Plastic-steel door and window production management system and method based on computer

Abstract

The invention relates to the field of intelligent management of plastic-steel doors and windows, and particularly discloses a plastic-steel door and window production management system and method based on a computer, which uses an artificial intelligence technology based on the deep learning field to carry out feature extraction and coding on an image of a section of a profile so as to obtain a classification result about whether the cutting effect of the profile is qualified or not. Thus, by intelligently judging the profile cutting effect, the profile cutting detection fineness is improved, meanwhile, the labor cost is reduced, careless omission of the detection result is avoided, and the profile cutting quality is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of intelligent management of plastic-steel doors and windows, and in particular to a computer-based production management system and method for plastic-steel doors and windows. Background Art

[0002] Plastic-steel doors and windows, a new type of building window and door design, offer advantages such as durability, long service life, excellent sound and heat insulation, waterproof and moisture-resistant properties, and a stylish appearance. With rising living standards, plastic-steel doors and windows are gradually becoming part of people's daily lives. Plastic-steel doors and windows are primarily made of polyvinyl chloride (PVC-U) resin. After extrusion, the profiles are cut and reinforced with steel. The frames and sashes are then assembled through welding or screwing, followed by sealing strips, hardware, and glass. The cutting process for plastic-steel doors and windows requires strict control standards, such as burr-free and precise length and angle. During the cutting process, the quality of the cut profiles must be inspected to check for any substandard cuts. Traditional inspection methods often rely on manual labor, which can easily overlook minor defects and lead to fatigue, resulting in inaccuracies in the inspection results.

[0003] Therefore, an optimized production management system for plastic-steel doors and windows is expected. Summary of the Invention

[0004] The present application is proposed to address the aforementioned technical issues. The embodiments of the present application provide a computer-based plastic-steel door and window production management system and method, which utilizes artificial intelligence technology based on deep learning to extract and encode features from images of profile sections to obtain a classification result indicating whether the profile cutting effect is acceptable. This intelligent judgment of the profile cutting effect improves the precision of profile cutting detection, reduces labor costs, avoids omissions in inspection results, and ensures the quality of profile cutting.

[0005] According to one aspect of the present application, a computer-based plastic-steel door and window production management system is provided, which includes:

[0006] A profile cross-section image acquisition module is used to acquire images of the profile cross-section;

[0007] A profile section image correction module, configured to correct the profile section image to obtain a corrected profile section image;

[0008] A profile section feature analysis module, configured to analyze the corrected profile section image to obtain a profile section associated feature vector;

[0009] The cutting effect analysis module is used to analyze the profile section-related feature vectors to obtain a result indicating whether the profile cutting effect is qualified.

[0010] According to another aspect of the present application, a computer-based production management method for plastic-steel doors and windows is provided, comprising:

[0011] Acquire an image of the cross section of the profile;

[0012] Correcting the image of the profile section to obtain a corrected profile section image;

[0013] Analyzing the corrected profile section image to obtain a profile section correlation feature vector;

[0014] The profile section associated feature vector is analyzed to obtain a result indicating whether the profile cutting effect is qualified.

[0015] In summary, the computer-based plastic-steel door and window production management system and method provided by this application uses artificial intelligence technology based on deep learning to extract and encode features from images of profile sections, thereby obtaining a classification result that determines whether the profile cutting effect is qualified. In this way, by intelligently judging the profile cutting effect, the precision of profile cutting detection is improved, while labor costs are reduced, omissions in inspection results are avoided, and the quality of profile cutting is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. 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. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without inventive effort. In the drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 This is a block diagram of a computer-based plastic-steel door and window production management system according to an embodiment of the present application.

[0018] Figure 2 This is a block diagram of a profile section feature analysis module in a computer-based plastic-steel door and window production management system according to an embodiment of the present application.

[0019] Figure 3 This is a block diagram of a directional gradient histogram analysis unit in a computer-based plastic-steel door and window production management system according to an embodiment of the present application.

[0020] Figure 4 This is a flowchart of a computer-based plastic-steel door and window production management method according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] Below, the exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings, clearly and completely describing the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0022] Figure 1 FIG is a block diagram of a computer-based plastic-steel door and window production management system according to an embodiment of the present application. Figure 1 As shown, according to an embodiment of the present application, the computer-based plastic-steel door and window production management system 100 includes: a profile section image acquisition module 110, used to obtain an image of a profile section; a profile section image correction module 120, used to correct the image of the profile section to obtain a corrected profile section image; a profile section feature analysis module 130, used to analyze the corrected profile section image to obtain a profile section associated feature vector; and a cutting effect analysis module 140, used to analyze the profile section associated feature vector to obtain a result indicating whether the profile cutting effect is qualified.

[0023] In the computer-based plastic-steel door and window production management system 100, the profile cross-section image acquisition module 110 is used to capture images of the profile cross-section. As discussed in the background art, during the profile cutting process, it is necessary to inspect the quality of the cut profile to see if any substandard cutting quality is present. Traditional inspection methods often rely on manual labor, which can lead to inattentive inspections of minor defects and manual fatigue, resulting in omissions in the inspection results. Therefore, an optimized plastic-steel door and window production management system is desired.

[0024] To address these technical issues, a computer-based plastic-steel door and window production management system has been proposed. This system uses deep learning-based artificial intelligence (AI) to extract and encode features from images of profile sections, categorizing whether the profile cuts are acceptable. This intelligent assessment of profile cutting improves the precision of profile cutting inspections, reduces labor costs, avoids oversights in inspection results, and ensures the quality of profile cutting.

[0025] Currently, deep learning and neural networks are widely used in fields such as computer vision, natural language processing, and speech signal processing. Furthermore, deep learning and neural networks have demonstrated capabilities approaching or even surpassing those of humans in areas such as image classification, object detection, semantic segmentation, and text translation.

[0026] In recent years, the development of deep learning and neural networks has provided new solutions and plans for computer-based plastic steel door and window production management systems.

[0027] Specifically, first, obtain an image of the profile cross section. By obtaining the image of the profile cross section, the quality of the profile cutting can be monitored in real time, including surface flatness, cutting accuracy, etc., so as to promptly discover and correct possible problems and ensure product quality. By using image processing and deep learning technology, it is possible to realize automated detection and analysis of the profile cross section, improve production efficiency, and reduce labor costs. Specifically, the above-mentioned images can be obtained through the following methods: 1. Camera shooting: Install a camera or industrial camera on the production line to capture images of the profile cross section in real time. 2. Sensor acquisition: Use sensor technology, such as optical sensors, to obtain the shape and surface information of the profile cross section and convert it into image data. 3. Use a professional profile cross section scanner to scan the profile cross section to obtain high-definition image data. Through the above methods, images of the profile cross section can be effectively obtained and used in subsequent quality control and production management processes.

[0028] In the computer-based plastic-steel door and window production management system 100 described above, the profile section image correction module 120 is used to correct the profile section image to obtain a corrected profile section image. Correction can help eliminate distortion, noise, and other undesirable factors in the image, thereby improving image clarity and quality. Furthermore, correction can reduce interference factors in the image, enabling the system to process image data more stably and reducing errors and uncertainties. The corrected image conforms more closely to a standardized format, facilitating subsequent image processing and feature extraction, and improving the accuracy and reliability of analysis.

[0029] Specifically, in the embodiment of the present application, the profile section image correction module 120 is used to: pass the profile section image through an image distortion corrector based on an automatic codec to obtain a corrected profile section image.

[0030] More specifically, the image of the profile cross section is corrected using an automatic codec-based image distortion corrector. Considering that various factors may cause image distortion and blurring during the production process, distortion correction can improve image quality, making subsequent processing more accurate and reliable. The corrected image better meets standardized image requirements, facilitating subsequent feature extraction, analysis, and processing, ensuring data consistency and comparability. The corrected image eliminates errors and interference caused by distortion, improving the accuracy and stability of subsequent algorithms and thus better supporting the functions of the production management system. Using an automatic codec for image distortion correction enables automated processing, reduces manual intervention, and enhances the intelligence of the production management system.

[0031] Specifically, in an embodiment of the present application, the profile section image correction module 120 includes: a profile section image encoding unit, used to input the image of the profile section into the encoder of the image distortion correction, wherein the encoder uses a convolution layer to perform explicit spatial encoding on the image of the profile section to obtain a profile section feature map; and a profile section feature decoding unit, used to input the profile section feature map into the decoder of the image distortion correction, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the profile section feature map to obtain the corrected profile section image.

[0032] In the aforementioned computer-based plastic-steel door and window production management system 100, the profile section feature analysis module 130 is used to analyze the corrected profile section image to obtain profile section-related feature vectors. Corrected profile section images may contain a wealth of information, and feature extraction can transform this information into more representative feature vectors, helping the system better understand and process image data. By analyzing these feature vectors, the system can assess the quality and accuracy of profile cutting, helping production managers identify problems and make adjustments promptly.

[0033] Figure 2 The block diagram of the profile section feature analysis module in the computer-based plastic-steel door and window production management system according to an embodiment of the present application is shown. The profile section feature analysis module 130 includes: a directional gradient histogram analysis unit 131, which is used to extract a profile section directional gradient histogram from the corrected profile section image and analyze it to obtain a spatially enhanced profile section feature map; and a profile section correlation feature analysis unit 132, which is used to divide the spatially enhanced profile section feature map into multiple spatially enhanced profile section sub-feature maps and analyze them to obtain a profile section correlation feature vector.

[0034] Figure 3 The block diagram of the directional gradient histogram analysis unit in the computer-based plastic-steel door and window production management system according to an embodiment of the present application is shown. The directional gradient histogram analysis unit 131 includes: a profile section directional gradient histogram extraction subunit 11, which is used to extract a profile section directional gradient histogram from the corrected profile section image; a profile section multi-channel aggregation subunit 12, which is used to aggregate the corrected profile section image and the profile section directional gradient histogram along the channel dimension to obtain a multi-channel profile section input image; and a profile section spatial feature enhancement subunit 13, which is used to pass the multi-channel profile section input image through a profile section spatial feature extraction module to obtain a spatially enhanced profile section feature map.

[0035] More specifically, a histogram of directional gradients of the profile cross section is extracted from the corrected profile cross section image. This histogram can help extract important features from the profile cross section image, including edges, texture, and other information, facilitating a deeper analysis and understanding of the profile cross section. The histogram of directional gradients can be used to describe the shape and structural characteristics of the profile cross section, providing a foundation for subsequent shape recognition and analysis. Analysis of the histogram of directional gradients can detect defects, flaws, or anomalies in the profile cross section image, enabling quality control and anomaly detection, and helping to identify and address problems promptly.

[0036] More specifically, the rectified profile cross-section image and the profile cross-section directional gradient histogram are aggregated along the channel dimension to produce a multi-channel profile cross-section input image. By aggregating the rectified profile cross-section image and the directional gradient histogram along the channel dimension, features from different information sources can be integrated into a single multi-channel image, enriching the image's information representation and improving the model's understanding of the profile cross-section. Multi-channel images can fuse different feature information, helping to extract richer and more diverse feature representations, enhancing the model's ability to learn features of the profile cross-section and improving classification and recognition accuracy. Furthermore, information from different channels can complement and reinforce each other, improving a comprehensive understanding of the profile cross-section, facilitating the discovery of potential connections and patterns, and enhancing data utilization efficiency. Combining multiple single-channel images into a single multi-channel image helps mitigate the impact of the curse of data dimensionality, reducing the computational complexity and number of parameters in model training, and improving model efficiency and generalization. Multi-channel input images better meet the input requirements of deep learning models, facilitate model training and optimization, and enhance model performance and stability.

[0037] Specifically, in an embodiment of the present application, the profile section spatial feature extraction module is a convolutional neural network model using a spatial attention mechanism.

[0038] More specifically, a multi-channel cross-section image is input to a convolutional neural network model using a spatial attention mechanism to generate a spatially enhanced cross-section feature map. The spatial attention mechanism helps the model focus on specific regions of the input image, improving the recognition and utilization of key features, thereby enhancing the model's representational capabilities. Through the spatial attention mechanism, the model learns the importance weights of different regions in the input image, specifically enhancing the representation of key features, and improving feature discrimination and expressiveness. The spatial attention mechanism helps the model effectively integrate information from different channels and spatial locations, improving the comprehensive utilization of global and local features in the cross-section image and enhancing the model's perception capabilities. Furthermore, the spatial attention mechanism dynamically adjusts attention weights based on the specific task and input data, enabling adaptive focus on different features and increasing the model's flexibility and generalization capabilities. The introduction of the spatial attention mechanism improves the model's performance in cross-section feature extraction and classification tasks, deepens the model's understanding and abstraction of cross-section images, and enhances the model's accuracy and robustness.

[0039] Specifically, in an embodiment of the present application, the profile section spatial feature enhancement subunit 13 is used to: use each layer of the profile section spatial feature extraction module to perform the following on the input data in the forward pass of the layer: convolve the input data based on the convolution kernel to obtain a convolution feature map; pass the convolution feature map through the spatial attention module to obtain the spatial attention score matrix; multiply the spatial attention score matrix and each feature matrix of the convolution feature map along the channel dimension by the spatial attention feature map at each position point; perform pooling processing on the spatial attention feature map based on the local feature matrix to obtain a pooled feature map; and perform nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the input of the first layer of the profile section spatial feature extraction module is the multi-channel profile section input image, and the output of the last layer of the profile section spatial feature extraction module is the spatially enhanced profile section feature map.

[0040] Specifically, in an embodiment of the present application, the profile section association feature analysis unit 132 includes: a spatially enhanced profile section feature map segmentation subunit, used to segment the spatially enhanced profile section feature map into multiple spatially enhanced profile section sub-feature maps; a profile section depth feature extraction subunit, used to pass the multiple spatially enhanced profile section sub-feature maps through a profile section depth feature extraction module to obtain multiple profile section depth fusion feature vectors; and a profile section feature association subunit, used to pass the multiple profile section depth fusion feature vectors through a converter-based profile section semantic encoding module to obtain a profile section association feature vector.

[0041] More specifically, the spatially enhanced profile cross-section feature map is divided into multiple spatially enhanced profile cross-section sub-feature maps. First, dividing the feature map into multiple sub-feature maps can help the model better capture local features in the profile cross-section image, which helps improve the model's perception of detail information. At the same time, by dividing the feature map, the size of each sub-feature map can be reduced, thereby reducing the amount of computational effort for subsequent processing and improving the model's operating efficiency. Each sub-feature map can focus more on the feature representation of the local area, which is beneficial to improving the feature's distinctiveness and expressiveness, thereby improving the model's classification and recognition performance. Multiple sub-feature maps can provide more sources of information, help the model's resistance to noise and interference, and improve the model's robustness and generalization capabilities. At the same time, each sub-feature map can contain specific semantic information. By combining information from different sub-feature maps, the complex feature associations in the profile cross-section image can be better captured, thereby improving the model's representation capabilities.

[0042] More specifically, a plurality of spatially enhanced profile section sub-feature maps are processed separately by the profile section depth feature extraction module to obtain a plurality of profile section depth fusion feature vectors. The profile section depth feature extraction module is a depth feature extraction model. In the technical solution of the present application, different spatially enhanced profile section sub-feature maps may capture different aspects and detail information of the profile section, and richer features can be extracted by processing these sub-feature maps separately. Processing different sub-feature maps separately can enable the model to better distinguish features at different spatial positions, avoid information confusion, and help improve feature recognition. By processing multiple sub-feature maps through the profile section depth feature extraction module, the information in different sub-feature maps can be fused together to obtain a more comprehensive profile section feature representation. There is a spatial relationship between different spatially enhanced profile section sub-feature maps, and processing these sub-feature maps separately can better capture the spatial relationship and mutual influence between different areas.

[0043] Specifically, in an embodiment of the present application, the profile section depth feature extraction subunit includes: a secondary subunit for shallow profile section feature extraction, for extracting a profile section shallow feature map from the Mth layer of the profile section depth feature extraction module, wherein M is greater than or equal to 1 and less than or equal to 6; a secondary subunit for deep profile section feature extraction, for extracting a profile section deep feature map from the Nth layer of the profile section depth feature extraction module, wherein N / M is greater than or equal to 5 and less than or equal to 10; a secondary subunit for profile section depth feature fusion, for using the profile section depth feature extraction module to fuse the profile section shallow feature map and the profile section deep feature map to obtain a profile section fusion feature map; and a secondary subunit for dimensionality reduction, for performing global pooling processing on the local feature matrix along the channel dimension of the profile section fusion feature map to obtain the multiple profile section depth fusion feature vectors.

[0044] Specifically, in an embodiment of the present application, the profile section semantic encoding module is a converter-based context encoder model.

[0045] More specifically, multiple profile cross-section deep and shallow fusion feature vectors are processed through a transformer-based context encoder model to obtain profile cross-section association feature vectors. The transformer-based context encoder model can effectively capture global dependencies and contextual information in text or images, and helps understand the associations between profile cross-section deep and shallow fusion feature vectors. The transformer model is suitable for processing sequential data and can effectively encode and understand input sequences, thereby better capturing the sequential relationships between profile cross-section deep and shallow fusion feature vectors. The transformer model can achieve interaction and fusion of global information, which helps the model better understand the non-local associations between profile cross-section deep and shallow fusion feature vectors and improve the feature representation capability. The attention mechanism in the transformer model can effectively learn the importance weights between different features, helping the model focus on feature information that is more critical to the current task and improving feature utilization efficiency. The transformer-based context encoder model has strong flexibility and scalability, can adapt to different types of input data and complex associations, and is conducive to improving the performance of the model in profile cross-section image processing tasks. At the same time, by processing the deep and shallow fusion feature vectors of multiple profile sections through the converter-based context encoder model, global interaction and fusion of features can be achieved, thereby obtaining more representative and rich profile section association feature vectors.

[0046] Specifically, in an embodiment of the present application, the profile section feature association subunit includes: a query vector construction secondary subunit, which is used to arrange the multiple profile section depth and shallow fusion feature vectors into an input vector; a vector conversion secondary subunit, which is used to convert the input vector into a query vector and a key vector respectively through a learnable embedding matrix; a self-attention secondary subunit, which is used to calculate the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; a normalization secondary subunit, which is used to normalize the self-attention association matrix to obtain a standardized self-attention association matrix; an attention calculation secondary subunit, which is used to input the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and an attention application secondary subunit, which is used to multiply the self-attention feature matrix with each profile section depth and shallow fusion feature vector in the multiple profile section depth and shallow fusion feature vectors to obtain the multiple profile section semantic feature vectors; and cascade the multiple profile section semantic feature vectors to obtain a profile section association feature vector.

[0047] In the computer-based plastic-steel door and window production management system 100, the cutting effect analysis module 140 is used to analyze the profile section-related feature vectors to obtain a result indicating whether the profile cutting effect is qualified.

[0048] Specifically, in the embodiment of the present application, the cutting effect analysis module 140 includes: a profile section-related feature compensation unit, which is used to perform kernel space projection reorganization of the profile section-related feature vector based on basis function regression to obtain a compensated profile section-related feature vector; and a cutting effect analysis unit, which is used to pass the compensated profile section-related feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the profile cutting effect is qualified.

[0049] In particular, in the feature extraction stage, the Histogram of Oriented Gradients (HOG) is mainly used to capture the directional information of edges and lines. For certain types of profile sections, this feature description method may not be sufficient to fully characterize their internal structural features, especially when the profile has a complex texture or shape. In addition, in the process of spatial feature enhancement and depth feature extraction, if the model or algorithm adopted fails to accurately adapt to the specific characteristics of the profile section, it may also lead to the neglect of internal structural information. Furthermore, although the converter-based semantic encoding module can effectively capture long-range dependencies, if the training data lacks sufficient diversity or the model parameters are improperly set, it will also affect the quality of the final profile section-related feature vector, thereby causing the problem of insufficient utilization of the feature internal structural information. Therefore, in order to overcome this problem, the profile section-related feature vector is subjected to kernel space projection reconstruction based on basis function regression to obtain the compensated profile section-related feature vector.

[0050] Specifically, in the embodiment of the present application, the profile section associated feature compensation unit is used to: first, construct a pixel coupling matrix of the profile section associated feature vector, which is expressed as:

[0051]

[0052] Wherein, V represents the profile section associated feature vector, v i and v j Respectively represent the eigenvalues of the i-th and j-th positions of the profile section associated eigenvector, d(v i ,v j ) represents the calculation of Euclidean distance, D i,j Represents the eigenvalue at position (i,j) of the pixel coupling matrix.

[0053] Specifically, by quantifying the response coordination and pattern consistency between pixels, isolated image feature points are transformed into a dynamic feature field with contextual associations. The resulting pixel coupling matrix not only enhances the ability to characterize cross-regional morphological features of cut surface defects, but also, through a pixel-level relational topological network, provides an interpretable feature interaction path for subsequent convolution kernel analysis and processing.

[0054] Secondly, the pixel coupling matrix is mined based on the convolution layer to obtain the nonlinear activation matrix of the profile cross-section correlation kernel space, which is expressed as follows:

[0055] M=Conv(D)

[0056] Wherein, D represents the pixel coupling matrix, Conv represents the convolutional layer, and M represents the nonlinear activation matrix of the profile section correlation kernel space.

[0057] Specifically, the pixel coupling matrix is converted into a physically interpretable kernel space feature field. Through an adaptive learning mechanism for convolution kernel weights, multi-order correlation feature patterns closely related to cutting quality judgment are extracted while preserving the spatial topological relationships of the pixel coupling matrix. This kernel space mapping approach establishes an implicit mathematical relationship between the correlation intensity distribution and the cutting process parameters. The resulting nonlinear activation matrix for the profile cross-section correlation kernel space selectively enhances the correlation features, providing a high-order feature substrate with process mechanism relevance for subsequent quality judgment.

[0058] Then, the pixel coupling matrix is decomposed in the spectral domain to obtain a set of profile section correlation basis function feature coding vectors, which can be expressed as follows:

[0059]

[0060] Where T represents the transpose of the vector, Λ represents the diagonal matrix, λ1 and λ m They represent the first and mth eigenvalues of the diagonal matrix respectively, U represents the set of characteristic coding vectors of the profile section correlation basis function, x1, x2, x m Represent the first, second and mth profile section correlation basis function feature coding vectors respectively.

[0061] That is, the high-dimensional pixel coupling matrix is mapped to the orthogonal basis function space, and by extracting the eigenmode clusters that characterize the key factors of cutting quality, an interpretable defect feature atom library is constructed. It should be understood that the basis function encodes the implicit mapping relationship between the geometric accuracy of the cutting surface and the production process parameters. It can not only decouple the coupling interference between environmental noise and real process defects in the correlation matrix, but also, through the linear combination characteristics of the orthogonal basis vectors, transform the unstructured correlation matrix into a dynamic feature field with process physical significance, that is, a collection of profile section correlation basis function feature encoding vectors, providing a traceable pattern separation basis for subsequent dynamic weight allocation.

[0062] Next, each profile section-related basis function feature encoding vector in the set of profile section-related basis function feature encoding vectors is input into the transformer model based on the self-attention mechanism to obtain a set of profile section-related core attention encoding vectors, which is expressed as follows:

[0063] Y=Transformer{[x1,x2,…,x m ]}=[y1,y2,…,y m ]

[0064] Among them, Transformer represents a sequence model based on the self-attention mechanism, Y represents the set of profile section associated core attention encoding vectors, y1, y2, y m Represent the first, second and m-th profile slice associated kernel attention encoding vectors respectively.

[0065] Specifically, a dynamic response channel is constructed between process parameters and feature saliency. By real-time evaluating the correlation entropy between basis function patterns and current cutting quality indicators, interpretable defect feature modulation rules are established. This dynamic enhancement mechanism not only automatically adjusts the basis function attention distribution based on the real-time acquired profile mechanical parameters, but also suppresses pseudo-basis function responses triggered by environmental noise. This ensures that the feature enhancement process is both data-driven and controlled by physical process boundary conditions. This generates a collection of profile cross-section correlation kernel attention encoding vectors, providing a robust feature basis for defect classification under complex working conditions.

[0066] Then, each profile section associated core attention coding vector in the set of the profile section associated core attention coding vectors is projected onto the profile section associated core space nonlinear activation matrix to obtain a set of profile section associated core masked coding vectors, which is expressed as follows:

[0067]

[0068] in, represents matrix multiplication, S represents the characteristic scale of the nonlinear activation matrix of the profile cross-section correlation kernel space, y irepresents the i-th profile section associated core attention encoding vector, L represents the length of the profile section associated core attention encoding vector, z i represents the kernel masked encoding vector associated with the i-th profile slice.

[0069] Specifically, a multi-scale feature field with dual process-material driving characteristics is constructed, and masked modulation of the profile section-related kernel attention encoding vectors is performed via a kernel space nonlinear matrix to achieve controllable generation of cutting quality judgment features. This masking mechanism not only physically associates the tool spindle vibration frequency with the material stress concentration area, but also suppresses pseudo-process defect signals caused by ambient temperature and humidity fluctuations through nonlinear interaction. This allows the generated collection of profile section-related kernel masked encoding vectors to be traced back to physical parameter anomalies in the mechanical transmission system and to anchor the critical failure point of the material microstructure.

[0070] Finally, the set of profile section-related kernel masked coding vectors is fused to obtain the compensated profile section-related feature vector, which is expressed as follows:

[0071] V'=Concat{z1,z2,…,z m}

[0072] Among them, Concat represents the cascade function, z1, z2, z m They represent the first, second and mth profile section associated kernel masked coding vectors respectively, and V' represents the profile section associated feature vector after compensation.

[0073] That is, a feature hyperspace with process parameter traceability is constructed, and the compensated profile section associated feature vector is generated by cascading each profile section associated core masked coding vector in the set of profile section associated core masked coding vectors.

[0074] More specifically, the feature vectors associated with the compensated profile sections are passed through a classifier to obtain a classification result, which is used to determine whether the profile cutting effect is satisfactory. The classifier processes the feature vectors associated with the compensated profile sections, transforming the abstract feature representation into a concrete classification result, helping to assess whether the profile cutting effect meets expectations. Furthermore, using the classifier to classify the feature vectors associated with the compensated profile sections enables automated assessment of the profile cutting effect, improving efficiency and accuracy.

[0075] Specifically, in the embodiment of the present application, the cutting effect analysis module 140 is used to: use the classifier to process the associated feature vector of the compensated profile section using the following classification formula to obtain the classification result; wherein the classification formula is:

[0076] O=softmax{(W n ,Bn ):…:(W1,B1)|V′}

[0077] Among them, W1 to W n is the weight matrix, B1 to B n is the bias vector, V' is the associated feature vector of the profile section after compensation, softmax represents the softmax function, and O represents the classification result.

[0078] In summary, the computer-based plastic-steel door and window production management system according to the embodiments of the present application has been described. It uses artificial intelligence technology based on deep learning to extract and encode features from images of profile sections, thereby obtaining a classification result that determines whether the profile cutting effect is acceptable. This intelligent judgment of profile cutting results improves the precision of profile cutting inspection, reduces labor costs, avoids omissions in inspection results, and ensures the quality of profile cutting.

[0079] As described above, the computer-based plastic-steel door and window production management system 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a computer-based plastic-steel door and window production management server. In one example, the computer-based plastic-steel door and window production management system 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or hardware module. For example, the computer-based plastic-steel door and window production management system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the computer-based plastic-steel door and window production management system 100 can also be one of the many hardware modules of the terminal device.

[0080] Alternatively, in another example, the computer-based plastic-steel door and window production management system 100 and the terminal device may also be separate devices, and the computer-based plastic-steel door and window production management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0081] Based on the same inventive concept, the embodiments of the present application also provide a computer-based plastic-steel door and window production management method, which can be used to implement the system described in the above embodiments, as described in the following embodiments.

[0082] Figure 4 FIG is a flowchart of a computer-based plastic-steel door and window production management method according to an embodiment of the present application. Figure 4As shown, the computer-based plastic-steel door and window production management method according to the embodiment of the present application includes the steps of: S110, acquiring an image of a profile cross section; S120, correcting the image of the profile cross section to obtain a corrected profile cross section image; S130, analyzing the corrected profile cross section image to obtain a profile cross section associated feature vector; and S140, analyzing the profile cross section associated feature vector to obtain a result indicating whether the profile cutting effect is qualified.

[0083] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. Here, for the computer-based plastic-steel door and window production management method disclosed in the embodiment, those skilled in the art will understand that the specific operations of each step in the above-mentioned computer-based plastic-steel door and window production management method have been referred to above. Figures 1 to 3 It has been introduced in detail in the description of the computer-based plastic steel door and window production management system, so the description is relatively simple. For relevant details, please refer to the description of the computer-based plastic steel door and window production management part, and therefore, its repeated description will be omitted.

[0084] In summary, the computer-based production management method for plastic-steel doors and windows according to the embodiments of this application has been described. This method uses deep learning-based artificial intelligence technology to extract and encode features from images of profile sections, thereby classifying whether the profile cutting effect is acceptable. This intelligent judgment of profile cutting results improves the precision of profile cutting inspection, reduces labor costs, avoids omissions in inspection results, and ensures the quality of profile cutting.

[0085] In summary, after reading this detailed disclosure, those skilled in the art will appreciate that the foregoing detailed disclosure may be presented by way of example only and may not be limiting. Although not explicitly stated herein, those skilled in the art will understand that this application is intended to encompass various reasonable changes, improvements, and modifications to the embodiments. Such changes, improvements, and modifications are intended to be proposed by this application and are within the spirit and scope of the exemplary embodiments of this application.

[0086] In addition, certain terms in this application have been used to describe embodiments of the present application. For example, "one embodiment," "an embodiment," and / or "some embodiments" mean that a particular feature, structure, or characteristic described in conjunction with that embodiment may be included in at least one embodiment of the present application. Therefore, it is emphasized and should be understood that two or more references to "an embodiment," "one embodiment," or "an alternative embodiment" in various parts of this specification do not necessarily refer to the same embodiment. Furthermore, particular features, structures, or characteristics may be appropriately combined in one or more embodiments of the present application.

[0087] Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that an article or device comprising a list of elements includes not only those elements but also other elements not explicitly listed or inherent to such article or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the article or device comprising the aforementioned elements.

[0088] It should be understood that in the foregoing description of the embodiments of this application, in order to facilitate understanding of a feature and to simplify this application, this application combines various features into a single embodiment, figure, or description thereof. However, this does not mean that the combination of these features is required. When reading this application, it is entirely possible for those skilled in the art to extract some of the features and understand them as separate embodiments. In other words, the embodiments of this application can also be understood as the integration of multiple secondary embodiments. This also applies when the content of each secondary embodiment is less than all the features of a single aforementioned disclosed embodiment.

[0089] Finally, it should be understood that the embodiments of the application disclosed herein are illustrations of the principles of the embodiments of the present application. Other modified embodiments are also within the scope of the present application. Therefore, the embodiments disclosed in this application are merely examples and not limitations. Those skilled in the art can adopt alternative configurations to implement the application in this application based on the embodiments in this application.

[0090] Therefore, the embodiments of the present application are not limited to the precise embodiments described in the application.

Claims

1. A computer-based plastic-steel door and window production management system, characterized in that: include: A profile cross-section image acquisition module is used to acquire images of the profile cross-section; A profile section image correction module, configured to correct the profile section image to obtain a corrected profile section image; A profile section feature analysis module, configured to analyze the corrected profile section image to obtain a profile section associated feature vector; The cutting effect analysis module is used to analyze the profile section-related feature vectors to obtain a result indicating whether the profile cutting effect is qualified.

2. The computer-based plastic-steel door and window production management system according to claim 1, characterized in that: The profile section image correction module is used to: The profile section image is passed through an image distortion corrector based on an automatic codec to obtain a corrected profile section image.

3. The computer-based plastic-steel door and window production management system according to claim 2, characterized in that: The profile section feature analysis module includes: a directional gradient histogram analysis unit, configured to extract a profile section directional gradient histogram from the corrected profile section image and then analyze the extracted directional gradient histogram to obtain a spatially enhanced profile section feature map; The profile section correlation feature analysis unit is used to divide the spatially enhanced profile section feature map into multiple spatially enhanced profile section sub-feature maps and then analyze them to obtain a profile section correlation feature vector.

4. The computer-based plastic-steel door and window production management system according to claim 3, characterized in that: The directional gradient histogram analysis unit comprises: a profile section directional gradient histogram extraction subunit, configured to extract a profile section directional gradient histogram from the corrected profile section image; a profile section multi-channel aggregation subunit, configured to aggregate the corrected profile section image and the profile section directional gradient histogram along a channel dimension to obtain a multi-channel profile section input image; The profile section spatial feature enhancement subunit is used to pass the multi-channel profile section input image through the profile section spatial feature extraction module to obtain a spatially enhanced profile section feature map.

5. The computer-based plastic-steel door and window production management system according to claim 4, characterized in that: The profile section spatial feature extraction module is a convolutional neural network model that uses a spatial attention mechanism.

6. The computer-based plastic-steel door and window production management system according to claim 5, characterized in that: The profile section correlation feature analysis unit includes: A spatially enhanced profile section feature graph slicing subunit, configured to slice the spatially enhanced profile section feature graph into a plurality of spatially enhanced profile section sub-feature graphs; A profile section depth feature extraction subunit, configured to pass the plurality of spatially enhanced profile section sub-feature maps through a profile section depth feature extraction module to obtain a plurality of profile section depth fusion feature vectors; The profile section feature association subunit is used to pass the multiple profile section depth and shallow fusion feature vectors through the converter-based profile section semantic encoding module to obtain a profile section association feature vector.

7. The computer-based plastic-steel door and window production management system according to claim 6, characterized in that: The cutting effect analysis module includes: a profile section-related feature compensation unit, configured to perform kernel space projection reorganization based on basis function regression on the profile section-related feature vector to obtain a compensated profile section-related feature vector; The cutting effect analysis unit is used to pass the compensated profile section associated feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the profile cutting effect is qualified.

8. The computer-based plastic-steel door and window production management system according to claim 7, characterized in that: The profile section associated feature compensation unit is used to: Constructing a pixel coupling matrix of the profile section-related feature vectors; Performing kernel feature mining on the pixel coupling matrix based on a convolutional layer to obtain a profile section correlation kernel space nonlinear activation matrix; Performing spectral domain decomposition on the pixel coupling matrix to obtain a set of profile section correlation basis function feature coding vectors; Inputting each profile section-related basis function feature encoding vector in the set of profile section-related basis function feature encoding vectors into a transformer model based on a self-attention mechanism to obtain a set of profile section-related core attention encoding vectors; Projecting each profile section associated core attention coding vector in the set of the profile section associated core attention coding vectors onto the profile section associated core space nonlinear activation matrix to obtain a set of profile section associated core masked coding vectors; The set of the profile section associated kernel masked coding vectors is fused to obtain the compensated profile section associated feature vector.

9. A computer-based production management method for plastic-steel doors and windows, characterized in that: include: Acquire an image of the cross section of the profile; Correcting the image of the profile section to obtain a corrected profile section image; Analyzing the corrected profile section image to obtain a profile section correlation feature vector; The profile section associated feature vector is analyzed to obtain a result indicating whether the profile cutting effect is qualified.

Images