human intervention and reduces errors and time consumption in the detection process.

[0064] Preferably, the supplementary detection data is substituted into the detection information feedback model, and the detection object is subjected to defect detection evaluation of the same production batch to obtain a detection evaluation label of the detection object, and the step of describing the detection result of the detection object through the detection evaluation label comprises: S51: Substitute the supplementary detection data into the detection information feedback model, and let the detection information feedback model analyze the data content of the supplementary detection data to obtain a supplementary detection feature matrix of the supplementary detection data relative to the detection information feedback model; wherein the supplementary detection feature matrix comprises a plurality of supplementary detection feature vectors having a space-time distribution relationship; S52: According to the necessary detection position and the selected detection position corresponding to the supplementary detection data, analyze the information fusion value of the supplementary detection feature matrix to obtain an information fusion mode of the supplementary detection feature matrix; S53: According to the information fusion mode of the supplementary detection feature matrix of the same production batch, mutually interactively fuse each supplementary detection feature matrix of the same production batch to obtain comprehensive feedback information, and based on the comprehensive feedback information, evaluate and process various defects in the preset standard to obtain a detection evaluation label of the detection object.

[0065] Specifically, the first supplementary detection data and the second supplementary detection data are substituted into the pre-constructed detection information feedback model, and the detection information feedback model is used to analyze the data content of the supplementary detection data. The detection information feedback model analyzes the supplementary detection data to generate a supplementary detection feature matrix, and the supplementary detection feature matrix comprises a plurality of supplementary detection feature vectors having a space-time distribution relationship.

[0066] More specifically, according to the necessary detection position and the selected detection position corresponding to the supplementary detection data, the supplementary detection feature matrix is analyzed, the information fusion value is evaluated, the information fusion mode of the supplementary detection feature matrix is obtained, the information fusion of each supplementary detection feature matrix of the same production batch is interacted according to the information fusion mode of the supplementary detection feature matrix of the batch, and comprehensive feedback information is generated, which contains all related detection features and space-time distribution information.

[0067] More specifically, based on the comprehensive feedback information, the detection object is evaluated for each type of defect in the preset standard, and all detection features and information fusion modes are considered in the evaluation process to ensure the accuracy and comprehensiveness of the evaluation. According to the results of the defect evaluation process, a detection evaluation label of the detection object is generated, which describes the detection results of the detection object in detail, including the types, severity and distribution of defects.

[0068] It can be understood that by analyzing the supplementary detection feature matrix, the characteristics of the detection object at different detection positions can be comprehensively understood. The application of the information fusion mode makes the evaluation result more accurate and reduces the deviation that may be caused by single detection data. The comprehensive feedback information generated by the information fusion process covers all related data of the detection object, provides multi-dimensional and multi-angle evaluation basis, ensures the comprehensiveness of the detection evaluation, and greatly improves the processing efficiency of the supplementary detection data. The application of the detection information feedback model greatly improves the processing efficiency of the supplementary detection data, and the automatic data analysis and fusion process reduces human intervention and improves the efficiency of the overall detection process.

[0069] Referring to Figure 2 The second aspect, the present application provides a kind of automobile parts defect detection device, for realizing the automobile parts defect detection method of any one of the first aspect, it includes: Object marking module, for the basic information identification of automobile parts to be detected by pre-deployed image recognition unit, according to the basic information identified by the automobile parts to be detected Mark as detection object; Image detection module, for the specified form of clamping operation to the detection object by pre-deployed clamping structure, and the image acquisition of the detection object by the image recognition unit, to obtain the image detection data of the detection object; Preliminary evaluation module, for the digital modeling of the detection object according to the image detection data, to obtain the detection information feedback model of the detection object, and the defect detection evaluation of the same production batch to the detection information feedback model by pre-trained machine learning algorithm, to obtain subsequent process detection scheme; The supplementary detection module is configured to sequentially transmit the detection object to necessary detection positions and selection detection positions through the conveying structure according to the subsequent process detection scheme, and perform defect detection on the detection object by the detection units pre-deployed at the necessary detection positions and the selection detection positions, to obtain supplementary detection data. The final evaluation module is configured to substitute the supplementary detection data into the detection information feedback model, perform defect detection evaluation on the detection object of the same production batch, to obtain a detection evaluation label of the detection object, and describe the detection result of the detection object by using the detection evaluation label.

[0070] In the embodiment, the specific implementation of each module in the device embodiment is described above in the method embodiment, and will not be described here.

[0071] In a third aspect, the present application provides a computer device, comprising a memory and a processor, the memory stores a computer program capable of running on the processor, and the processor implements the defect detection method of the automobile part according to any one of the first aspect when executing the computer program.

[0072] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement and improvement within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A defect detection method for automobile parts, characterized in that: include: Identify basic information of the automobile parts to be inspected by using a pre-deployed image recognition unit, and mark the automobile parts as inspection objects based on the identified basic information; Performing a specified clamping operation on the detection object by using a pre-deployed clamping structure, and collecting an image of the detection object by using the image recognition unit to obtain image detection data of the detection object; Digitally modeling the inspection object based on the image inspection data to obtain an inspection information feedback model of the inspection object, and performing defect detection evaluation of the same production batch on the inspection information feedback model using a pre-trained machine learning algorithm to obtain a subsequent process inspection plan; According to the subsequent process inspection plan, the inspection object is sequentially transported to the necessary inspection position and the selected inspection position through the conveying structure, and the inspection units pre-deployed at the necessary inspection position and the selected inspection position perform defect inspection on the inspection object to obtain supplementary inspection data; The supplementary inspection data is substituted into the inspection information feedback model, and defect inspection and evaluation of the same production batch is performed on the inspection object to obtain an inspection evaluation label of the inspection object, and the inspection result of the inspection object is described by the inspection evaluation label.

2. The defect detection method for automobile parts according to claim 1, characterized in that: The steps of identifying basic information of the automobile parts to be inspected by a pre-deployed image recognition unit and marking the automobile parts as inspection objects according to the identified basic information include: The automobile component to be inspected is transported to a first inspection position by a transport structure, and an image of the component morphology of the automobile component at the first inspection position is captured by a pre-deployed image recognition unit to obtain morphology image data of the automobile component at the first inspection position; Performing information matching on the topographic image data according to a detection database, and parsing the type and specification of the automotive parts according to the result of the information matching to obtain basic information consisting of type information and specification information of the automotive parts; The basic information is registered through a designated data processing platform to mark the automobile parts as detection objects.

3. The defect detection method for automobile parts according to claim 1, wherein: The steps of performing a specified clamping operation on the detection object by using a pre-deployed clamping structure and acquiring an image of the detection object by using the image recognition unit to obtain image detection data of the detection object include: Analyzing the image acquisition form of the inspection object according to the automobile parts image acquisition inspection standard preset on the data processing platform to obtain an image acquisition form set of the inspection object; wherein the image acquisition form set includes a plurality of preliminary acquisition forms and image acquisition effects corresponding to the various preliminary acquisition forms; performing multiple combination simulations of image acquisition effects on each of the preliminary acquisition forms in the image acquisition form set to obtain an image detection form sequence composed of each of the preliminary acquisition forms and an overall detection effect corresponding to each of the image detection form sequences, and performing a weighted evaluation of the execution feasibility of the image detection form sequence based on the overall detection effect to obtain an evaluation reference selection weight for each of the image detection form sequences; Retrieving, from a historical database, image detection form sequences used by detection objects of the same type and specification as the detection object as historical reference information, analyzing the number of executions and execution frequencies of each image detection form sequence of the detection object based on the historical reference information, and generating selection weights for each image detection form of the detection object based on the analysis results to obtain historical reference selection weights for each image detection form; The evaluation reference selection weights of each of the image detection form sequences and the historical reference selection weights are weightedly fused and interactively evaluated to determine a final image detection form sequence to be executed, and the pre-deployed clamping structure is used to sequentially perform clamping operations on the detection object in accordance with the image detection form sequence, so that the detection object is sequentially positioned in front of the image recognition unit in a position corresponding to the image detection form; The image recognition unit is used to collect images of the detection object in various positions and forms to obtain morphological image data of the detection object in various positions and forms. The morphological image data of the detection object in various positions and forms together constitute the image detection data of the detection object.

4. The defect detection method for automobile parts according to claim 3, characterized in that: The steps of digitally modeling the inspection object based on the image inspection data to obtain an inspection information feedback model of the inspection object, and performing defect detection evaluation of the same production batch on the inspection information feedback model using a pre-trained machine learning algorithm to obtain a subsequent process inspection plan include: Substituting the image detection data into a data processing platform, the data processing platform simultaneously obtains an image detection form sequence of the clamping operation performed by the clamping structure on the detection object, and instructs the data processing platform to position the image detection data in an information feedback manner according to the image detection form sequence, thereby obtaining an image data feedback mode for each form image data in the image detection data; Retrieving a standard digital model of the detection object from a prepared model database, and performing spatiotemporal distribution analysis of feedback information on each of the morphological image data according to an image data feedback mode corresponding to each of the morphological image data in the image detection data, so as to obtain an image detection feature matrix for each of the morphological image data; wherein the image detection feature matrix includes a plurality of image detection feature vectors having a spatiotemporal distribution relationship; performing spatiotemporal mapping of image detection feature vectors on the standard digital model according to each of the image detection feature matrices to assign detection information feedback parameters corresponding to each of the image detection feature matrices to the standard digital model, and performing multi-dimensional confidence analysis on the detection information feedback parameters corresponding to each specific position of the standard digital model to obtain final feedback parameters for each specific position of the standard digital model; Reconstructing detailed model parameters of the standard digital model according to final feedback parameters of specific locations of the standard digital model to obtain a detection information feedback model for providing digital feedback to the detection object; Performing surface defect detection information analysis and spatiotemporal correlation prediction analysis of other defects on the detection information feedback model using a pre-trained machine learning algorithm to obtain surface defect analysis feature distribution and potential defect analysis feature distribution of the detection object; Uploading information on the surface defect analysis feature distribution and the potential defect analysis feature distribution of the inspection object, so as to perform time-series integration of the surface defect analysis feature distribution and the potential defect analysis feature distribution of the inspection object through a data processing platform, and obtaining an analysis sequence consisting of the surface defect distribution features and potential defect analysis feature distributions between the current inspection object and several inspection objects of the same production batch; Analyzing the production defect factors of the production equipment of the inspection object according to the analysis sequence to obtain possible production defect factors of the production equipment producing the inspection object under the current production batch, and analyzing the usage requirements of specific inspection units of the inspection object according to the possible production defect factors to obtain specific inspection unit requirement characteristics of the inspection object corresponding to the possible production defect factors; Retrieving the usage records of specific testing units that have been executed for the test objects of the same production batch, and performing a demand realization analysis on the demand characteristics of the specific testing units based on the usage records of the specific testing units, so as to obtain the specific testing units that the test objects need to use in subsequent testing processes; wherein the specific testing units are deployed at the selected testing location; According to preset standards, the necessary detection units required for the detection object in the subsequent detection process are determined, and the subsequent detection instruments of the detection object are arranged according to the specific detection units and the necessary detection units to obtain a subsequent process detection plan; wherein, the necessary detection units are deployed at necessary detection positions.

5. The defect detection method for automobile parts according to claim 1, wherein: The steps of sequentially transporting the inspection object to a necessary inspection position and a selected inspection position through a conveying structure according to the subsequent process inspection plan, and performing defect inspection on the inspection object by inspection units pre-deployed at the necessary inspection position and the selected inspection position to obtain supplementary inspection data include: According to the subsequent process detection scheme, the detection object is transported to a necessary detection position through a conveying structure, and the detection object is detected by a necessary detection unit pre-deployed at the necessary detection position to obtain first supplementary detection data of the detection object; According to the subsequent process detection scheme, the detection object is transported to the selected detection position through the conveying structure, and the detection object is detected by the selected detection instrument pre-deployed at the selected detection position to obtain second supplementary detection data of the detection object; The first supplementary detection data and the second supplementary detection data are combined to obtain supplementary detection data of the detection object.

6. The defect detection method for automobile parts according to claim 1, wherein: Substituting the supplementary inspection data into the inspection information feedback model, performing defect inspection and evaluation of the same production batch on the inspection object to obtain an inspection evaluation label for the inspection object, and describing the inspection result of the inspection object using the inspection evaluation label includes the following steps: Substituting the supplementary detection data into the detection information feedback model, and causing the detection information feedback model to perform data content analysis on the supplementary detection data to obtain a supplementary detection feature matrix of the supplementary detection data relative to the detection information feedback model; wherein the supplementary detection feature matrix includes a plurality of supplementary detection feature vectors having a spatiotemporal distribution relationship; Analyzing the information fusion value of the supplementary detection feature matrix according to the necessary detection positions and the selected detection positions corresponding to the supplementary detection data to obtain an information fusion mode of the supplementary detection feature matrix; According to the information fusion mode of the supplementary inspection feature matrices of the same production batch, the information fusion of each supplementary inspection feature matrix of the same production batch is interactively performed to obtain comprehensive feedback information. Based on the comprehensive feedback information, the inspection object is evaluated for various defects in the preset standards to obtain the inspection evaluation label of the inspection object.

7. A defect detection device for automobile parts, characterized in that: A method for detecting defects in automotive parts according to any one of claims 1 to 6, comprising: An object marking module is used to identify basic information of the automobile parts to be inspected through a pre-deployed image recognition unit, and mark the automobile parts as inspection objects based on the basic information obtained by identification; An image detection module is configured to perform a specified clamping operation on the detection object through a pre-deployed clamping structure, and to capture an image of the detection object through the image recognition unit to obtain image detection data of the detection object; a preliminary evaluation module, configured to digitally model the inspection object based on the image inspection data to obtain an inspection information feedback model of the inspection object, and to perform defect detection evaluation of the same production batch on the inspection information feedback model using a pre-trained machine learning algorithm to obtain a subsequent process inspection plan; a supplementary inspection module, configured to sequentially transport the inspection object to a necessary inspection position and a selected inspection position via a conveying structure according to the subsequent process inspection plan, and to perform defect inspection on the inspection object by inspection units pre-deployed at the necessary inspection position and the selected inspection position to obtain supplementary inspection data; The final evaluation module is used to substitute the supplementary detection data into the detection information feedback model, perform defect detection evaluation on the detection object of the same production batch, so as to obtain the detection evaluation label of the detection object, and describe the detection result of the detection object through the detection evaluation label.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the method for detecting defects in automotive parts according to any one of claims 1 to 6 is implemented.

Images