Visual detection system and detection method based on automobile mold

Abstract

The invention relates to the technical field of automobile mold detection, and particularly discloses a visual detection system and detection method based on an automobile mold, the detection system comprises an image acquisition module, a data processing module, a target analysis module and a result output module; the method specifically comprises the steps that historical standard image information of the surface of a standard automobile mold is collected through an image sensor, a three-dimensional vision model is input for construction, standard image contour data are output, historical image contour data and standard image contour data of the same batch of automobile molds are collected for digital processing, and an image matrix of the automobile molds is obtained; constructing a preset analysis model, collecting an image matrix of the current batch of automobile molds in real time to carry out factor matching analysis on a problem area, obtaining a target problem risk strategy, carrying out risk assessment on a simulation operation surface problem state of the current batch of automobile molds, and matching a corresponding adjustment strategy, and outputting a real-time processing adjustment mode corresponding to the automobile die adjustment strategy and outputting an early warning signal.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile mold detection, and in particular to a system and method for detecting automobile molds based on visual inspection. Background Art

[0002] The automotive mold visual inspection system is a tool that uses machine vision technology to inspect automotive molds. It usually uses a high-precision camera and light source to shoot the automotive mold, and then detects and evaluates the quality of the mold through image processing and algorithm analysis. This type of inspection method can achieve a full range of inspections of the mold, including the identification of structural defects, size measurement, shape recognition, etc. on the surface of the automotive mold.

[0003] However, in the existing process of visual inspection and identification of automobile molds, only the surface structure shape and other aspects are identified to improve the judgment and reduce the detection error, so as to achieve more accurate measurement of automobile molds and precise output of mold problem defects. However, the influence of factors such as the processing methods and mold structural materials of different batches of automobile molds is usually ignored, and the standards for judging structural defects are also different. In addition, the method of identifying the problem factors on the surface of a single automobile mold cannot comprehensively express the causes of mold defects caused by internal factors such as the diverse mold production materials and production methods and external environmental factors in the entire automobile mold visual inspection process; and there is a lack of analysis process for adjusting and improving the mold processing method, which makes it impossible to realize dynamic and batch automobile mold visual inspection effect prediction and processing adjustment. Summary of the invention

[0004] The purpose of the present invention is to provide a visual inspection system and method for automotive molds to solve the following technical problems:

[0005] How to use visual inspection systems to achieve batch detection of comprehensive defects in automotive molds and improve the predictive effect of improvements to mold processing methods.

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

[0007] A visual inspection system based on automobile molds, comprising:

[0008] An image acquisition module is used to acquire historical standard image information of the surface of a standard automobile mold through an image sensor, input the historical standard image information into a three-dimensional visual model to construct and output standard image contour data;

[0009] And it is used to collect historical image contour data and standard image contour data of the same batch of automobile molds for digital processing to obtain the image matrix of the automobile mold; the automobile mold surface includes the automobile mold problem area;

[0010] The data processing module is used to build a preset analysis model and analyze the problem area of ​​the automobile mold based on the image matrix to obtain the factors corresponding to the problem area of ​​the automobile mold;

[0011] And it is used to collect the image matrix of the current batch of automobile molds in real time to conduct factor matching analysis of the problem area and obtain the target problem risk strategy;

[0012] The target analysis module is used to conduct risk assessment on the simulated operation surface problem status of the current batch of automobile molds according to the target problem risk strategy; the risk assessment result is obtained through simulation training based on the preset risk assessment model, and is used to predict the risk prediction of target problems of future automobile molds and estimate the severity of existing problems;

[0013] The result output module matches the corresponding adjustment strategy according to the risk assessment results, and outputs the real-time processing adjustment method corresponding to the automobile mold adjustment strategy and outputs the early warning signal.

[0014] Preferably, the image acquisition module includes:

[0015] The solid-state image sensor of a high-definition camera is used to shoot the automobile mold, obtain the original image of the automobile mold, and pre-process the original image:

[0016] Grayscale the current original image, denoise the grayscale image through bilateral filtering, extract the edge contour of the grayscale image using the Blob algorithm, and select and determine the final image contour boundary area of ​​the grayscale image through erosion and dilation operations;

[0017] Extracting the region of interest (ROI) from the selected image contour boundary area;

[0018] Perform feature extraction on the preprocessed image to obtain the image features of the automobile mold; wherein the image features of the automobile mold include damage and missing features, convex and concave features, and dark and fine line features;

[0019] Segment the image contour boundary area through the automobile mold image features to obtain a segmented image;

[0020] Perform vector transformation on the segmented image through image processing technology to obtain image transformation vector;

[0021] Construct the image matrix of the automobile mold based on the image transformation vector.

[0022] Preferably, the data processing module:

[0023] The image matrix of the automobile mold is analyzed dimensionlessly and quantitatively by a preset analysis model to obtain the regional parameters of the current automobile mold problem; wherein the regional parameters include the regional area, regional structure and regional depth of the automobile mold problem;

[0024] Determine the level of the automobile mold problem area according to the regional parameters, and conduct multi-dimensional analysis on the level of the automobile mold problem area to obtain the factor characteristics corresponding to the automobile mold problem area, which include: internal factors and external factors;

[0025] And conduct a fusion analysis of external and internal factors to obtain the factors corresponding to the problem areas of automobile molds.

[0026] Preferably, the data processing module also includes obtaining the target problem risk strategy, and the specific process is:

[0027] Input the current batch of automobile mold image contour data into the trained analysis model and output the level of the automobile mold problem area;

[0028] According to the level of the automobile mold problem area and the factors of the current automobile mold problem area, a matching analysis is performed to obtain a list of target problem risk factors.

[0029] Preferably, the target analysis module optimizes the risk assessment model in the following manner:

[0030] Obtain a target risk factor list, and input the target risk factor list into a preset risk assessment model;

[0031] Obtain a pre-calibrated risk label in the target risk factor list; the risk label includes a first risk result corresponding to the basic information of the automobile mold and a second risk result corresponding to the current factor in the automobile mold processing information;

[0032] Compare the first risk result with the predicted first risk result, and calculate and obtain a first comparison parameter;

[0033] Comparing the second risk result with the predicted second risk result, and calculating and obtaining a second comparison parameter;

[0034] And iteratively adjust and optimize the parameters of the initial risk assessment model; obtain the minimum first comparison parameter and second comparison parameter to complete the optimization of the risk assessment model.

[0035] Preferably, the calculation process of the processing adjustment method corresponding to the adjustment strategy output of the result output module includes:

[0036] By formula Calculate the real-time processing adjustment coefficient P(t);

[0037] Among them, Δt is the preset time period for visual inspection of the current batch of automobile molds; R e (t) is the real-time automobile mold problem area level of the current batch; H e (t) is the real-time processing strength level of the problem area of ​​the current batch of automobile molds; T e (t) is the processing temperature level of the real-time automobile mold problem area of ​​the current batch; Q τ1 is the first conversion function corresponding to the level of the automobile mold problem area; Q τ2 is the second conversion function corresponding to the processing intensity level of the problem area of ​​the automobile mold; Q τ3 is the third conversion function corresponding to the processing temperature level of the problem area of ​​the automobile mold; ρ k is the correction coefficient for the visual inspection of the kth batch of automotive molds;

[0038] X is the judgment function, when When X=E, Q th (Δt) is the allowable threshold corresponding to the current time period Δt, and E is the warning value, according to which the warning signal is issued.

[0039] Preferably, it also includes:

[0040] when hour,

[0041] The real-time processing adjustment coefficient P(t) and the preset threshold P thr To compare:

[0042] If P(t)≥P thr , it is judged that it is necessary to strengthen the adjustment of the processing strength level and temperature level of the problem area of ​​the automobile mold;

[0043] If P(t) <P thr , it is judged that the processing strength level and temperature level of the problem area of ​​the automobile mold need to be weakened and adjusted.

[0044] A method for visual inspection of automobile molds is applied to a system for visual inspection of automobile molds, and includes the following steps:

[0045] Step 1: collect historical standard image information of the surface of the standard automobile mold through an image sensor, input the historical standard image information into a three-dimensional visual model to construct and output standard image contour data; and collect historical image contour data and standard image contour data of automobile molds of the same batch for digital processing to obtain an image matrix of the automobile mold;

[0046] Step 2: Build a preset analysis model, and analyze the problem area of ​​the automobile mold based on the image matrix to obtain the factors corresponding to the problem area of ​​the automobile mold; and further collect the image matrix of the current batch of automobile molds in real time to perform factor matching analysis on the problem area to obtain the target problem risk strategy;

[0047] Step 3: Conduct risk assessment on the simulated operation surface problem status of the current batch of automobile molds according to the target problem risk strategy; the result of risk assessment is obtained through simulation training based on the preset risk assessment model, which is used to predict the risk prediction of target problems of future automobile molds and estimate the severity of existing problems;

[0048] Step 4: Match the corresponding adjustment strategy according to the risk assessment results, and output the real-time processing adjustment method corresponding to the automobile mold adjustment strategy and output early warning signal.

[0049] Beneficial effects of the present invention:

[0050] (1) The present invention collects the image matrix of the current batch of automobile molds in real time through the data processing module to perform factor matching analysis on the problem area and obtain the target problem risk strategy; by performing risk matching on the problem area factors of mass-produced automobile molds, it is possible to adopt flexible risk prediction and control processing methods according to the accumulated problem changes.

[0051] (2) The target analysis module can accurately analyze the normal and abnormal areas of the automobile mold and extract the features of the abnormal area. By converting the surface features of the automobile mold into quantifiable digital indicators, it ensures a more objective reflection of the mold quality and facilitates further comparison and analysis to predict the improved production methods of the mold, thereby improving the quality of mass production of automobile molds, production and use.

[0052] (3) The present invention ensures the batch analysis process of the current automobile mold visual inspection by designing the calculation of the real-time processing adjustment coefficient, allowing the connection between the actual processing information of the current batch of automobile molds, such as the temperature in the processing environment, the processing intensity, etc., and the specific problem statistics of the surface area of ​​the current batch of molds, and compares the regional problem level or proportion data with the processing level or proportion data. By judging the adjustment direction of the processing method, the prediction of the overall visual inspection system is guaranteed to be more flexible and timely, and while accurately inspecting the current batch, the mold processing information and problem areas of the next batch can be predicted and adjusted in advance.

[0053] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0055] Figure 1 This is a module diagram of a visual inspection system for automotive molds according to the present invention;

[0056] Figure 2 This is a step diagram of a method for visual inspection of automobile molds according to the present invention;

[0057] Figure 3 This is a diagram of the optimization steps of the risk assessment model of the target analysis module of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] However, in the existing process of visual inspection and identification of automobile molds, only the surface structure shape and other aspects are identified to improve the judgment and reduce the detection error, so as to achieve more accurate measurement of automobile molds and precise output of mold problem defects. However, the influence of factors such as the processing methods and mold structural materials of different batches of automobile molds is usually ignored, and the standards for judging structural defects are also different. In addition, the method of identifying the problem factors on the surface of a single automobile mold cannot comprehensively express the causes of mold defects caused by internal factors such as the diverse mold production materials and production methods and external environmental factors in the entire automobile mold visual inspection process; and there is a lack of analysis process for adjusting and improving the mold processing method, which makes it impossible to realize dynamic and batch automobile mold visual inspection effect prediction and processing adjustment.

[0060] To solve the above technical problems, please refer to Figure 1 , 3 As shown, the present invention is a visual inspection system based on automobile molds, comprising:

[0061] An image acquisition module is used to acquire historical standard image information of the surface of a standard automobile mold through an image sensor, input the historical standard image information into a three-dimensional visual model to construct and output standard image contour data;

[0062] And it is used to collect historical image contour data and standard image contour data of the same batch of automobile molds for digital processing to obtain the image matrix of the automobile mold; the automobile mold surface includes the automobile mold problem area;

[0063] The data processing module is used to build a preset analysis model and analyze the problem area of ​​the automobile mold based on the image matrix to obtain the factors corresponding to the problem area of ​​the automobile mold;

[0064] And it is used to collect the image matrix of the current batch of automobile molds in real time to conduct factor matching analysis of the problem area and obtain the target problem risk strategy;

[0065] The target analysis module is used to conduct risk assessment on the simulated operation surface problem status of the current batch of automobile molds according to the target problem risk strategy; the risk assessment result is obtained through simulation training based on the preset risk assessment model, and is used to predict the risk prediction of target problems of future automobile molds and estimate the severity of existing problems;

[0066] The result output module matches the corresponding adjustment strategy according to the risk assessment results, and outputs the real-time processing adjustment method corresponding to the automobile mold adjustment strategy and outputs the early warning signal.

[0067] In the above technical solution, in this embodiment, the systematized processing process of the visual inspection of automobile molds is realized by designing four modules, namely, an image acquisition module, a data processing module, a target analysis module, and a result output module.

[0068] Specifically, firstly, this embodiment sets an image acquisition module to collect historical standard image information of the surface of a standard automobile mold through an image sensor, inputs the historical standard image information into a three-dimensional visual model to construct and output standard image contour data; and is used to collect historical image contour data and standard image contour data of automobile molds of the same batch for digital processing to obtain an image matrix of the automobile mold; wherein the automobile mold surface includes the automobile mold problem area.

[0069] In order to better achieve the above effects mentioned in the image acquisition module, specifically, the automobile mold is photographed by using a solid-state image sensor of a high-definition camera. In order to make the photographed automobile mold have clearer surface structural features, a 3D camera is used to assist the image sensor to identify the mold surface in the current posture and correct it through a three-dimensional algorithm to ensure the accuracy of the final detection result. The three-dimensional algorithm is implemented based on a three-dimensional visual model; because the 3D camera detection system is different from the 2D camera detection system, the 3D camera is specifically used to obtain the corresponding three-dimensional coordinates of the appearance of the automobile mold, and the dimensions of the points, lines, surfaces and other features on the key parts are calculated according to the three-dimensional coordinates, and finally the image contour data is obtained; because the 3D camera detection system does not have high requirements for the placement of the workpiece, and can detect multiple automobile mold surfaces at one time.

[0070] In addition, by obtaining the original image of the automobile mold, the original image needs to be preprocessed to improve the accuracy of subsequent analysis; the specific preprocessing includes denoising the grayscale image through bilateral filtering, and using the Blob algorithm to extract the edge contour of the grayscale image, and then using corrosion and expansion operations to select and determine the final image contour boundary area of ​​the grayscale image; and extracting the region of interest ROI from the selected image contour boundary area.

[0071] Furthermore, the features of the automobile mold are obtained through image processing technology, specifically, the features of the automobile mold image are obtained by extracting features of the preprocessed image; wherein the features of the automobile mold image include damage and missing features, convex and concave features, and dark lines and fine marks; the image contour boundary area is segmented through the automobile mold image features to obtain a segmented image; thereby, the division of the automobile mold image area is realized, especially the analysis of problem areas, such as surface structure damage, scratches, missing and structural redundancy.

[0072] After segmentation through image processing technology, the image is also converted into a digital vector to obtain the image conversion vector, and the image matrix of the automobile mold is constructed based on the image conversion vector; by converting the automobile mold image into a digital image matrix, various image processing, machine learning and data mining technologies can be used for in-depth analysis to provide more comprehensive problem detection and analysis results for automobile processing molds.

[0073] Secondly, this embodiment sets a data processing module, constructs a preset analysis model, and analyzes the problem area of ​​the automobile mold based on the image matrix to obtain factors corresponding to the problem area of ​​the automobile mold.

[0074] In order to better achieve the above-mentioned effects mentioned in the image acquisition module, specifically, the image matrix of the automobile mold is dimensionlessly quantified and analyzed through a preset analysis model to obtain the regional parameters of the current automobile mold problem; the image matrix is ​​subjected to data analysis through a preset analysis model to distinguish the normal area and the problem area of ​​the mold surface, and the current image matrix obtains the characteristics of the automobile mold problem by extracting characteristic factors from the obtained problem area. This type of image segmentation will rely on existing technologies such as cluster analysis, automatic threshold method, Blob analysis, edge detection, etc.; feature extraction usually relies on image processing and computer vision algorithms, such as grayscale co-occurrence matrix, bilateral filtering, dilation operation, etc.

[0075] By performing dimensionless quantification processing on the extracted features of automobile molds, numerical factors that can reflect the current real-time status of the automobile mold problem area are obtained, such as the area of ​​the current automobile mold problem area, the size of the regional structure, and the regional depth value; and by determining the level of the automobile mold problem area according to the regional parameters, and performing multi-dimensional analysis on the level of the automobile mold problem area, more and more comprehensive problem feature types can be obtained.

[0076] In this embodiment, the factor matching analysis of the problem area is further performed by real-time acquisition of the image matrix of the current batch of automobile molds to obtain the target problem risk strategy; by performing risk matching of the problem area factors of the mass-produced automobile molds, it is convenient to adopt flexible risk prediction and control processing methods according to the cumulative problem changes. The specific process of factor matching of the problem area is: first, the image contour data of the current batch of automobile molds is input into the trained analysis model, and the level of the automobile mold problem area is output; then, matching analysis is performed based on the level of the automobile mold problem area and the factors of the current automobile mold problem area, and a list of target problem risk factors is obtained by matching.

[0077] In the above-mentioned further implementation of the technical solution, the basis for the level of the automobile mold problem area in the trained analysis model that relies on the method of matching the problem area factors is: through multi-dimensional analysis of actual related factors, mainly based on the regional area, regional structure size and regional depth value to establish a regional standard threshold system, the automobile mold problems are divided into levels, such as level one slight damage (slight scratches, burr marks, etc.), level two moderate damage (obvious scratches on the surface, obvious marks, etc.), level three severe damage (structural loss, deformation, size discrepancy, etc.); for automobile mold problem areas of different levels, combined with production operations and processing experience, analyze the different factors that may cause, according to the factors affecting the quality of automobile molds, the factor characteristics corresponding to the automobile mold problem area will be obtained, and the factor characteristic types will be analyzed, mainly including: internal factors and external factors; and the external factors and internal factors are fused and analyzed, and the correlation analysis is combined with different factors using statistical methods and machine algorithms; the factors corresponding to the automobile mold problem area are analyzed.

[0078] The above technical solution can realize accurate analysis of normal and abnormal areas of automobile molds and extract features of abnormal areas. By converting the surface features of automobile molds into quantifiable digital indicators, it is ensured that the quality of molds is reflected more objectively, and it is convenient for further comparison and analysis to predict the improved production methods of molds, thereby improving the quality of mass production of automobile molds, production and use.

[0079] Thirdly, this embodiment also provides a target analysis module for performing risk assessment on the simulated operation surface problem status of the current batch of automobile molds according to the target problem risk strategy; the result of the risk assessment is obtained through simulation training based on a preset risk assessment model, which is used to predict the risk of target problems of future automobile molds and estimate the severity of existing problems.

[0080] In the specific implementation, please refer to Figure 2 As shown, the target analysis module optimizes the risk assessment model in the following manner: first, obtain a list of target risk factors, and input the list of target risk factors into a preset risk assessment model; based on the basic information of the actual processing of the current batch of automobile molds, including the size, specifications, shape, material, physical strength, previous processing surface problems, processing environment conditions, and basic parameters of current automobile mold accessories and other historical information; the multi-source data sets collected above are combined as basic data and input into a preset risk assessment model for pre-training. The preset risk assessment model is constructed based on a conventional convolutional neural network model.

[0081] Then, the output of the target risk factor related parameters is obtained, and the key visual inspection targets of the current batch of automobile molds are obtained, based on the risk labels pre-calibrated in the target risk factor list; the risk labels include the first risk result corresponding to the basic information of the automobile mold and the second risk result corresponding to the current factor in the automobile mold processing information. The basic information data of the automobile mold is input through the initial basic information model to generate the first risk result, and then the current result is used as the input of the factor model to finally output the second risk result; then the first risk result is compared with the predicted first risk result, and the first comparison parameter is calculated; the second risk result is compared with the predicted second risk result, and the second comparison parameter is calculated; and the parameters of the initial risk assessment model are iteratively adjusted and optimized;

[0082] Finally, the minimum first comparison parameter and the second comparison parameter are obtained to complete the optimization process of the risk assessment model. In the specific calculation process, the optimization training process of the risk assessment model is completed through repeated iterative training until the loss of the model on the training set reaches the optimal value or converges to a predetermined threshold.

[0083] Thirdly, the present embodiment further provides a result output module, which matches the corresponding adjustment strategy according to the risk assessment result, and outputs the real-time processing adjustment method corresponding to the automobile mold adjustment strategy and outputs an early warning signal.

[0084] In a specific implementation, the calculation process of the processing adjustment method corresponding to the adjustment strategy output of the result output module includes:

[0085] By formula Calculate the real-time processing adjustment coefficient P(t);

[0086] Among them, Δt is the preset time period for visual inspection of the current batch of automobile molds; R e (t) is the real-time automobile mold problem area level of the current batch; H e (t) is the real-time processing strength level of the problem area of ​​the current batch of automobile molds; T e (t) is the processing temperature level of the real-time automobile mold problem area of ​​the current batch; Q τ1 is the first conversion function corresponding to the level of the automobile mold problem area; Q τ2 is the second conversion function corresponding to the processing intensity level of the problem area of ​​the automobile mold; Q τ3 is the third conversion function corresponding to the processing temperature level of the problem area of ​​the automobile mold; ρ k is the correction coefficient for the visual inspection of the kth batch of automotive molds;

[0087] X is the judgment function, when When X=E, Q th (Δt) is the allowable threshold corresponding to the current time period Δt, and E is the warning value, according to which the warning signal is issued.

[0088] In the above technical scheme, the batch analysis process of the current automobile mold visual inspection is guaranteed by calculating the real-time processing adjustment coefficient, allowing the actual processing information of the current batch of automobile molds, such as the temperature in the processing environment, the processing intensity, etc., and the specific problem statistics of the surface area of ​​the current batch of molds to be linked, and compared according to the regional problem level or proportion data and the processing level or proportion data. By judging the adjustment direction of the processing method, the overall visual inspection system is guaranteed to be more flexible and timely, and the mold processing information and problem areas of the next batch can be predicted and adjusted in advance while accurately detecting the current batch.

[0089] It needs further explanation that the first conversion function Q in the formula τ1 , the second conversion function Q τ2 And the third conversion function Q τ2 The data are obtained by fitting the test data under the normal operation state of the visual inspection process of the automobile mold of this batch. Therefore, the real-time automobile mold problem area level R of the current batch can be obtained by bringing in the influencing conditions. e (t), the real-time processing strength level H of the problem area of ​​the current batch of automobile molds e (t), the real-time processing temperature level T of the problem area of ​​the current batch of automobile molds e (t) is calculated and the judgment function is used to determine whether there is a warning of processing problems in the current automotive mold visual inspection process; to ensure early prediction. In addition, the preset time period Δt is selected and set according to empirical data, and can be adaptively adjusted according to the specific automotive mold visual inspection batch information; the correction coefficient ρ of the kth batch of automotive mold visual inspection k When data calculations and formula expressions deviate between ideal and reality, reality and investigation, etc., the coefficients are added to the calculation formulas to make them reflect the real performance as much as possible. Specifically, this design is mainly based on the deviation in the duration of visual inspection work consumed by different batches of automotive molds in reality, the deviation in visual inspection position, the proportion of the current batch in the entire visual inspection work, and the overall state deviation of the inspection sequence. It has the effect of adjusting the data output results, which will not be elaborated here.

[0090] Specifically, it also includes:

[0091] when hour,

[0092] The real-time processing adjustment coefficient P(t) and the preset threshold P thr To compare:

[0093] If P(t)≥P thr , it is judged that it is necessary to strengthen the adjustment of the processing strength level and temperature level of the problem area of ​​the automobile mold;

[0094] If P(t)<P thr , it is judged that the processing strength level and temperature level of the problem area of ​​the automobile mold need to be weakened and adjusted.

[0095] The above is an analysis of a specific embodiment of the present invention based on a visual inspection system for automotive molds. Figure 3 As shown, a method for visual inspection of automobile molds based on the system configuration includes the following steps:

[0096] Step 1: collect historical standard image information of the surface of the standard automobile mold through an image sensor, input the historical standard image information into a three-dimensional visual model to construct and output standard image contour data; and collect historical image contour data and standard image contour data of automobile molds of the same batch for digital processing to obtain an image matrix of the automobile mold;

[0097] Step 2: Build a preset analysis model, and analyze the problem area of ​​the automobile mold based on the image matrix to obtain the factors corresponding to the problem area of ​​the automobile mold; and further collect the image matrix of the current batch of automobile molds in real time to perform factor matching analysis on the problem area to obtain the target problem risk strategy;

[0098] Step 3: Conduct risk assessment on the simulated operation surface problem status of the current batch of automobile molds according to the target problem risk strategy; the result of risk assessment is obtained through simulation training based on the preset risk assessment model, which is used to predict the risk prediction of target problems of future automobile molds and estimate the severity of existing problems;

[0099] Step 4: Match the corresponding adjustment strategy according to the risk assessment results, and output the real-time processing adjustment method corresponding to the automobile mold adjustment strategy and output early warning signal.

[0100] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined in this application, they shall all fall within the protection scope of the present invention.

Claims

1. A visual inspection system based on automobile molds, characterized in that: include: An image acquisition module is used to acquire historical standard image information of the surface of a standard automobile mold through an image sensor, input the historical standard image information into a three-dimensional visual model to construct and output standard image contour data; and used to collect historical image contour data and standard image contour data of the same batch of automobile molds for digital processing to obtain an image matrix of the automobile mold; the automobile mold surface includes the automobile mold problem area; The data processing module is used to build a preset analysis model and analyze the problem area of ​​the automobile mold based on the image matrix to obtain the factors corresponding to the problem area of ​​the automobile mold; And it is used to collect the image matrix of the current batch of automobile molds in real time to conduct factor matching analysis of the problem area and obtain the target problem risk strategy; The target analysis module is used to conduct risk assessment on the simulated operation surface problem status of the current batch of automobile molds according to the target problem risk strategy; The results of the risk assessment are obtained through simulation training based on a preset risk assessment model, and are used to predict the risk of target problems of future automotive molds and estimate the severity of existing problems; The result output module matches the corresponding adjustment strategy according to the risk assessment results, and outputs the real-time processing adjustment method corresponding to the automobile mold adjustment strategy and outputs the early warning signal.

2. According to claim 1, the automotive mold visual inspection system is characterized in that: The image acquisition module comprises: The solid-state image sensor of a high-definition camera is used to shoot the automobile mold, obtain the original image of the automobile mold, and pre-process the original image: Grayscale the current original image, denoise the grayscale image through bilateral filtering, extract the edge contour of the grayscale image using the Blob algorithm, and select and determine the final image contour boundary area of ​​the grayscale image through erosion and dilation operations; Extracting the region of interest (ROI) from the selected image contour boundary area; Extracting features from the preprocessed image to obtain features of the automobile mold image; wherein the features of the automobile mold image include features of damage and missing, features of protrusions and depressions, and features of dark lines and fine marks; Segmenting the image contour boundary area according to the automobile mold image features to obtain a segmented image; Performing vector conversion on the segmented image by using image processing technology to obtain an image conversion vector; An image matrix of the automobile mold is constructed based on the image conversion vector.

3. The automotive mold visual inspection system according to claim 1 is characterized in that: The data processing module: Perform dimensionless quantitative analysis on the image matrix of the automobile mold through a preset analysis model to obtain regional parameters of the current automobile mold problem; wherein the regional parameters include the regional area, regional structure and regional depth of the automobile mold problem; Determine the level of the automobile mold problem area according to the area parameters, and perform multi-dimensional analysis on the level of the automobile mold problem area to obtain factor characteristics corresponding to the automobile mold problem area, wherein the factor characteristics include: internal factors and external factors; The external factors and internal factors are integrated and analyzed to obtain the factors corresponding to the problem areas of automobile molds.

4. The automotive mold visual inspection system according to claim 3 is characterized in that: The data processing module also includes obtaining the target problem risk strategy, and the specific process is as follows: Input the current batch of automobile mold image contour data into the trained analysis model and output the level of the automobile mold problem area; A matching analysis is performed based on the level of the automobile mold problem area and the factors of the current automobile mold problem area, and a list of target problem risk factors is obtained through matching.

5. The automotive mold visual inspection system according to claim 4 is characterized in that: The target analysis module optimizes the risk assessment model in the following way: Obtaining a target risk factor list, and inputting the target risk factor list into a preset risk assessment model; Obtain pre-calibrated risk labels in the target risk factor list; The risk tag includes a first risk result corresponding to the basic information of the automobile mold and a second risk result corresponding to the current factor in the automobile mold processing information; Comparing the first risk result with the predicted first risk result, and calculating and obtaining a first comparison parameter; Comparing the second risk result with the predicted second risk result, and calculating and obtaining a second comparison parameter; And iteratively adjust and optimize the parameters of the initial risk assessment model; obtain the minimum first comparison parameter and second comparison parameter to complete the optimization of the risk assessment model.

6. The automotive mold visual inspection system according to claim 1 is characterized in that: The calculation process of the processing adjustment method corresponding to the adjustment strategy output of the result output module includes: By formula Calculate the real-time processing adjustment coefficient P(t); Among them, Δt is the preset time period for visual inspection of the current batch of automobile molds; R e (t) is the real-time automobile mold problem area level of the current batch; H e (t) is the real-time processing strength level of the problem area of ​​the current batch of automobile molds; T e (t) is the processing temperature level of the real-time automobile mold problem area of ​​the current batch; Q τ1 is the first conversion function corresponding to the level of the automobile mold problem area; Q τ2 is the second conversion function corresponding to the processing intensity level of the problem area of ​​the automobile mold; Q τ3 is the third conversion function corresponding to the processing temperature level of the problem area of ​​the automobile mold; ρ k is the correction coefficient for the visual inspection of the kth batch of automotive molds; X is the judgment function, when When X=E, Q th (Δt) is the allowable threshold corresponding to the current time period Δt, and E is the warning value, according to which the warning signal is issued.

7. The automotive mold visual inspection system according to claim 6 is characterized in that: Also includes: when hour, The real-time processing adjustment coefficient P(t) and the preset threshold P thr To compare: If P(t)≥P thr , it is judged that it is necessary to strengthen the adjustment of the processing strength level and temperature level of the problem area of ​​the automobile mold; If P(t)<P thr , it is judged that the processing strength level and temperature level of the problem area of ​​the automobile mold need to be weakened and adjusted.

8. A method for visual inspection of automobile molds, characterized in that: The automotive mold visual inspection system according to any one of claims 1 to 7 comprises the following steps: Step 1: collect historical standard image information of the surface of the standard automobile mold through an image sensor, input the historical standard image information into a three-dimensional visual model to construct and output standard image contour data; and collect historical image contour data and standard image contour data of automobile molds of the same batch for digital processing to obtain an image matrix of the automobile mold; Step 2: Build a preset analysis model, and analyze the problem area of ​​the automobile mold based on the image matrix to obtain the factors corresponding to the problem area of ​​the automobile mold; and further collect the image matrix of the current batch of automobile molds in real time to perform factor matching analysis on the problem area to obtain the target problem risk strategy; Step 3: Perform risk assessment on the simulated operation surface problem status of the current batch of automobile molds according to the target problem risk strategy; the result of the risk assessment is obtained by simulation training based on a preset risk assessment model, and is used to predict the risk prediction of target problems of future automobile molds and estimate the severity of existing problems; Step 4: Match the corresponding adjustment strategy according to the risk assessment results, and output the real-time processing adjustment method corresponding to the automobile mold adjustment strategy and output early warning signal.

Images