Stamping part quality detection method and system based on computer vision

Abstract

The invention provides a stamping part quality detection method and system based on computer vision. The method comprises the following steps: extracting an appearance feature set of a stamping part from a high-definition image based on a computer vision technology; processing and analyzing the appearance feature set by using a first quality inspection model to obtain a first qualification evaluation value, and if the first qualification evaluation value is higher than a first evaluation threshold, generating and outputting quality inspection qualification conclusion information; if the first qualified evaluation value is lower than a second evaluation threshold value, quality inspection disqualification conclusion information is generated and output; if the first qualified evaluation value is lower than the first evaluation threshold but higher than the second evaluation threshold, inputting the appearance feature set into a second quality inspection model, and outputting a second qualified evaluation value; if the second qualification evaluation value is higher than a third evaluation threshold value, quality inspection qualification conclusion information is generated and output, and otherwise, quality inspection disqualification conclusion information is generated and output. According to the invention, identification of obvious defects and non-obvious defects of the stamping part is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of quality inspection, and in particular to a stamping part quality inspection method and system based on computer vision. Background Art

[0002] The quality inspection of stamping parts is a key link, which is directly related to the final quality and reliability of the product. At present, there are some solutions that use computer vision processing technology to inspect the quality of stamping parts. These existing methods basically extract the contour features of the stamping area of ​​the stamping parts, and then calculate the deviation between the extracted contour features and the designed contour features. When the deviation is greater than the deviation threshold, it is considered unqualified.

[0003] However, the above-mentioned method of quality inspection of stamping parts based on a single factor only inspects those stamping parts with obvious stamping quality problems, while those stamping parts with no obvious stamping quality problems are easily misjudged as qualified. This will result in an excessively high ratio of unqualified products in stamping parts products, making it difficult to meet actual production needs. Summary of the invention

[0004] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a stamping part quality inspection method, system, electronic equipment and computer storage medium based on computer vision.

[0005] The present invention provides a stamping parts quality inspection method based on computer vision, comprising the following steps: Obtain a high-definition image of the stamped part after stamping, and extract an appearance feature set of the stamped part from the high-definition image based on computer vision technology; wherein the appearance feature set includes contour information, depth information, color information, and texture information; Processing and analyzing the appearance feature set using a first quality inspection model to obtain a first qualified evaluation value, generating and outputting quality inspection qualified conclusion information if the first qualified evaluation value is higher than a first evaluation threshold; and generating and outputting quality inspection unqualified conclusion information if the first qualified evaluation value is lower than a second evaluation threshold; If the first qualified evaluation value is lower than a first evaluation threshold but higher than a second evaluation threshold, inputting the appearance feature set into a second quality inspection model, and the second quality inspection model outputs a second qualified evaluation value; If the second qualified evaluation value is higher than the third evaluation threshold, then generating and outputting quality inspection qualified conclusion information, otherwise generating and outputting quality inspection unqualified conclusion information; The first quality inspection model has a lower recognition capability for non-obvious defects of stamping parts than the second quality inspection model.

[0006] Optionally, the first quality inspection model is constructed based on any one of support vector machine, Gaussian naive Bayes, hidden Markov model, and random forest; the second quality inspection model is constructed based on a deep learning algorithm, and the deep learning algorithm is any one of convolutional neural network, recurrent neural network, long short-term memory network, and Transformer.

[0007] Optionally, extracting the appearance feature set of the stamping part from the high-definition image based on computer vision technology includes: Performing orientation correction processing on the high-definition image, retrieving preset stamping area distribution information, and determining each stamping area from the high-definition image according to the stamping area distribution information; Based on computer vision technology, a sub-appearance feature set is extracted from the image corresponding to the stamping area, and each of the sub-appearance feature sets is subjected to matrix transformation and matrix fusion to obtain the appearance feature set.

[0008] Optionally, the second quality inspection model is trained in the following manner: The local network node uses the deep learning algorithm to construct a basic model of the second quality inspection model, and publishes the basic model in the collaborative network; The collaborative network nodes in the collaborative network receive the basic model, and collect corresponding training data sets according to the model application scenario information in the published information, and use the training data sets to train the basic model; In the process of each of the collaborative network nodes training the basic model, the local network node publishes training strategy update information to each of the collaborative network nodes through the collaborative network, and each of the collaborative network nodes adjusts the training mode according to the training strategy update information until the basic model reaches the training termination condition, extracts the model key parameters of each of the basic models and feeds them back to the local network node; The local network node receives each of the model key parameters, imports each of the model key parameters into a target basic model, and uses a test data set to test the target basic model, and determines the target model key parameters from each of the model key parameters according to the test results; wherein the target basic model is obtained by training the local network node; The target model key parameters are imported into the target basic model to obtain the second quality inspection model.

[0009] Optionally, the local network node publishes training strategy update information to each of the collaborative network nodes through the collaborative network, including: The local network node acquires, through the collaborative network, a first number of training data types included in a training data set of each collaborative network node, and a second number of training data types that have been used for training by each collaborative network node; The severity of the training termination condition is determined according to the first quantity and the second quantity, the training strategy update information is generated according to the severity of the training termination condition, and the training strategy update information is published to each of the collaborative network nodes through the collaborative network.

[0010] Optionally, the severity is positively correlated with the first quantity, and negatively correlated with the second quantity.

[0011] The present invention also provides a stamping parts quality inspection system based on computer vision, comprising a feature extraction module, a first processing module, and a second processing module; The feature extraction module is used to obtain a high-definition image of the stamped part after stamping, and extract the appearance feature set of the stamped part from the high-definition image based on computer vision technology; wherein the appearance feature set includes contour information, depth information, color information, and texture information; The first processing module is used to perform a first processing analysis on the appearance feature set using a first quality inspection model to obtain a first qualified evaluation value, and if the first qualified evaluation value is higher than a first evaluation threshold, generate and output quality inspection qualified conclusion information; and if the first qualified evaluation value is lower than a second evaluation threshold, generate and output quality inspection unqualified conclusion information; The second processing module is configured to input the appearance feature set into a second quality inspection model if the first qualified evaluation value is lower than a first evaluation threshold but higher than a second evaluation threshold, and the second quality inspection model outputs a second qualified evaluation value; if the second qualified evaluation value is higher than a third evaluation threshold, generate and output quality inspection qualified conclusion information, otherwise generate and output quality inspection unqualified conclusion information; The first quality inspection model has a lower recognition capability for non-obvious defects of stamping parts than the second quality inspection model.

[0012] The present invention also provides an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method as described in any of the preceding items.

[0013] The present invention also provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, any of the above methods is executed.

[0014] The present invention also provides a computer program product, which includes computer code. When the computer code is executed by a processor of an electronic device, the method described in any one of the above items is implemented.

[0015] The beneficial effects of the present invention are: The scheme of the present invention sets up two quality inspection models, namely the first quality inspection model and the second quality inspection model. There is a difference in the processing capabilities of the two models. The processing capability of the first quality inspection model is weaker than that of the second quality inspection model. It uses a simple identification method to process stamping parts with obvious quality problems, while the second quality inspection model uses a more complex method to identify stamping parts without obvious quality problems. Therefore, the present invention realizes the identification of stamping parts with obvious or subtle quality problems by using two quality inspection models with different processing capabilities to cooperate with each other, thereby significantly improving the yield rate of stamping parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 It is a flow chart of a stamping parts quality inspection method based on computer vision disclosed in an embodiment of the present invention.

[0018] Figure 2 The present invention discloses a structural schematic diagram of a stamping parts quality inspection system based on computer vision. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0021] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0022] In the description of the present invention, it should be noted that if the terms "upper", "lower", "inner", "outer", etc. appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the drawings, or the orientation or position relationship in which the product of the invention is usually placed when in use. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, if the terms "first", "second", etc. appear, they are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0023] like Figure 1 As shown, a stamping part quality inspection method based on computer vision according to an embodiment of the present invention comprises the following steps: Obtain a high-definition image of the stamped part after stamping, and extract an appearance feature set of the stamped part from the high-definition image based on computer vision technology; wherein the appearance feature set includes contour information, depth information, color information, and texture information; Performing a first processing analysis on the appearance feature set using a first quality inspection model to obtain a first qualified evaluation value, generating and outputting quality inspection qualified conclusion information if the first qualified evaluation value is higher than a first evaluation threshold; and generating and outputting quality inspection unqualified conclusion information if the first qualified evaluation value is lower than a second evaluation threshold; If the first qualified evaluation value is lower than a first evaluation threshold but higher than a second evaluation threshold, inputting the appearance feature set into a second quality inspection model, and the second quality inspection model outputs a second qualified evaluation value; If the second qualified evaluation value is higher than the third evaluation threshold, then generating and outputting quality inspection qualified conclusion information, otherwise generating and outputting quality inspection unqualified conclusion information; The first quality inspection model has a lower recognition capability for non-obvious defects of stamping parts than the second quality inspection model.

[0024] The scheme of the present invention sets up two quality inspection models, namely the first quality inspection model and the second quality inspection model. The two models have different recognition capabilities for non-obvious defects of stamping parts. The recognition capability of the first quality inspection model is weaker than that of the second quality inspection model. It uses a simple recognition method to process stamping parts with obvious defects, while the second quality inspection model uses a more complex method to identify stamping parts with non-obvious defects. Therefore, the present invention realizes the recognition of stamping parts with obvious or non-obvious defects by using two quality inspection models with different processing capabilities to cooperate with each other, thereby significantly improving the yield rate of stamping parts.

[0025] Specifically, the present invention first uses a high-definition camera to shoot a high-definition image of a stamped part after stamping, and uses computer vision technology to extract the contour information, depth information, color information, and texture information of the stamped part from the high-definition image, and these information constitute an appearance feature set. First, the first quality inspection model is used to analyze each appearance element in the appearance feature set based on a simple method. If the stamped part has obvious defects (such as cracking, obvious position deviation), the first qualified evaluation value output by the first quality inspection model will be lower than the second evaluation threshold, and the quality inspection unqualified conclusion information is generated and output at this time; and when the stamped part does not have any defects, the first qualified evaluation value output by the first quality inspection model will be higher than the first evaluation threshold, and the quality inspection qualified conclusion information is generated and output at this time. However, when the stamped part has non-obvious defects (such as slight cracking, and the cracking is mainly concentrated inside or on the back of the stamped part), the first quality inspection model cannot distinguish this, and the first qualified evaluation value output by it will be between the first evaluation threshold and the second evaluation threshold. At this time, the second quality inspection model with stronger processing power is used for subsequent processing. If the second qualified evaluation value output by the second quality inspection model is higher than the third evaluation threshold (the third evaluation threshold can be the same as the first evaluation threshold, or it can be a threshold set separately for the second quality inspection model), then the quality inspection qualified conclusion information is generated and output; otherwise, the quality inspection unqualified conclusion information is generated and output.

[0026] Optionally, the first quality inspection model is constructed based on any one of support vector machine, Gaussian naive Bayes, hidden Markov model, and random forest; the second quality inspection model is constructed based on a deep learning algorithm, and the deep learning algorithm is any one of convolutional neural network, recurrent neural network, long short-term memory network, and Transformer.

[0027] In the embodiment of the present invention, support vector machine, Gaussian naive Bayes, hidden Markov model, random forest are traditional machine learning algorithms, and their architectures have been studied very thoroughly. At present, it is actually difficult to make a qualitative breakthrough in the above algorithms, resulting in a low upper limit of the processing capacity of the first quality inspection model constructed based on these algorithms. Basically, it can only identify those very obvious or frequently occurring quality defects in stampings, and it is difficult to accurately identify those non-obvious defects (mainly internal defects) or new defects that have not been trained / identified. Deep learning algorithm is an algorithm that has become increasingly popular in recent years. It is a type of algorithm inspired by the structure of the human brain. It uses multi-layer neural networks to learn complex representations of data and is particularly suitable for the identification of the above-mentioned non-obvious defects or new defects that have not been trained / identified in stampings. Of course, the processing capacity of the first quality inspection model and the second quality inspection model can also be artificially changed by training intensity (the amount of training data, training strategy), etc. Generally speaking, the classification / prediction accuracy of the model trained with a smaller amount of data will be significantly lower than that of the model trained with a larger amount of data. The difference in training strategies will be explained in the following content.

[0028] Optionally, extracting the appearance feature set of the stamping part from the high-definition image based on computer vision technology includes: Performing orientation correction processing on the high-definition image, retrieving preset stamping area distribution information, and determining each stamping area from the high-definition image according to the stamping area distribution information; Based on computer vision technology, a sub-appearance feature set is extracted from the image corresponding to the stamping area, and each of the sub-appearance feature sets is subjected to matrix transformation and matrix fusion to obtain the appearance feature set.

[0029] In an embodiment of the present invention, when the stamped parts are transferred to the quality inspection area, their orientation may be non-standard. In this case, it is necessary to first identify the key feature information, determine the current orientation of the stamped parts based on the key feature information, and then compare the current orientation with the standard orientation. If they are inconsistent, the high-definition image is adjusted at an angle to achieve orientation correction. At the same time, the stamping scheme information of each stamped part is preset, which contains detailed stamping area distribution information. The stamping area distribution information contains the various sub-areas where the punch head will exert stamping pressure on the stamping part template, and the stamping area can be determined from the high-definition image. Then, based on computer vision technology, image processing is performed on each stamping area sub-image, and a corresponding number of sub-appearance feature sets are extracted. The features contained in each sub-appearance feature set are organized into a sub-matrix, and then each sub-matrix is ​​fused to obtain an appearance feature set for use in the first quality inspection model analysis.

[0030] Among them, by first identifying the processing method of the stamping area, it is possible to avoid comprehensive computer vision processing of the entire high-definition image, thereby reducing the ineffective image processing load and improving quality inspection efficiency.

[0031] It should be noted that when the first qualified evaluation value of the first quality inspection model for the classification and identification of the sub-appearance features corresponding to each stamping area in the appearance feature set is lower than the second evaluation threshold, the quality inspection unqualified conclusion information is generated and output; and when all the first qualified evaluation values ​​are higher than the first evaluation threshold, the quality inspection qualified conclusion information is generated and output. Of course, the first quality inspection model can also process each group of sub-feature sets in the appearance feature set separately and obtain the corresponding first qualified evaluation value, and then perform threshold comparison, which is not limited by the present invention.

[0032] Optionally, the second quality inspection model is trained in the following manner: The local network node uses the deep learning algorithm to construct a basic model of the second quality inspection model, and publishes the basic model in the collaborative network; The collaborative network nodes in the collaborative network receive the basic model, and collect corresponding training data sets according to the model application scenario information in the published information, and use the training data sets to train the basic model; In the process of each of the collaborative network nodes training the basic model, the local network node publishes training strategy update information to each of the collaborative network nodes through the collaborative network, and each of the collaborative network nodes adjusts the training mode according to the training strategy update information until the basic model reaches the training termination condition, extracts the model key parameters of each of the basic models and feeds them back to the local network node; The local network node receives each of the model key parameters, imports each of the model key parameters into a target basic model, and uses a test data set to test the target basic model, and determines the target model key parameters from each of the model key parameters according to the test results; wherein the target basic model is obtained by training the local network node; The target model key parameters are imported into the target basic model to obtain the second quality inspection model.

[0033] In an embodiment of the present invention, the present invention adopts a collaborative distributed training method to train the above-mentioned second quality inspection model. Specifically, first use any of the deep learning algorithms mentioned above to construct a basic model of the second quality inspection model, and publish the basic model in the collaborative network, which includes local nodes and other collaborative network nodes. Some of the other collaborative network nodes will respond to the collaborative training information (generally based on the remaining amount of their own training resources, the rewards of collaborative training, etc.). The published information will also include model application scenario information of the basic model (i.e., the identification of non-obvious defects in stamping parts in the present invention), and these responding collaborative network nodes will collect data related to stamping parts quality inspection on their own, and form corresponding training data to train the basic model published by the local network node.

[0034] In addition to adopting the collaborative distributed training method, another improvement of the present invention is that the local network node can also publish training strategy update information to each collaborative network node so that it can adjust the training method, that is, the local network node can appropriately intervene in the training process of the collaborative network node, which is conducive to obtaining the best training effect. After each collaborative network node completes the training, the key model parameters in the basic model obtained through training are extracted and fed back to the local network node. While the collaborative network node is training, the local network node also trains the basic model based on the training data collected by itself to obtain the target basic model, imports each group of model key parameters into the target basic model one by one, and then uses the test data to test the target basic model. According to the test results, the optimal group of model key parameters can be determined, and it is imported into the target basic model to obtain the above-mentioned second quality inspection model.

[0035] The test data set is part of the training data collected by the local network node. In other words, the local network node sets part of the collected training data as the training data set and the other part as the test data set. In addition, the training data in the training data set determined by the local network node and the collaborative network node are data pairs of the appearance features of the stamping parts and the probability labels of non-obvious defects. The probability labels of non-obvious defects are marked by humans or labeling machines.

[0036] Optionally, the local network node publishes training strategy update information to each of the collaborative network nodes through the collaborative network, including: The local network node acquires, through the collaborative network, a first number of training data types included in a training data set of each collaborative network node, and a second number of training data types that have been used for training by each collaborative network node; The severity of the training termination condition is determined according to the first quantity and the second quantity, the training strategy update information is generated according to the severity of the training termination condition, and the training strategy update information is published to each of the collaborative network nodes through the collaborative network.

[0037] In an embodiment of the present invention, during the training process of each collaborative network node, the local network node can obtain two aspects of information, namely, the first number of training data types contained in the training data set of each collaborative network node, and the second number of training data types that each collaborative network node has used for training. Among them, metal stamping parts include stainless steel, cold-rolled plate, aluminum plate, hot-dip galvanized steel plate, copper plate, carbon steel plate, low titanium alloy, magnesium alloy and other types. The information released by the local network node only includes the stamping scene, and does not specifically limit the type of stamping (because the stamping machine generally needs to stamp multiple types of plates), so each collaborative network node will collect and obtain training data corresponding to different plate types based on its own experience and understanding, which leads to the difference in the first number mentioned above. When training, each collaborative network node also inputs the corresponding training data grouping one by one according to the plate type. For example, the basic model is trained using the training data of the aluminum plate first, and then the basic model is trained using the training data of the hot-dip galvanized steel plate, which leads to the difference in the second number mentioned above.

[0038] According to the above-mentioned first quantity and second quantity, the severity of the training termination condition can be determined. The severity of the training termination condition refers to the threshold of the value of the objective function in the training termination condition, that is, when the value of the objective function is lower than the threshold, it can be determined that the training has met the standard and the training can be terminated. The higher the severity, the smaller the corresponding threshold, and vice versa. The severity can be a coefficient located near 1. Using this coefficient multiplied by the initial threshold above, the adjustment of the initial threshold is achieved. Each collaborative network node adjusts its training termination condition based on the published training strategy update information.

[0039] The above-mentioned collaborative network preferably uses a blockchain network, such as a Public blockchain network, a Private blockchain network, and a Hybrid blockchain network, without specific limitation.

[0040] Optionally, the severity is positively correlated with the first quantity, and negatively correlated with the second quantity.

[0041] In the embodiment of the present invention, when the number of training data types included in the training data set is greater, the basic model needs to have more stable analysis capabilities. At this time, the severity is set higher, that is, the threshold in the training termination condition is increased; conversely, the severity is set higher, that is, the threshold in the training termination condition is appropriately lowered. At the same time, the severity is negatively correlated with the second quantity, that is, the more data types have been put into training, the higher the training completion progress, and at this time, the strict training of the basic model can be appropriately reduced to avoid failing to meet the threshold in the training termination condition; and when the number of data types have been put into training is less, the training completion progress is lower, and at this time, the strict training of the basic model needs to be maintained to avoid terminating the training too early.

[0042] The calculation formula of severity, first quantity and second quantity is, for example: ; In the formula, For strictness, is the first quantity, is the second quantity, , is the coefficient, and .For example =0.05, =0.025. The above is only an example and is not intended to limit the scope of protection of the present invention.

[0043] like Figure 2 As shown, the embodiment of the present invention also discloses a stamping part quality inspection system based on computer vision, including a feature extraction module, a first processing module, and a second processing module; The feature extraction module is used to obtain a high-definition image of the stamped part after stamping, and extract the appearance feature set of the stamped part from the high-definition image based on computer vision technology; wherein the appearance feature set includes contour information, depth information, color information, and texture information; The first processing module is used to perform a first processing analysis on the appearance feature set using a first quality inspection model to obtain a first qualified evaluation value, and if the first qualified evaluation value is higher than a first evaluation threshold, generate and output quality inspection qualified conclusion information; and if the first qualified evaluation value is lower than a second evaluation threshold, generate and output quality inspection unqualified conclusion information; The second processing module is configured to input the appearance feature set into a second quality inspection model if the first qualified evaluation value is lower than a first evaluation threshold but higher than a second evaluation threshold, and the second quality inspection model outputs a second qualified evaluation value; if the second qualified evaluation value is higher than a third evaluation threshold, generate and output quality inspection qualified conclusion information, otherwise generate and output quality inspection unqualified conclusion information; The first quality inspection model has a lower recognition capability for non-obvious defects of stamping parts than the second quality inspection model.

[0044] An embodiment of the present invention further discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method described in the above embodiment.

[0045] The embodiment of the present invention further discloses a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in the above embodiment is executed.

[0046] An embodiment of the present invention further discloses a computer program product, which includes computer code. When the computer code is executed by a processor of an electronic device, the method described in the above embodiment is implemented.

[0047] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0048] The method and related apparatus provided by the embodiment of the present application are described with reference to the method flow chart and / or structural diagram provided by the embodiment of the present application. Specifically, each process and / or box in the method flow chart and / or structural diagram, as well as the combination of the processes and / or boxes in the flow chart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the process. Figure 1 Schematic diagram of one or more processes and / or structures Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 Schematic diagram of one or more processes and / or structures Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A flow or multiple flows and / or structures illustrate the steps of the functions specified in one block or multiple blocks.

[0049] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A stamping parts quality inspection method based on computer vision, It is characterized in that The steps include: Obtain a high-definition image of the stamped part after stamping, and extract an appearance feature set of the stamped part from the high-definition image based on computer vision technology; wherein the appearance feature set includes contour information, depth information, color information, and texture information; Processing and analyzing the appearance feature set using a first quality inspection model to obtain a first qualified evaluation value, generating and outputting quality inspection qualified conclusion information if the first qualified evaluation value is higher than a first evaluation threshold; and generating and outputting quality inspection unqualified conclusion information if the first qualified evaluation value is lower than a second evaluation threshold; If the first qualified evaluation value is lower than a first evaluation threshold but higher than a second evaluation threshold, inputting the appearance feature set into a second quality inspection model, and the second quality inspection model outputs a second qualified evaluation value; If the second qualified evaluation value is higher than the third evaluation threshold, then generating and outputting quality inspection qualified conclusion information, otherwise generating and outputting quality inspection unqualified conclusion information; The first quality inspection model has a lower recognition capability for non-obvious defects of stamping parts than the second quality inspection model.

2. According to claim 1, a stamping parts quality inspection method based on computer vision, Features: The first quality inspection model is constructed based on any one of support vector machine, Gaussian naive Bayes, hidden Markov model, and random forest; the second quality inspection model is constructed based on a deep learning algorithm, and the deep learning algorithm is any one of convolutional neural network, recurrent neural network, long short-term memory network, and Transformer.

3. According to claim 2, a stamping parts quality inspection method based on computer vision, Features: The appearance feature set of the stamping part is extracted from the high-definition image based on computer vision technology, including: Performing orientation correction processing on the high-definition image, retrieving preset stamping area distribution information, and determining each stamping area from the high-definition image according to the stamping area distribution information; Based on computer vision technology, a sub-appearance feature set is extracted from the image corresponding to the stamping area, and each of the sub-appearance feature sets is subjected to matrix transformation and matrix fusion to obtain the appearance feature set.

4. According to claim 2, a stamping parts quality inspection method based on computer vision, Features: The second quality inspection model is trained in the following way: The local network node uses the deep learning algorithm to construct a basic model of the second quality inspection model, and publishes the basic model in the collaborative network; The collaborative network nodes in the collaborative network receive the basic model, and collect corresponding training data sets according to the model application scenario information in the published information, and use the training data sets to train the basic model; In the process of each of the collaborative network nodes training the basic model, the local network node publishes training strategy update information to each of the collaborative network nodes through the collaborative network, and each of the collaborative network nodes adjusts the training mode according to the training strategy update information until the basic model reaches the training termination condition, extracts the model key parameters of each of the basic models and feeds them back to the local network node; The local network node receives each of the model key parameters, imports each of the model key parameters into a target basic model, and uses a test data set to test the target basic model, and determines the target model key parameters from each of the model key parameters according to the test results; wherein the target basic model is obtained by training the local network node; The target model key parameters are imported into the target basic model to obtain the second quality inspection model.

5. According to claim 4, a stamping parts quality inspection method based on computer vision, Features: The local network node publishes training strategy update information to each of the collaborative network nodes through the collaborative network, including: The local network node acquires, through the collaborative network, a first number of training data types included in a training data set of each collaborative network node, and a second number of training data types that have been used for training by each collaborative network node; Determining the severity of the training termination condition according to the first number and the second number, and generating the training strategy update information according to the severity of the training termination condition; The training strategy update information is published to each of the collaborative network nodes through the collaborative network.

6. A stamping parts quality inspection method based on computer vision according to claim 5, Features: The severity is positively correlated with the first quantity and negatively correlated with the second quantity.

7. A stamping parts quality inspection system based on computer vision, comprising a feature extraction module, a first processing module, and a second processing module; Features: The feature extraction module is used to obtain a high-definition image of the stamped part after stamping, and extract the appearance feature set of the stamped part from the high-definition image based on computer vision technology; wherein the appearance feature set includes contour information, depth information, color information, and texture information; The first processing module is used to perform a first processing analysis on the appearance feature set using a first quality inspection model to obtain a first qualified evaluation value, and if the first qualified evaluation value is higher than a first evaluation threshold, generate and output quality inspection qualified conclusion information; and if the first qualified evaluation value is lower than a second evaluation threshold, generate and output quality inspection unqualified conclusion information; The second processing module is configured to input the appearance feature set into a second quality inspection model if the first qualified evaluation value is lower than a first evaluation threshold but higher than a second evaluation threshold, and the second quality inspection model outputs a second qualified evaluation value; if the second qualified evaluation value is higher than a third evaluation threshold, generate and output quality inspection qualified conclusion information, otherwise generate and output quality inspection unqualified conclusion information; The first quality inspection model has a lower recognition capability for non-obvious defects of stamping parts than the second quality inspection model.

8. An electronic device, include: A memory storing executable program code; A processor coupled to the memory; characterized in that: the processor calls the executable program code stored in the memory to execute the method according to any one of claims 1-6.

9. A computer storage medium having a computer program stored thereon, Features: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is executed.

10. A computer program product, Features: The computer program product includes computer codes, and when the computer codes are executed by a processor of an electronic device, the method according to any one of claims 1 to 6 is implemented.

Images