Washing machine valve production monitoring management system and method

Through machine vision technology based on deep learning, automated quality inspection of washing machine valves is solved, and the instability and inefficiency of traditional monitoring methods are achieved, and efficient quality monitoring and production improvement are achieved.

CN118658110BActive Publication Date: 2024-11-22ZHEJIANG JINSHUO VALVE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410654026.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-24
Publication Date
2024-11-22
Estimated Expiration
2044-05-24

AI Technical Summary

Technical Problem

The quality monitoring of valves in traditional washing machines relies on manual inspection and simple mechanical inspection, which is susceptible to work experience and skill level, resulting in unstable and inaccurate inspection results, making it difficult to meet large-scale and efficient production needs.

Method used

The digital images of washing machine valves are analyzed using machine vision technology based on deep learning. Through depth and shallow feature extraction, multi-scale semantic enhancement and semantic guidance fusion, we can intelligently identify whether the quality of the valve is qualified and realize automated quality inspection.

Benefits of technology

It improves the quality monitoring efficiency of washing machine valves, reduces manual intervention, and improves overall production efficiency and product quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118658110B_ABST
    Figure CN118658110B_ABST
Patent Text Reader

Abstract

The present application relates to the field of industrial data analysis technology, and specifically discloses a washing machine valve production monitoring management system and method, which uses machine vision technology based on deep learning to analyze and process the digital image of the washing machine valve to capture the multi-level semantic features of the valve digital image, and mine the deep information of the surface state of the washing machine valve, so as to intelligently identify whether the valve quality is qualified. In this way, automated valve quality inspection can be achieved to improve the quality monitoring efficiency of the washing machine valve, reduce manual intervention, and thus improve the overall production efficiency and product quality of the washing machine valve.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of industrial data analysis, and more specifically, to a washing machine valve production monitoring management system and method. Background Art

[0002] With the rapid development of modern manufacturing industry and the increasing demand of consumers for product quality, the quality of washing machine valves, as an important part of washing machines, is directly related to the performance and service life of the entire washing machine. Therefore, in the production process of washing machine valves, the monitoring of valve quality is an important link to ensure product quality and stability.

[0003] Traditional washing machine valve quality monitoring mainly relies on manual inspection and simple mechanical inspection methods. However, such inspection methods are easily restricted by subjective factors such as the work experience and skill level of the inspectors, resulting in instability and inaccuracy of the inspection results, and the inspection efficiency is low, which makes it difficult to meet the needs of large-scale and high-efficiency production. Therefore, an optimized washing machine valve production monitoring management system and method are expected. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a washing machine valve production monitoring management system and method, which uses machine vision technology based on deep learning to analyze and process the digital image of the washing machine valve to capture the multi-level semantic features of the valve digital image, and mine the deep information of the surface state of the washing machine valve, so as to intelligently identify whether the quality of the valve is qualified. In this way, automated valve quality inspection can be achieved to improve the quality monitoring efficiency of the washing machine valve, reduce manual intervention, and thus improve the overall production efficiency and product quality of the washing machine valve.

[0005] Accordingly, according to one aspect of the present application, a washing machine valve production monitoring and management system is provided, which includes:

[0006] A washing machine valve image acquisition module, used to acquire a valve digital image of the washing machine valve collected by an industrial camera;

[0007] A depth feature extraction module, used for performing depth feature extraction on the valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state;

[0008] A multi-scale semantic enhancement module, used for performing multi-scale semantic enhancement on the valve surface state semantic feature map to obtain a multi-scale semantic representation vector of the valve surface state;

[0009] A semantically guided enhanced fusion module, used for obtaining a valve surface state semantically guided valve state enhanced fusion feature map by passing the multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state through a semantically guided attention feature selection network;

[0010] The washing machine valve quality diagnosis module is used to guide the valve state reinforcement fusion feature map based on the valve surface state semantics to determine the quality inspection result of the washing machine valve.

[0011] According to another aspect of the present application, a method for monitoring and managing the production of washing machine valves is provided, comprising:

[0012] Acquire a valve digital image of a washing machine valve captured by an industrial camera;

[0013] Extracting shallow and deep features from the valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state;

[0014] Performing multi-scale semantic enhancement on the valve surface state semantic feature map to obtain a multi-scale semantic representation vector of the valve surface state;

[0015] The multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state are passed through a semantically guided attention feature selection network to obtain a valve surface state semantically guided valve state enhanced fusion feature map;

[0016] The quality inspection result of the washing machine valve is determined based on the valve surface state semantics-guided valve state enhanced fusion feature map.

[0017] Compared with the prior art, the washing machine valve production monitoring management system and method provided by the present application uses machine vision technology based on deep learning to analyze and process the digital image of the washing machine valve to capture the multi-level semantic features of the valve digital image and mine the deep information of the surface state of the washing machine valve, thereby intelligently identifying whether the valve quality is qualified. In this way, automated valve quality inspection can be achieved to improve the quality monitoring efficiency of the washing machine valve, reduce manual intervention, and thus improve the overall production efficiency and product quality of the washing machine valve. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 14 is a block diagram of a washing machine valve production monitoring and management system according to an embodiment of the present application.

[0020] Figure 2 Schematic diagram of the architecture of a washing machine valve production monitoring and management system according to an embodiment of the present application.

[0021] Figure 3 4 is a block diagram of a depth feature extraction module in a washing machine valve production monitoring and management system according to an embodiment of the present application.

[0022] Figure 4 4 is a block diagram of a multi-scale semantic enhancement module in a washing machine valve production monitoring and management system according to an embodiment of the present application.

[0023] Figure 5 It is a block diagram of the semantic-guided enhanced fusion module in the washing machine valve production monitoring and management system according to an embodiment of the present application.

[0024] Figure 6 4 is a block diagram of a training module in a washing machine valve production monitoring and management system according to an embodiment of the present application.

[0025] Figure 7 Flow chart of a method for monitoring and managing production of washing machine valves according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0027] As mentioned in the above background technology, traditional washing machine valve quality monitoring mainly relies on manual inspection and simple mechanical inspection methods. However, this type of detection method is easily restricted by subjective factors such as the work experience and skill level of the inspectors, which leads to instability and inaccuracy of the detection results, and the detection efficiency is low, which makes it difficult to meet the needs of large-scale and high-efficiency production. In response to the above technical problems, the technical concept of the present application is to use machine vision technology based on deep learning to analyze and process the digital image of the washing machine valve, so as to capture the multi-level semantic features of the valve digital image, and mine the deep information of the surface state of the washing machine valve, so as to intelligently identify whether the quality of the valve is qualified. In this way, automated valve quality inspection can be achieved to improve the quality monitoring efficiency of the washing machine valve, reduce manual intervention, and thereby improve the overall production efficiency and product quality of the washing machine valve.

[0028] Figure 1 4 is a block diagram of a washing machine valve production monitoring and management system according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of a washing machine valve production monitoring and management system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the washing machine valve production monitoring and management system 100 includes: a washing machine valve image acquisition module 110, which is used to acquire a valve digital image of a washing machine valve collected by an industrial camera; a depth feature extraction module 120, which is used to perform depth feature extraction on the valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state; a multi-scale semantic enhancement module 130, which is used to perform multi-scale semantic enhancement on the valve surface state semantic feature map to obtain a multi-scale semantic representation vector of the valve surface state; a semantic-guided enhancement fusion module 140, which is used to pass the multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state through a semantic-guided attention feature selection network to obtain a valve surface state semantic-guided valve state enhancement fusion feature map; a washing machine valve quality diagnosis module 150, which is used to determine the quality inspection result of the washing machine valve based on the valve surface state semantic-guided valve state enhancement fusion feature map.

[0029] In the above-mentioned washing machine valve production monitoring and management system 100, the washing machine valve image acquisition module 110 is used to acquire the valve digital image of the washing machine valve collected by the industrial camera. It should be understood that the industrial camera has higher image quality and stability than the traditional camera. By using the industrial camera to collect the valve digital image of the washing machine valve, the subtle features and textures on the surface of the washing machine valve can be captured, and non-contact quality monitoring can be achieved to reduce the labor intensity of the workers. Based on this, in the technical solution of the present application, the physical object (i.e., the washing machine valve) is converted into a digital form by an industrial camera, so that it can be analyzed and processed using machine vision technology, thereby realizing automated washing machine valve quality inspection.

[0030] In the above-mentioned washing machine valve production monitoring and management system 100, the depth feature extraction module 120 is used to extract depth features from the valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state. Figure 3 FIG. 1 is a block diagram of a depth feature extraction module in a washing machine valve production monitoring and management system according to an embodiment of the present application. Figure 3 As shown, the depth feature extraction module 120 includes: an image detail enhancement processing unit 121, which is used to perform contrast normalization and generalized Gaussian fitting on the valve digital image to obtain a detail enhanced valve digital image; an image multi-scale feature extraction unit 122, which is used to pass the detail enhanced valve digital image through a valve image multi-scale feature extractor based on a pyramid network to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state.

[0031] Specifically, the image detail enhancement processing unit 121 is used to perform contrast normalization and generalized Gaussian fitting on the valve digital image to obtain a detail enhanced valve digital image. It should be understood that, considering the influence of factors such as the shooting environment, light conditions, and valve surface material during the production process of the washing machine valve, the valve digital image may have problems such as inconsistent contrast and uneven brightness, which in turn interferes with subsequent image analysis and feature extraction. Therefore, in the technical solution of the present application, the valve digital image is further subjected to contrast normalization and generalized Gaussian fitting processing to eliminate the contrast and brightness differences in the image, enhance the image details, and make the feature information on the surface of the washing machine valve clearer and more prominent. Specifically, contrast normalization is an image enhancement technology that can map the pixel values ​​in the valve digital image to a new range, and improve the image contrast by adjusting the range of the image pixel values, so that the brightness distribution in different areas is more uniform, thereby improving the visual quality of the image. Generalized Gaussian fitting is a nonlinear fitting method that uses generalized Gaussian distribution to model the distribution of image pixel values. By adjusting the fitting parameters, different areas in the image are enhanced to varying degrees, thereby making the detail information in the valve digital image more prominent, providing a more accurate and reliable data basis for subsequent image feature extraction and valve quality analysis.

[0032] Specifically, the image multi-scale feature extraction unit 122 is used to pass the detail-enhanced valve digital image through a valve image multi-scale feature extractor based on a pyramid network to obtain the shallow feature map of the valve surface state and the semantic feature map of the valve surface state. It should be understood that considering that the defects or quality problems that may exist on the surface of the washing machine valve usually have different scales and complexities, such as local surface scratches and overall contour deformation. Therefore, in order to more comprehensively capture the state information of the valve surface, in the technical solution of the present application, a valve image multi-scale feature extractor based on a pyramid network is used to perform feature mining of different scales on the detail-enhanced valve digital image. Specifically, the pyramid network is a deep network structure with multi-scale feature extraction capabilities. It constructs feature pyramids of different scales by using convolutional layers and pooling layers of different scales, and can simultaneously capture shallow local detail information and global semantic information in the detail-enhanced valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state. Among them, the shallow feature map of the valve surface state mainly reflects the local detail information of the valve surface, such as texture, color, edge, corner point, etc.; while the semantic feature map of the valve surface state focuses more on the global structure and semantic information of the valve surface, such as overall shape, contour, etc.

[0033] In the above-mentioned washing machine valve production monitoring and management system 100, the multi-scale semantic enhancement module 130 is used to perform multi-scale semantic enhancement on the valve surface state semantic feature map to obtain a multi-scale semantic representation vector of the valve surface state. In a specific example of the present application, the encoding method for multi-scale semantic enhancement of the valve surface state semantic feature map is to pass the valve surface state semantic feature map through a semantic distiller based on a multi-scale semantic feature extraction network to obtain a multi-scale semantic representation vector of the valve surface state. That is, in order to further enhance the semantic expression of the valve surface state semantic feature map, a multi-scale semantic feature extraction network is used to further perform semantic feature distillation enhancement on the valve surface state semantic feature map. Specifically, the multi-scale semantic feature extraction network can effectively capture the global semantic features and local semantic features of the valve surface state semantic feature map through a variety of convolution and pooling structures, and comprehensively represent the deep-level semantic features of the valve surface state by fusing its global semantic features and local semantic features, thereby improving the understanding of the valve surface state and providing a richer and more accurate information basis for subsequent valve quality identification.

[0034] Figure 4 FIG. 1 is a block diagram of a multi-scale semantic enhancement module in a washing machine valve production monitoring and management system according to an embodiment of the present application. Figure 4 As shown, the multi-scale semantic enhancement module 130 includes: a multi-scale semantic encoding unit 131, which is used to respectively extract the global semantic features and the local semantic features of the valve surface state semantic feature map to obtain a global semantic feature vector of the valve surface state and a local semantic feature vector of the valve surface state; and a multi-scale semantic fusion unit 132, which is used to fuse the global semantic feature vector of the valve surface state and the local semantic feature vector of the valve surface state to obtain a multi-scale semantic representation vector of the valve surface state.

[0035] More specifically, the multi-scale semantic encoding unit 131 includes: an upsampling subunit, used to upsample the valve surface state semantic feature map to obtain an upsampled valve surface state semantic feature map; a global mean pooling subunit, used to perform global mean pooling processing on the upsampled valve surface state semantic feature map to obtain an upsampled valve surface state semantic channel feature vector; a semantic channel point convolution encoding subunit, used to perform point convolution encoding on the upsampled valve surface state semantic channel feature vector to obtain the valve surface state global semantic feature vector; a channel modulation subunit, used to perform point convolution encoding on the upsampled valve surface state semantic feature map to obtain a channel-modulated upsampled valve surface state semantic feature map; and a local semantic feature extraction subunit, used to perform two-dimensional convolution encoding on the channel-modulated upsampled valve surface state semantic feature map to obtain the valve surface state local semantic feature vector.

[0036] In the above-mentioned washing machine valve production monitoring and management system 100, the semantic guided enhanced fusion module 140 is used to pass the multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state through the semantic guided attention feature selection network to obtain the valve surface state semantic guided valve state enhanced fusion feature map. It should be understood that the multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state respectively represent different levels of features of the valve surface state. By fusing the two, the understanding and recognition accuracy of the valve surface state can be further improved. However, traditional fusion methods such as feature splicing often cannot effectively utilize the complementarity between the two features, resulting in the fused features may contain redundant information and noise. Based on this, in the technical solution of the present application, a semantic guided attention feature selection network is introduced to realize the feature fusion task of the multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state. Specifically, the semantic guided attention feature selection network uses the global semantic information in the multi-scale semantic representation vector of the valve surface state as the guiding information, guides the feature selection of the shallow feature map of the valve surface state through the attention mechanism, and guides the position information in the shallow feature map of the valve surface state to restore the category. In this way, the multi-scale semantic representation vector of the valve surface state and the enhanced shallow feature map of the valve surface state are feature fused, so as to fully utilize the complementarity between the global semantic features and the shallow detail features, and provide a more accurate and reliable feature basis for the subsequent valve quality analysis.

[0037] Figure 5 FIG. 1 is a block diagram of a semantically guided enhanced fusion module in a washing machine valve production monitoring and management system according to an embodiment of the present application. Figure 5 As shown, the semantically guided enhancement fusion module 140 includes: a semantically guided feature enhancement unit 141, which is used to perform feature enhancement processing on the shallow feature map of the valve surface state based on the multi-scale semantic representation vector of the valve surface state to obtain an enhanced shallow feature map of the valve surface state; and a fusion unit 142, which is used to fuse the enhanced shallow feature map of the valve surface state and the multi-scale semantic representation vector of the valve surface state to obtain the valve surface state semantically guided valve state enhancement fusion feature map.

[0038] More specifically, the semantically guided feature enhancement unit 141 is used to: determine the weight coefficient matrix and bias vector of the multi-scale semantic representation vector of the valve surface state; multiply the multi-scale semantic representation vector of the valve surface state by the weight coefficient matrix, and then positionally add the product result and the bias vector to obtain a biased multi-scale semantic representation vector of the valve surface state; pass the biased multi-scale semantic representation vector of the valve surface state through a tanh activation function to obtain a weight probability vector; and use the weight probability vector as a weight to perform weighted processing on the shallow feature map of the valve surface state to obtain the enhanced shallow feature map of the valve surface state. That is, the shallow feature map of the valve surface state is subjected to feature enhancement processing using the following semantic information guided enhancement formula to obtain an enhanced shallow feature map of the valve surface state; wherein the semantic information guided enhancement formula is:

[0039] a c =tanh(W i V+B i )

[0040] F'=attention(a c ,F)

[0041] Where V is the multi-scale semantic representation vector of the valve surface state, W i is the weight coefficient matrix, B i is the bias vector, tanh is the nonlinear activation function, a c is the weight probability vector, F is the shallow feature map of the valve surface state, attention(a c , F) represents the weighted processing of the shallow feature map of the valve surface state using the weight probability vector, and F' is the enhanced shallow feature map of the valve surface state.

[0042] In the above-mentioned washing machine valve production monitoring and management system 100, the washing machine valve quality diagnosis module 150 is used to determine the quality inspection result of the washing machine valve based on the valve surface state semantics-guided valve state enhanced fusion feature map. In a specific example of the present application, the method of determining the quality inspection result of the washing machine valve based on the valve surface state semantics-guided valve state enhanced fusion feature map is to pass the valve surface state semantics-guided valve state enhanced fusion feature map through a quality inspector based on a classifier to obtain the quality inspection result, and the quality inspection result is used to indicate whether the washing machine valve is qualified. It should be understood that the quality inspector adopts a classification algorithm, and by using a large amount of sample data for training, it can learn the mapping relationship between the valve surface state characteristics and the valve quality type label. Based on this, the valve surface state semantics-guided valve state enhanced fusion feature map is input into the trained quality inspector, and the classifier model is used to identify the feature patterns in the valve surface state semantics-guided valve state enhanced fusion feature map. According to the mapping rules learned during the training process, the feature map is mapped to the corresponding label, that is, "washing machine valve is qualified" or "washing machine valve is unqualified", so as to obtain the quality inspection result of the valve and realize the automatic detection and diagnosis of the quality of the washing machine valve.

[0043] It should be understood that before using the above neural network model, it needs to be trained. That is, the washing machine valve production monitoring and management system of the present application also includes a training module for training the valve image multi-scale feature extractor based on the pyramid network, the semantic distiller based on the multi-scale semantic feature extraction network, the semantic guided attention feature selection network and the classifier-based quality inspector.

[0044] Figure 6 FIG. 1 is a block diagram of a training module in a washing machine valve production monitoring and management system according to an embodiment of the present application. Figure 6As shown, the training module 200 includes: a training data acquisition unit 210, which is used to acquire training data, wherein the training data includes a training valve digital image of a washing machine valve collected by an industrial camera, and a true value of whether the washing machine valve is qualified; a training image detail enhancement unit 220, which is used to perform contrast normalization and generalized Gaussian fitting on the training valve digital image to obtain a training detail enhanced valve digital image; a training image multi-scale feature extraction unit 230, which is used to pass the training detail enhanced valve digital image through the valve image multi-scale feature extractor based on the pyramid network to obtain a training valve surface state shallow feature map and a training valve surface state semantic feature map; a training image deep semantic feature distillation unit 240, which is used to pass the training valve surface state semantic feature map through the semantic distiller based on the multi-scale semantic feature extraction network to obtain a training valve surface state multi-scale semantic representation vector; a training image multi-scale semantic feature extraction unit 230, which is used to pass the training valve surface state semantic feature map through the semantic distiller based on the multi-scale semantic feature extraction network to obtain a training valve surface state multi-scale semantic representation vector; a training image multi-scale semantic feature extraction ...30, which is used to pass the training valve surface state semantic feature map through the semantic distiller based on the multi-scale semantic feature extraction network to obtain a training valve surface state multi-scale semantic representation vector; a training image multi-scale semantic feature extraction unit 230, which is used to pass the training valve surface state semantic feature map through the semantic distiller based on the multi-scale semantic feature extraction network to obtain a training valve surface state multi- A fusion unit 250 is used to pass the multi-scale semantic representation vector of the training valve surface state and the shallow feature map of the training valve surface state through the semantic-guided attention feature selection network to obtain a semantic-guided valve state enhanced fusion feature map of the training valve surface state; a classification loss calculation unit 260 is used to pass the semantic-guided valve state enhanced fusion feature map of the training valve surface state through the classifier-based quality inspector to obtain a probability score and a classification loss function value; a class probability related loss calculation unit 270 is used to calculate a class probability related loss function value based on the probability score; a model training unit 280 is used to train the pyramid network-based valve image multi-scale feature extractor, the multi-scale semantic feature extraction network-based semantic distiller, the semantic-guided attention feature selection network and the classifier-based quality inspector using the weighted sum of the classification loss function value and the class probability related loss function value as the loss function value.

[0045] In an optimized embodiment of the present application, during the model training process, a class probability related loss function is introduced for the probability score obtained by the classifier of the training valve surface state semantics-guided valve state enhanced fusion feature map at each iteration. The class probability related loss function is calculated specifically as follows: obtain the probability score obtained by the classifier of the training valve surface state semantics-guided valve state enhanced fusion feature map at each iteration, and subtract the probability score from one to obtain a score difference; determine the feature mean and feature variance of all feature values ​​of the training valve surface state semantics-guided valve state enhanced fusion feature map as a feature set; calculate the product of each feature value of the training valve surface state semantics-guided valve state enhanced fusion feature map and the probability score in the form of class cross entropy, and then subtract the product of the score difference and the feature mean to obtain a class cross entropy difference; The squares of the class cross entropy differences corresponding to each eigenvalue in the combined feature map are summed and divided by the scale of the training valve surface state semantics-guided valve state enhanced fusion feature map to obtain a first loss term, wherein the scale of the training valve surface state semantics-guided valve state enhanced fusion feature map is the width of the feature matrix of the training valve surface state semantics-guided valve state enhanced fusion feature map multiplied by the height and then multiplied by the number of channels of the training valve surface state semantics-guided valve state enhanced fusion feature map; the square sum of the probability score and the score difference is multiplied by the feature variance to obtain a second loss term; the first loss term is subtracted from the second loss term to obtain the class probability related loss function value.

[0046] Here, for the training detail enhanced valve digital image respectively expressed by the training valve surface state shallow feature map and the training valve surface state semantic feature map, the multi-scale and multi-depth image semantic feature representation based on the pyramid network, when the semantically guided attention is used to guide the local image semantic feature distribution through the multi-scale semantic representation of the deep image semantic feature, it will also cause the class distribution noise caused by the inconsistency of the in-attention-outside attention reinforcement feature distribution, causing the class distribution noise of the overall training valve surface state semantic guided valve state reinforcement fusion feature map, thereby affecting the accuracy of the classification result.

[0047] Therefore, the above-mentioned class probability related loss function sets the approximate expected variance characteristics of the out-of-distribution noise of the feature set of the training valve surface state semantics guided valve state enhanced fusion feature map relative to the Bernoulli sampling instance of the feature set based on the class probability distribution characteristics expressed by the probability score and score difference of the training valve surface state semantics guided valve state enhanced fusion feature map through the classifier. Therefore, the class probability related loss function is used to measure the distribution feature noise based on the class probability distribution characteristics expressed by the probability score and score difference of the training valve surface state semantics guided valve state enhanced fusion feature map. Based on the class probability distribution of each feature value of the training valve surface state semantics guided valve state enhanced fusion feature map during the classification process, the probability cross random selection characteristics of the expected variance of the training valve surface state semantics guided valve state enhanced fusion feature map, including the class cross entropy compound difference calculation, can be used to degrade possible sampling noise features, thereby realizing the class distribution noise reduction of self-supervised learning in the process of updating model parameters, so as to improve the accuracy of the quality inspection results obtained by the training valve surface state semantics guided valve state enhanced fusion feature map through the classifier-based quality inspector.

[0048] In summary, the washing machine valve production monitoring and management system according to the embodiment of the present application is explained, which uses machine vision technology based on deep learning to analyze and process the digital image of the washing machine valve to capture the multi-level semantic features of the valve digital image, and mine the deep information of the surface state of the washing machine valve, so as to intelligently identify whether the quality of the valve is qualified. In this way, automated valve quality inspection can be achieved to improve the quality monitoring efficiency of the washing machine valve, reduce manual intervention, and thus improve the overall production efficiency and product quality of the washing machine valve.

[0049] Figure 7 FIG. 1 is a flow chart of a method for monitoring and managing the production of washing machine valves according to an embodiment of the present application. Figure 7 As shown, according to the embodiment of the present application, the production monitoring and management method of the washing machine valve includes the following steps: S1, obtaining a valve digital image of the washing machine valve collected by an industrial camera; S2, performing deep and shallow feature extraction on the valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state; S3, performing multi-scale semantic enhancement on the semantic feature map of the valve surface state to obtain a multi-scale semantic representation vector of the valve surface state; S4, passing the multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state through a semantically guided attention feature selection network to obtain a valve surface state semantically guided valve state enhanced fusion feature map; S5, determining the quality inspection result of the washing machine valve based on the valve surface state semantically guided valve state enhanced fusion feature map.

[0050] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned washing machine valve production monitoring and management method have been referred to above. Figures 1 to 6 The invention has been described in detail in the description of the washing machine valve production monitoring and management system, and therefore, its repeated description will be omitted.

[0051] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0052] In the above embodiments, the description of each embodiment has its own emphasis. For the parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical sub-unit, that is, it may be located in one place, or it may be distributed on multiple network sub-units. Some or all of the units can be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0053] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0054] In addition, it is obvious that the word "comprising" does not exclude other sub-units or steps, and the singular does not exclude the plural. Multiple sub-units stated in the system claim can also be implemented by one sub-unit through software or hardware.

[0055] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A washing machine valve production monitoring and management system, characterized in that: include: A washing machine valve image acquisition module, used to acquire a valve digital image of the washing machine valve collected by an industrial camera; A depth feature extraction module, used for performing depth feature extraction on the valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state; A multi-scale semantic enhancement module, used for performing multi-scale semantic enhancement on the valve surface state semantic feature map to obtain a multi-scale semantic representation vector of the valve surface state; A semantically guided enhanced fusion module, used for obtaining a valve surface state semantically guided valve state enhanced fusion feature map by passing the multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state through a semantically guided attention feature selection network; A washing machine valve quality diagnosis module, used to guide the valve state reinforcement fusion feature map based on the valve surface state semantics to determine the quality inspection result of the washing machine valve; Wherein, the multi-scale semantic enhancement module is used to: The valve surface state semantic feature map is passed through a semantic distiller based on a multi-scale semantic feature extraction network to obtain a multi-scale semantic representation vector of the valve surface state; Wherein, the multi-scale semantic enhancement module includes: A multi-scale semantic encoding unit, used for respectively extracting global semantic features and local semantic features of the valve surface state semantic feature map to obtain a valve surface state global semantic feature vector and a valve surface state local semantic feature vector; A multi-scale semantic fusion unit, used for fusing the global semantic feature vector of the valve surface state and the local semantic feature vector of the valve surface state to obtain a multi-scale semantic representation vector of the valve surface state; Wherein, the multi-scale semantic coding unit includes: An upsampling subunit, used for upsampling the valve surface state semantic feature map to obtain an upsampled valve surface state semantic feature map; A global mean pooling subunit, used for performing global mean pooling processing on the up-sampled valve surface state semantic feature map to obtain an up-sampled valve surface state semantic channel feature vector; A semantic channel point convolution encoding subunit, used for performing point convolution encoding on the upsampled valve surface state semantic channel feature vector to obtain the valve surface state global semantic feature vector; A channel modulation subunit, used for performing point convolution encoding on the up-sampled valve surface state semantic feature map to obtain a channel-modulated up-sampled valve surface state semantic feature map; The local semantic feature extraction subunit is used to perform two-dimensional convolution encoding on the channel-modulated up-sampled valve surface state semantic feature map to obtain the valve surface state local semantic feature vector.

2. The washing machine valve production monitoring and management system according to claim 1, characterized in that: The depth feature extraction module comprises: An image detail enhancement processing unit, used for performing contrast normalization and generalized Gaussian fitting on the valve digital image to obtain a detail enhanced valve digital image; The image multi-scale feature extraction unit is used to pass the detail-enhanced valve digital image through a valve image multi-scale feature extractor based on a pyramid network to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state.

3. The washing machine valve production monitoring and management system according to claim 2, characterized in that: The semantic guided enhanced fusion module includes: A semantic guidance feature enhancement unit, configured to perform feature enhancement processing on the shallow feature map of the valve surface state based on the multi-scale semantic representation vector of the valve surface state to obtain an enhanced shallow feature map of the valve surface state; A fusion unit is used to fuse the enhanced valve surface state shallow feature map and the valve surface state multi-scale semantic representation vector to obtain the valve surface state semantics-guided valve state enhanced fusion feature map.

4. The washing machine valve production monitoring and management system according to claim 3, characterized in that: The semantic guidance feature enhancement unit is used to: Determining a weight coefficient matrix and a bias vector of the multi-scale semantic representation vector of the valve surface state; After multiplying the valve surface state multi-scale semantic representation vector and the weight coefficient matrix, the multiplication result and the bias vector are added according to the position to obtain the biased valve surface state multi-scale semantic representation vector; Passing the multi-scale semantic representation vector of the biased valve surface state through a tanh activation function to obtain a weight probability vector; The weighted probability vector is used as a weight to perform weighted processing on the shallow feature map of the valve surface state to obtain the enhanced shallow feature map of the valve surface state.

5. The washing machine valve production monitoring and management system according to claim 4, characterized in that: The washing machine valve quality diagnosis module is used to: The valve surface state semantics-guided valve state enhanced fusion feature map is passed through a classifier-based quality inspector to obtain the quality inspection result, and the quality inspection result is used to indicate whether the washing machine valve is qualified.

6. The washing machine valve production monitoring and management system according to claim 5, characterized in that: It also includes a training module for training the valve image multi-scale feature extractor based on the pyramid network, the semantic distiller based on the multi-scale semantic feature extraction network, the semantic guided attention feature selection network and the quality inspector based on the classifier; wherein the training module includes: A training data acquisition unit, used to acquire training data, wherein the training data includes a training valve digital image of a washing machine valve collected by an industrial camera, and a true value of whether the washing machine valve is qualified; A training image detail enhancement unit, used for performing contrast normalization and generalized Gaussian fitting on the training valve digital image to obtain a training detail enhanced valve digital image; A training image multi-scale feature extraction unit, used for passing the training detail-enhanced valve digital image through the valve image multi-scale feature extractor based on the pyramid network to obtain a training valve surface state shallow feature map and a training valve surface state semantic feature map; A training image deep semantic feature distillation unit, used for passing the training valve surface state semantic feature map through the semantic distiller based on the multi-scale semantic feature extraction network to obtain a training valve surface state multi-scale semantic representation vector; A training image multi-scale semantic fusion unit, used for passing the multi-scale semantic representation vector of the training valve surface state and the shallow feature map of the training valve surface state through the semantic guided attention feature selection network to obtain a training valve surface state semantic guided valve state enhanced fusion feature map; A classification loss calculation unit, used for passing the training valve surface state semantics-guided valve state enhanced fusion feature map through the classifier-based quality inspector to obtain a probability score and a classification loss function value; A class probability related loss calculation unit, used to calculate a class probability related loss function value based on the probability score; A model training unit is used to train the valve image multi-scale feature extractor based on the pyramid network, the semantic distiller based on the multi-scale semantic feature extraction network, the semantic-guided attention feature selection network and the classifier-based quality inspector using the weighted sum of the classification loss function value and the class probability related loss function value as the loss function value.

7. A method for monitoring and managing the production of washing machine valves, using the washing machine valve production monitoring and management system according to claim 1, characterized in that: include: Acquire a valve digital image of a washing machine valve captured by an industrial camera; Extracting shallow and deep features from the valve digital image to obtain a shallow feature map of the valve surface state and a semantic feature map of the valve surface state; Performing multi-scale semantic enhancement on the valve surface state semantic feature map to obtain a multi-scale semantic representation vector of the valve surface state; The multi-scale semantic representation vector of the valve surface state and the shallow feature map of the valve surface state are passed through a semantically guided attention feature selection network to obtain a valve surface state semantically guided valve state enhanced fusion feature map; The valve surface state semantics are used to guide the valve state enhanced fusion feature map, and the quality inspection result of the washing machine valve is determined.