MES-based Production Management System and Method for Window Decoration Products

Abstract

This application relates to the field of intelligent management. Specifically, it discloses a production management system and method for window decoration products based on MES. It collects the surface texture images of the window decoration products to be detected through a camera, extracts the designed texture pattern images from the background database, and uses image analysis and processing algorithms based on artificial intelligence technology to extract the texture features of the surface texture images of the window decoration products to be detected and the designed texture pattern images. Based on this, it can intelligently judge whether the window decoration products to be detected are qualified according to the differential semantic representation features of the textures of the two. In this way, it can automatically collect and process the images of the products, reduce the time and labor intensity of manual detection, and significantly improve the efficiency of production detection. At the same time, using artificial intelligence algorithms for image analysis can more accurately identify and extract deep and fine texture features, improving the accuracy of detection.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to a production management system and method for window decoration products based on MES. Background Art

[0002] Window decoration products, such as curtains, blinds, roller blinds, etc., are important components of interior decoration and architectural elements. They not only provide privacy protection, light control, and aesthetic decoration, but also have heat insulation and energy-saving functions. High-quality window decoration products are important factors in winning consumer trust and market competitiveness. Production management requires a strict quality control system to ensure that each product meets the standards.

[0003] However, in the production management of traditional window decoration products, the detection of window decoration products often can only detect surface or partial quality problems, and is often powerless for internal cracks or defects, resulting in difficulty in comprehensively covering all possible defects. In addition, the degree of intelligence of traditional quality detection is relatively low, which mainly relies on manual inspection. This not only has low efficiency, but is also easily affected by human judgment, resulting in difficulty in ensuring the consistency and accuracy of detection results.

[0004] Therefore, an optimized production management system for window decoration products is desired. Summary of the Invention

[0005] To solve the above technical problems, the present application is proposed. Embodiments of the present application provide a production management system and method for window decoration products based on MES. It collects the surface texture image of the window decoration product to be detected through a camera, extracts the designed texture pattern image from the background database, and uses an image analysis and processing algorithm based on artificial intelligence technology to extract the texture features of the surface texture image of the window decoration product to be detected and the designed texture pattern image. Based on this, it intelligently judges whether the window decoration product to be detected is qualified according to the differential semantic representation features of the surface texture image of the window decoration product to be detected and the designed texture pattern image in terms of texture. In this way, it can automatically collect and process the images of products, reduce the time and labor intensity of manual detection, and significantly improve the efficiency of production detection. At the same time, using artificial intelligence algorithms for image analysis can more accurately identify and extract deep and fine texture features, improving the accuracy of detection.

[0006] According to one aspect of the present application, a production management system for window decoration products based on MES is provided, which includes:

[0007] A module for collecting the texture image of the window decoration product to be detected, which is used to collect the surface texture image of the window decoration product to be detected through a camera;

[0008] A module for extracting the designed texture pattern image, which is used to extract the designed texture pattern image from the background database;

[0009] A module for extracting features of the reference texture image to be detected, which is used to input the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern;

[0010] A texture feature enhancement module, which is used to perform feature enhancement processing based on a compression suppression structure on the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern to obtain an enhanced surface texture feature map of the window decoration product to be detected and an enhanced reference feature map of the design texture pattern;

[0011] A texture differential semantic calculation module, which is used to calculate a texture differential semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the design texture pattern as a texture differential semantic representation feature;

[0012] A quality inspection result generation module, which is used to obtain a quality inspection result based on the texture differential semantic representation feature, and the quality inspection result is used to indicate whether the window decoration product to be detected is qualified.

[0013] According to another aspect of the present application, a production management method for window decoration products based on MES is provided, which includes:

[0014] Collect the surface texture image of the window decoration product to be detected through a camera;

[0015] Extract the design texture pattern image from the background database;

[0016] Input the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern;

[0017] Perform feature enhancement processing based on a compression suppression structure on the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern to obtain an enhanced surface texture feature map of the window decoration product to be detected and an enhanced reference feature map of the design texture pattern;

[0018] Calculate a texture differential semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the design texture pattern as a texture differential semantic representation feature;

[0019] Based on the texture differential semantic representation feature, obtain a quality inspection result, and the quality inspection result is used to indicate whether the window decoration product to be detected is qualified.

[0020] Compared with the prior art, a production management system and method for window decoration products based on MES provided by the present application collect surface texture images of the window decoration products to be detected through a camera, extract design texture pattern images from the background database, and use image analysis and processing algorithms based on artificial intelligence technology to extract texture features of the surface texture images of the window decoration products to be detected and the design texture pattern images. Based on this, it is possible to intelligently determine whether the window decoration products to be detected are qualified according to the differential semantic representation features of the surface texture images of the window decoration products to be detected and the design texture pattern images in terms of texture. In this way, it is possible to automatically collect and process product images, reduce the time and labor intensity of manual inspection, and significantly improve the efficiency of production inspection. At the same time, using artificial intelligence algorithms for image analysis can more accurately identify and extract deep and fine texture features, improving the accuracy of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, 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. They are used together with the embodiments of the present application to explain the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 It is a block diagram of a production management system for window decoration products based on MES according to an embodiment of the present application.

[0023] Figure 2 It is a schematic diagram of the architecture of a production management system for window decoration products based on MES according to an embodiment of the present application.

[0024] Figure 3 It is a block diagram of a texture feature enhancement module in a production management system for window decoration products based on MES according to an embodiment of the present application.

[0025] Figure 4 It is a block diagram of a surface texture feature weight calculation unit in a production management system for window decoration products based on MES according to an embodiment of the present application.

[0026] Figure 5 It is a flowchart of a production management method for window decoration products based on MES according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Instead, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0028] It should be understood that the various steps recited in the method embodiments of the present disclosure can be executed in a different order and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0029] In the description of the embodiments of the present disclosure, the term "including" and its like terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "an embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. There may also be other explicit and implicit definitions hereinafter.

[0030] It should be noted that the modifications referring to "one" and "plural" in the present disclosure are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly specified in the context, it should be understood as "one or more".

[0031] Window covering products, such as curtains, blinds, roller shades, etc., are an important part of interior decoration and architectural elements, providing not only privacy protection, light control, and aesthetic decoration, but also heat insulation and energy-saving functions. High-quality window covering products are an important factor in winning consumers' trust and market competitiveness. Production management requires a strict quality control system to ensure that each product meets the standards.

[0032] However, the detection of window covering products in traditional window covering product production management can often only detect surface or local quality problems of the products, and is often powerless for internal cracks or defects, resulting in difficulty in comprehensively covering all possible defects. In addition, the traditional quality detection has a low degree of intelligence and mainly relies on manual inspection, which is not only inefficient but also easily affected by human judgment, making it difficult to ensure the consistency and accuracy of the detection results.

[0033] It is worth mentioning that MES (Manufacturing Execution System) is a software system specifically designed for factory environments, which can coordinate, track, record, and monitor the entire manufacturing process from raw materials to semi-finished and finished products. In the production management of window decoration products, the MES system can formulate production plans according to orders and inventory situations and allocate them to each production line to ensure the smooth progress of the production process. In addition, the MES system can monitor product quality, promptly detect and correct quality problems in the production process, and ensure that products meet quality standards. Therefore, the MES system plays an important role in the production management of window decoration products, enabling enterprises to achieve automation, informatization, and intelligence in the production process, thereby improving production efficiency, reducing costs, enhancing product quality, and meeting customer needs.

[0034] Based on this, the present application proposes a production management system for window decoration products based on MES. Firstly, according to the production plan or customer requirements, a production order is officially issued. Then, the orders after placing are scheduled to arrange the production sequence. Next, the scheduled orders are handed over to the workshop for production. The workshop cuts the tracks, marks, and cuts the fabrics of the window decoration products to be produced in the order respectively, and then assembles the materials after marking and fabric cutting to obtain the window decoration products to be inspected. After that, the assembled window decoration products to be inspected are tested for quality compliance, and the qualified window decoration products are successively subjected to inner packaging and outer packaging. Finally, the packaged window decoration products are arranged for shipment for product transportation and distribution to ensure timely delivery to customers. In particular, the production management system for window decoration products based on MES can place various personalized orders. Operators at each step can scan the code to query the order content through the system for operation, and record each operator, which is convenient for tracing defective products.

[0035] Correspondingly, in the above-mentioned inspection of the assembled window decoration products to be inspected for quality compliance, the surface texture image of the window decoration products to be inspected is collected through a camera, and the designed texture pattern image is extracted from the background database, and an image analysis and processing algorithm based on artificial intelligence technology is used to extract the texture features of the surface texture image of the window decoration products to be inspected and the designed texture pattern image. Based on this, it is intelligently determined whether the window decoration products to be inspected are qualified according to the differential semantic representation features of the surface texture image of the window decoration products to be inspected and the designed texture pattern image in terms of texture. In this way, it is possible to automatically collect and process product images, reduce the time and labor intensity of manual inspection, and significantly improve the efficiency of production inspection. At the same time, using artificial intelligence algorithms for image analysis can more accurately identify and extract deep and fine texture features, improving the accuracy of inspection.

[0036] Figure 1It is a block diagram of a production management system for window decoration products based on MES according to an embodiment of the present application. Figure 2 It is a schematic diagram of the architecture of a production management system for window decoration products based on MES according to an embodiment of the present application. As Figure 1 and Figure 2 shown, the production management system 100 for window decoration products based on MES according to an embodiment of the present application includes: a texture image acquisition module 110 for the window decoration product to be detected, which is used to acquire the surface texture image of the window decoration product to be detected through a camera; a design texture pattern image extraction module 120, which is used to extract the design texture pattern image from the background database; a reference texture image feature extraction module 130 for the window decoration product to be detected, which is used to input the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern; a texture feature enhancement module 140, which is used to perform feature enhancement processing based on a compression suppression structure on the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern to obtain an enhanced surface texture feature map of the window decoration product to be detected and an enhanced reference feature map of the design texture pattern; a texture differential semantic calculation module 150, which is used to calculate the texture differential semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the design texture pattern as the texture differential semantic representation feature; a quality inspection result generation module 160, which is used to obtain a quality inspection result based on the texture differential semantic representation feature, and the quality inspection result is used to indicate whether the window decoration product to be detected is qualified.

[0037] In the embodiment of the present application, the texture image acquisition module 110 for the window decoration product to be detected is used to acquire the surface texture image of the window decoration product to be detected through a camera. The design texture pattern image extraction module 120 is used to extract the design texture pattern image from the background database. It should be understood that considering that the surface texture image of the window decoration product to be detected is directly taken by a camera, it shows the texture and pattern on the actual surface of the window decoration product. The design texture pattern image extracted from the background database is pre-stored in the background database and represents the original texture and pattern of the window decoration product design. Usually, it is an ideal texture image generated by design software and can be used as a standard reference for product quality qualification detection. Based on this, in the technical solution of the present application, by acquiring the surface texture image of the window decoration product to be detected through a camera, extracting the design texture pattern image from the background database, and analyzing and processing it, the difference between the actual product texture image and the design texture pattern image can be automatically compared, so as to detect whether the product is manufactured according to the design specifications, whether there are defects or deviations, thereby ensuring that the product quality meets the design and manufacturing standards.

[0038] In the embodiment of the present application, the module 130 for extracting features of the reference texture image to be detected is configured to input the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern. Specifically, in the embodiment of the present application, the module for extracting features of the reference texture image to be detected is configured to: input the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network including a first image encoder and a second image encoder to obtain the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern, where the first image encoder and the second image encoder have the same network structure. Correspondingly, considering that both the surface texture image of the window decoration product to be detected and the design texture pattern image contain feature information about the surface texture of the window decoration product, this surface texture information plays a crucial role in the subsequent judgment of the quality of the window decoration to be detected. Based on this, in the technical solution of the present application, the surface texture image of the window decoration product to be detected and the design texture pattern image are input into a texture twin detection network including a first image encoder and a second image encoder to respectively capture and extract the surface texture feature information of the product image and the design image, so as to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern. In particular, the first image encoder and the second image encoder described here have the same network structure.

[0039] In an embodiment of the present application, the texture feature enhancement module 140 is used to perform feature enhancement processing based on a compression suppression structure on the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern to obtain an enhanced surface texture feature map of the window decoration product to be detected and an enhanced reference feature map of the design texture pattern. Specifically, in an embodiment of the present application, the texture feature enhancement module is used to: input the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern into a feature attention selection enhancer under a compression suppression structure respectively to obtain the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the design texture pattern. It should be understood that, considering that the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern contain subtle key texture features that are crucial to the qualified rate judgment of the window decoration product to be detected, however, these two feature maps are also mixed with some non-surface texture key background features, and this information is not necessary for the texture recognition process and may even cause interference. Therefore, in order to pay special attention to and extract those tiny texture semantic features that play a decisive role, while excluding those irrelevant background features, so as to ensure that the quality judgment of the window decoration product to be detected is both accurate and efficient, in the technical solution of the present application, the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern are respectively input into the feature attention selection enhancer under the compression suppression structure to obtain the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the design texture pattern. Specifically, firstly, the surface texture feature map of the window decoration product to be detected is processed by global average pooling to compress and extract the most core surface texture semantic features, thereby forming a texture feature compression information representation vector. Then, the texture feature compression information representation vector is deeply analyzed using one-dimensional convolutional coding to explore and reveal the internal semantic connection about the surface texture in the vector. Subsequently, these features processed by convolutional coding are integrated with the texture feature compression information representation vector to enhance the model's recognition and understanding ability of the subtle semantic features of the local texture on the surface of the window decoration product to be detected, so as to obtain a multi-scale representation vector of the surface texture feature compression information of the window decoration product to be detected. Next, by constructing a multi-layer perceptron network containing two fully connected layers and SiLU activation function, the feature fine-grained semantic information of the multi-scale representation vector of the surface texture feature compression information of the window decoration product to be detected obtained through the cascade operation is further deepened and refined. Furthermore, the Sigmoid activation function is used to normalize the deepened and refined multi-scale correlation feature vector to obtain the relative importance weight of each feature channel. Finally, the surface texture feature map of the window decoration product to be detected is weighted and adjusted according to these weights to enhance the significance of the tiny surface texture semantic features and reduce the influence of non-critical features, thereby obtaining an enhanced surface texture feature map of the window decoration product to be detected.Specifically, the processing method of the reference feature map of the designed texture pattern is consistent with the processing method of the surface texture feature map of the window decoration product to be detected.

[0040] Figure 3 FIG. is a block diagram of a texture feature enhancement module in a window decoration product production management system based on MES according to an embodiment of the present application. Specifically, in the embodiment of the present application, as Figure 3 shown, the texture feature enhancement module 140 includes: a surface texture feature compression unit 141 of the window decoration product to be detected, configured to perform channel compression on the surface texture feature map of the window decoration product to be detected to obtain a surface texture feature compression information representation vector of the window decoration product to be detected; a surface texture feature compression correlation unit 142 of the window decoration product to be detected, configured to perform one-dimensional convolutional encoding on the surface texture feature compression information representation vector of the window decoration product to be detected to obtain a correlation representation feature vector between surface texture compression information of the window decoration product to be detected; a surface texture multi-scale feature fusion unit 143 of the window decoration to be detected, configured to concatenate the surface texture feature compression information representation vector of the window decoration product to be detected and the correlation representation feature vector between surface texture compression information of the window decoration product to be detected to obtain a multi-scale representation vector of surface texture compression information of the window decoration product to be detected; a texture feature multi-scale correlation unit 144, configured to input the multi-scale representation vector of surface texture compression information of the window decoration product to be detected into a compression information feature extraction module to obtain a multi-scale correlation feature vector of surface texture compression information of the window decoration product to be detected, where the compression information feature extraction module is a multi-layer perceptron including two fully connected layers and a SiLU activation function; a surface texture feature weight calculation unit 145, configured to perform a normalization operation on the multi-scale correlation feature vector of surface texture compression information of the window decoration product to be detected using a Sigmoid function to obtain a surface texture weight feature vector of the window decoration product to be detected; a texture feature amplification and suppression unit 146, configured to perform feature amplification and suppression operations on the surface texture feature map of the window decoration product to be detected based on the surface texture weight feature vector of the window decoration product to be detected to obtain the enhanced surface texture feature map of the window decoration product to be detected.

[0041] More specifically, in the embodiment of the present application, the surface texture feature compression unit of the window decoration product to be detected is configured to: calculate the global mean of each feature matrix of the surface texture feature map of the window decoration product to be detected along the channel dimension to obtain the surface texture feature compression information representation vector of the window decoration product to be detected.

[0042] Figure 4 FIG. is a block diagram of a surface texture feature weight calculation unit in a window decoration product production management system based on MES according to an embodiment of the present application. More specifically, in the embodiment of the present application, as Figure 4As shown, the surface texture feature weight calculation unit 145 includes: a surface texture compression information multi-scale exponential value calculation subunit 1451, which is used to use the negative of the feature value at each position in the multi-scale correlation feature vector of the surface texture compression information of the window decoration product to be detected as the exponent of the natural constant to calculate the exponential function value with the natural constant as the base for each position to obtain the multi-scale correlation class support feature vector of the surface texture compression information of the window decoration product to be detected; a surface texture weight value calculation subunit 1452, which is used to calculate the reciprocal of the sum of the feature value at each position in the multi-scale correlation class support feature vector of the surface texture compression information of the window decoration product to be detected and a constant one to obtain the surface texture weight feature vector of the window decoration product to be detected.

[0043] More specifically, in the embodiment of the present application, the texture feature amplification and suppression unit is used to: multiply the feature value at each position in the surface texture weight feature vector of the window decoration product to be detected by each feature matrix along the channel dimension of the surface texture feature map of the window decoration product to be detected to obtain the enhanced surface texture feature map of the window decoration product to be detected.

[0044] In the embodiment of the present application, specifically, the texture feature enhancement module is used to: input the surface texture feature map of the window decoration product to be detected into the feature attention selection enhancer under the compression and suppression structure, and process it with the following enhancement formula to obtain the enhanced surface texture feature map of the window decoration product to be detected; where the enhancement formula is:

[0045]

[0046] V s = concat(V s1 , Conv 1d (V s1 ))

[0047]

[0048] Where F is the surface texture feature map of the window decoration product to be detected, f xo (i o , j o ) represents the feature value of the point with coordinates (i o , j o , j o ) in the x s1 -th channel of the surface texture feature map F of the window decoration product to be detected, H and W are respectively the height and width of the surface texture feature map of the window decoration product to be detected, V 1d (·) is a one-dimensional convolutional encoding, concat(·,·) is a concatenation process for vectors, and V sis the multi-scale representation vector of the surface texture compression information of the window decoration product to be detected, and MLP is a multi-layer perceptron. denotes multiplication by position, and F' is the enhanced surface texture feature map of the window decoration product to be detected. In particular, the encoding method of the reference feature map of the designed texture pattern is consistent with the encoding method of the surface texture feature map of the window decoration product to be detected.

[0049] In the embodiment of the present application, the texture difference semantic calculation module 150 is used to calculate the texture difference semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the designed texture pattern as the texture difference semantic representation feature. Accordingly, in order to further compare and analyze the differences between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the designed texture pattern, such as texture missing, misalignment or extra texture, etc., so as to more accurately judge the qualification problem of the window decoration product to be detected. In the technical solution of the present application, the texture difference semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the designed texture pattern is calculated. That is, the difference feature map provides a quantitative method to evaluate the quality of the product surface texture, which helps to highlight those small but critical texture features in the window decoration products to be detected, thereby improving the accuracy of production inspection management.

[0050] In the embodiment of the present application, the quality inspection result generation module 160 is used to obtain a quality inspection result based on the texture difference semantic representation feature, and the quality inspection result is used to indicate whether the window decoration product to be detected is qualified. Specifically, in the embodiment of the present application, the quality inspection result generation module is used to: input the texture difference semantic representation feature map into the quality inspection module based on a classifier to obtain the quality inspection result, and the quality inspection result is used to indicate whether the window decoration product to be detected is qualified. That is, the texture difference semantic representation feature between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the designed texture pattern is used for classification processing, so as to intelligently judge whether the window decoration product to be detected is qualified. In this way, it is possible to automatically collect and process the images of the products, reduce the time and labor intensity of manual inspection, and significantly improve the efficiency of production inspection. At the same time, using artificial intelligence algorithms for image analysis can more accurately identify and extract deep and fine texture features, improving the accuracy of detection.

[0051] It should be understood that in the technical solution of this application, the surface texture feature map of the window decoration product to be detected and the reference feature map of the designed texture pattern respectively represent the image semantic features of the designed texture pattern image and the surface texture image of the window decoration product to be detected based on the texture twin detection network. Considering that in addition to the difference in the content of the image source domain between the designed texture pattern image and the surface texture image of the window decoration product to be detected, detection interferences introduced during the image acquisition process such as image noise and image distortion will also be introduced when collecting the surface texture image of the window decoration product to be detected. This makes the surface texture feature map of the window decoration product to be detected generated by encoding through the texture twin detection network have many types of out-of-distribution noises relative to the reference feature map of the designed texture pattern. After the feature attention selection enhancement and feature map differential processing under the compression suppression structure, the obtained texture differential semantic representation feature map will have a lack of local causal association in the posterior class probability relative to the image source domain differential representation between the surface texture image of the window decoration product to be detected and the designed texture pattern image, thus affecting the accuracy of the classification result.

[0052] Based on this, in a preferred embodiment of this application, inputting the texture differential semantic representation feature map into the quality inspection module based on a classifier to obtain the quality inspection result includes: probabilizing each eigenvalue of the texture differential semantic representation feature map based on a probability activation function, such as the sigmoid function or the softmax function, to obtain a probabilized texture differential semantic representation feature map; obtaining the qualified probability value obtained by inputting the texture differential semantic representation feature map into the quality inspection module based on the classifier; determining the class recognition symbol value based on the comparison between each eigenvalue of the probabilized texture differential semantic representation feature map and the qualified probability value, where the class recognition symbol value is equal to one, zero, and negative one respectively in response to the eigenvalue of the probabilized texture differential semantic representation feature map being greater than, equal to, and less than the qualified probability value; calculating the mean value of all eigenvalues of the probabilized texture differential semantic representation feature map to obtain the class overall phase shift value; after multiplying each eigenvalue of the probabilized texture differential semantic representation feature map by the class recognition symbol value and the class overall phase shift value respectively, performing weighted differential calculation and taking the absolute value to obtain the optimized eigenvalue of the probabilized texture differential semantic representation feature map; inputting the optimized texture differential semantic representation feature map composed of the optimized eigenvalues into the quality inspection module based on the classifier to obtain the quality inspection result.

[0053] Specifically, in this preferred embodiment, optimizing the probabilized texture differential semantic representation feature map to obtain an optimized texture differential semantic representation feature map, the process is represented by the following formula:

[0054]

[0055] Among them, f i,j,k is the eigenvalue at each position in the probability-based texture difference semantic representation feature map, p is the qualified probability value, sgn is the sign function, and α and β are weight hyperparameters. is the class overall phase shift value, and f' i,j,k is the eigenvalue at each position in the optimized texture difference semantic representation feature map.

[0056] Correspondingly, in the above optimization process, by comparing the probability amplitude of the eigenvalue of the texture difference semantic representation feature map with the class probability, the class cognitive phase transformation response of the texture difference semantic representation feature map is obtained, and the phase shift response of the eigenvalue of the texture difference semantic representation feature map relative to the class probability representation of the overall feature map is subjected to an invariant transformation of the feature distribution sequence based on the class difference distribution, so as to realize the causal constraint of the posterior class probability of the texture difference semantic representation feature map on its prior feature distribution representation, and improve the accuracy of the quality inspection result obtained by inputting the texture difference semantic representation feature map into the quality inspection module based on the classifier. In this way, the images of products can be automatically collected and processed, reducing the time and labor intensity of manual inspection and significantly improving the efficiency of production inspection. At the same time, using artificial intelligence algorithms for image analysis can more accurately identify and extract deep and fine texture features, improving the accuracy of detection.

[0057] In summary, the MES-based window decoration product production management system 100 according to the embodiments of the present application is elucidated. It collects the surface texture image of the window decoration product to be detected through a camera, extracts the designed texture pattern image from the background database, and uses image analysis and processing algorithms based on artificial intelligence technology to extract the texture features of the surface texture image of the window decoration product to be detected and the designed texture pattern image, and thus intelligently determines whether the window decoration product to be detected is qualified according to the differential semantic representation features of the surface texture image of the window decoration product to be detected and the designed texture pattern image in terms of texture. In this way, the images of products can be automatically collected and processed, reducing the time and labor intensity of manual inspection and significantly improving the efficiency of production inspection. At the same time, using artificial intelligence algorithms for image analysis can more accurately identify and extract deep and fine texture features, improving the accuracy of detection.

[0058] As described above, the MES-based production management system 100 for window decoration products according to the embodiments of the present application can be implemented in various wireless terminals, such as a server with an MES-based production management algorithm for window decoration products. In a possible implementation manner, the MES-based production management system 100 for window decoration products according to the embodiments of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the MES-based production management system 100 for window decoration products can be a software module in the operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the MES-based production management system 100 for window decoration products can also be one of the numerous hardware modules of the wireless terminal.

[0059] Alternatively, in another example, the MES-based production management system 100 for window decoration products and the wireless terminal can also be separate devices, and the MES-based production management system 100 for window decoration products can be connected to the wireless terminal through a wired and / or wireless network and transmit interaction information according to a predefined data format.

[0060] Figure 5 FIG. is a flowchart of an MES-based production management method for window decoration products according to the embodiments of the present application. As Figure 5 shown, the MES-based production management method for window decoration products according to the embodiments of the present application includes: S110, collecting a surface texture image of a window decoration product to be detected through a camera; S120, extracting a designed texture pattern image from a background database; S130, inputting the surface texture image of the window decoration product to be detected and the designed texture pattern image into a texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the designed texture pattern; S140, performing feature enhancement processing based on a compression suppression structure on the surface texture feature map of the window decoration product to be detected and the reference feature map of the designed texture pattern to obtain an enhanced surface texture feature map of the window decoration product to be detected and an enhanced reference feature map of the designed texture pattern; S150, calculating a texture difference semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the designed texture pattern as a texture difference semantic representation feature; S160, obtaining a quality inspection result based on the texture difference semantic representation feature, where the quality inspection result is used to indicate whether the window decoration product to be detected is qualified.

[0061] Here, those skilled in the art can understand that the specific operations of each step in the above MES-based production management method for window decoration products have been introduced in detail in the description of the MES-based production management system for window decoration products above, and therefore, the repeated description thereof will be omitted. Figures 1 to 4 of the MES-based production management system for window decoration products, and thus, the repeated description thereof will be omitted.

[0062] The various implementations of the present disclosure have been described above. The above description is exemplary and not exhaustive. Nor is it limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The choice of terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to technologies in the market, or to enable other ordinary skill in the art to understand the various implementation manners disclosed herein.

Claims

1. A window decoration product production management system based on MES, characterized in that: include: A texture image acquisition module for the window decoration product to be detected, used to acquire the surface texture image of the window decoration product to be detected through a camera; Design texture pattern image extraction module, used for extracting design texture pattern images from the background database; A reference texture image feature extraction module to be detected, used for inputting the surface texture image of the window decoration product to be detected and the design texture pattern image into the texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern; A texture feature enhancement module, used for performing feature enhancement processing on the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern based on a compression suppression structure to obtain an enhanced surface texture feature map of the window decoration product to be detected and an enhanced reference feature map of the design texture pattern; A texture difference semantic calculation module, used to calculate a texture difference semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be detected and the enhanced design texture pattern reference feature map as a texture difference semantic representation feature; A quality inspection result generating module, used to obtain a quality inspection result based on the texture difference semantic representation feature, wherein the quality inspection result is used to indicate whether the window decoration product to be inspected is qualified; Wherein, the texture feature enhancement module includes: A surface texture feature compression unit for the window decoration product to be detected, used for performing channel compression on the surface texture feature map of the window decoration product to be detected to obtain a surface texture feature compression information representation vector of the window decoration product to be detected; A surface texture feature compression and association unit for the window decoration product to be detected is used to perform one-dimensional convolution coding on the surface texture feature compression information representation vector of the window decoration product to be detected to obtain a correlation representation feature vector between the surface texture compression information of the window decoration product to be detected; A multi-scale feature fusion unit for the surface texture of the window decoration to be detected, used for cascading the representation vector of the surface texture feature compression information of the window decoration product to be detected and the feature vector representing the correlation between the surface texture compression information of the window decoration product to be detected to obtain a multi-scale representation vector of the surface texture compression information of the window decoration product to be detected; A texture feature multi-scale association unit, used for inputting the multi-scale representation vector of the surface texture compression information of the window decoration product to be detected into a compression information feature extraction module to obtain a multi-scale association feature vector of the surface texture compression information of the window decoration product to be detected, wherein the compression information feature extraction module is a multi-layer perceptron including two fully connected layers and a SiLU activation function; A surface texture feature weight calculation unit, used for using a Sigmoid function to perform a normalization operation on the multi-scale associated feature vector of the surface texture compression information of the window decoration product to be detected to obtain a surface texture weight feature vector of the window decoration product to be detected; The texture feature amplification and suppression unit is used to perform feature amplification and suppression operations on the surface texture feature map of the window decoration product to be detected based on the surface texture weight feature vector of the window decoration product to be detected to obtain the enhanced surface texture feature map of the window decoration product to be detected.

2. The MES-based window decoration product production management system according to claim 1 is characterized in that: The reference texture image feature extraction module to be detected is used to: input the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network including a first image encoder and a second image encoder to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern, wherein the first image encoder and the second image encoder have the same network structure.

3. The MES-based window decoration product production management system according to claim 2 is characterized in that: The texture feature enhancement module is used to: input the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern into the feature attention selection enhancer under the compression suppression structure respectively to obtain the enhanced surface texture feature map of the window decoration product to be detected and the enhanced reference feature map of the design texture pattern.

4. The MES-based window decoration product production management system according to claim 3 is characterized in that: The surface texture feature compression unit of the window decoration product to be detected is used to calculate the global mean of each feature matrix along the channel dimension of the surface texture feature map of the window decoration product to be detected to obtain the surface texture feature compression information representation vector of the window decoration product to be detected.

5. The MES-based window decoration product production management system according to claim 4 is characterized in that: The surface texture feature weight calculation unit comprises: A surface texture compression information multi-scale exponential value calculation subunit, used to use the negative number of the eigenvalue of each position in the multi-scale association eigenvector of the surface texture compression information of the window decoration product to be detected as the exponent of the natural constant to calculate the exponential function value based on the position with the natural constant as the base to obtain the multi-scale association class support eigenvector of the surface texture compression information of the window decoration product to be detected; The surface texture weight value calculation subunit is used to calculate the reciprocal of the sum of the eigenvalue of each position in the multi-scale association class support feature vector of the surface texture compression information of the window decoration product to be detected and the constant one to obtain the surface texture weight feature vector of the window decoration product to be detected.

6. The MES-based window decoration product production management system according to claim 5 is characterized in that: The texture feature amplification and suppression unit is used to: multiply the eigenvalues ​​of each position in the surface texture weight feature vector of the window decoration product to be detected by each feature matrix along the channel dimension of the surface texture feature map of the window decoration product to be detected by position to obtain the enhanced surface texture feature map of the window decoration product to be detected.

7. The MES-based window decoration product production management system according to claim 6 is characterized in that: The quality inspection result generating module is used to: input the texture difference semantic representation feature map into the classifier-based quality inspection module to obtain the quality inspection result, and the quality inspection result is used to indicate whether the window decoration product to be inspected is qualified.

8. A method for managing window decoration product production based on MES, using the window decoration product production management system based on MES according to claim 1, characterized in that: include: The surface texture image of the window decoration product to be inspected is collected by a camera; Extracting design texture pattern images from a background database; Inputting the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a design texture pattern reference feature map; Performing feature enhancement processing based on a compression suppression structure on the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern to obtain an enhanced surface texture feature map of the window decoration product to be detected and an enhanced reference feature map of the design texture pattern; Calculating a texture difference semantic representation feature map between the enhanced surface texture feature map of the window decoration product to be inspected and the enhanced design texture pattern reference feature map as a texture difference semantic representation feature; Based on the texture difference semantic representation feature, a quality inspection result is obtained, and the quality inspection result is used to indicate whether the window decoration product to be inspected is qualified.

9. The MES-based window decoration product production management method according to claim 8, characterized in that: The surface texture image of the window decoration product to be detected and the design texture pattern image are input into a texture twin detection network to obtain a surface texture feature map of the window decoration product to be detected and a reference feature map of the design texture pattern, including: inputting the surface texture image of the window decoration product to be detected and the design texture pattern image into a texture twin detection network including a first image encoder and a second image encoder to obtain the surface texture feature map of the window decoration product to be detected and the reference feature map of the design texture pattern, wherein the first image encoder and the second image encoder have the same network structure.

Images