LCD System Fault Detection System and Method Based on Image Analysis

Abstract

This application relates to the field of intelligent detection, and provides an LCD system fault detection system and method based on image analysis. First, it acquires the LCD panel image collected by a high-resolution industrial camera, then performs homomorphic filtering and CLAHE contrast enhancement processing on the image to obtain an enhanced LCD panel image, then conducts multi-scale feature enhancement fusion on the enhanced image to obtain a multi-scale fusion feature map of the LCD panel image, then conducts statistical analysis on this feature map based on the Mura segmentation region to obtain a Mura region statistical feature coding vector, and finally, based on this coding vector, obtains the Mura defect recognition result. In this way, it can more accurately capture the characteristics of Mura defects, reduce false judgments and missed detections caused by interference factors such as light sources and noise, improve the detection accuracy of Mura defects, and at the same time, the detection system can adapt to different production conditions and ensure the stability of the detection results.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to an LCD system fault detection system and method based on image analysis. Background Art

[0002] With the widespread adoption of liquid crystal display (LCD) technology in consumer electronics, industrial control, and other fields, quality inspection during the manufacturing process has become a critical step in ensuring product performance. Among the many types of LCD defects, mura (moire) presents a significant detection challenge due to its unique visual appearance. This defect manifests as cloud-like patterns with uneven brightness across the display area. Its causes involve multiple manufacturing processes, including abnormalities in the backlight assembly and disordered liquid crystal molecular alignment.

[0003] Traditional manual visual inspection methods rely on subjective judgment based on engineers' experience, resulting in inherent drawbacks such as low detection efficiency and inconsistent standards. This is particularly true for low-contrast, gradient-type mura. The human eye's threshold for brightness differences is easily affected by external factors such as ambient lighting and viewing angle, resulting in high rates of missed and false positives. Despite recent advances in automated inspection technology based on machine vision, practical applications still face multiple technical bottlenecks. During the image acquisition phase, abnormal brightness distribution caused by insufficient light source uniformity can easily be confused with actual defects, while camera noise and surface reflections significantly impact the image signal-to-noise ratio. During image processing, traditional algorithms lack sensitivity to the gradient characteristics of mura, making it difficult to accurately capture low-brightness defect areas using threshold segmentation. Simple edge detection operators are also unable to effectively distinguish artifacts caused by uneven backlighting from actual defects. More critically, existing methods lack the ability to integrate and analyze multi-scale defect features and fail to establish an objective, quantitative evaluation system consistent with subjective perception. This makes it difficult to ensure the robustness of inspection systems across varying production batches and environmental parameters.

[0004] Therefore, an optimized LCD system fault detection solution based on image analysis is desired. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the present application provides an LCD system fault detection system and method based on image analysis.

[0006] According to one aspect of the present application, there is provided a method for detecting faults in an LCD system based on image analysis, which includes: acquiring an LCD panel image collected by a high-resolution industrial camera; performing homomorphic filtering and CLAHE contrast enhancement on the LCD panel image to obtain an enhanced LCD panel image; performing multi-scale feature enhancement fusion on the enhanced LCD panel image to obtain a multi-scale fusion feature map of the LCD panel image, including: performing wavelet decomposition and feature extraction on the enhanced LCD panel image to obtain a low-frequency approximation sub-band feature map, a horizontal direction detail sub-band feature map, a vertical direction detail sub-band feature map, and a diagonal direction detail sub-band feature map; performing feature enhancement fusion processing based on the spatial-semantic two-dimensional dimension on the low-frequency approximation sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain the multi-scale fusion feature map of the LCD panel image; performing statistical analysis based on the Mura segmentation region on the multi-scale fusion feature map of the LCD panel image to obtain a Mura region statistical feature coding vector; and obtaining a Mura defect recognition result based on the Mura region statistical feature coding vector.

[0007] According to another aspect of the present application, there is provided a system for detecting faults in an LCD system based on image analysis, which includes: an image acquisition module for acquiring an LCD panel image collected by a high-resolution industrial camera; an image preprocessing module for performing homomorphic filtering and CLAHE contrast enhancement on the LCD panel image to obtain an enhanced LCD panel image; an image coding module for performing multi-scale feature enhancement fusion on the enhanced LCD panel image to obtain a multi-scale fusion feature map of the LCD panel image, wherein the image coding module is configured to: perform wavelet decomposition and feature extraction on the enhanced LCD panel image to obtain a low-frequency approximation sub-band feature map, a horizontal direction detail sub-band feature map, a vertical direction detail sub-band feature map, and a diagonal direction detail sub-band feature map; perform feature enhancement fusion processing based on the spatial-semantic two-dimensional dimension on the low-frequency approximation sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain the multi-scale fusion feature map of the LCD panel image; a Mura region analysis module for performing statistical analysis based on the Mura segmentation region on the multi-scale fusion feature map of the LCD panel image to obtain a Mura region statistical feature coding vector; and a defect recognition module for obtaining a Mura defect recognition result based on the Mura region statistical feature coding vector.

[0008] Due to the adoption of the above technical solutions, this application has remarkable technical effects: The LCD system fault detection system and method based on image analysis provided by this application first acquires the LCD panel image collected by a high-resolution industrial camera, then performs homomorphic filtering and CLAHE contrast enhancement processing on this image to obtain an enhanced LCD panel image, then conducts multi-scale feature enhancement fusion on the enhanced image to obtain a multi-scale fusion feature map of the LCD panel image, then performs statistical analysis on this feature map based on the Mura segmentation region to obtain a Mura region statistical feature coding vector, and finally, based on this coding vector, obtains the Mura defect recognition result. In this way, it can capture the features of Mura defects more accurately, reduce false judgments and missed detections caused by interference factors such as light sources and noises, improve the detection accuracy of Mura defects, and at the same time, the detection system can adapt to different production conditions and ensure the stability of the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] 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 obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of 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.

[0010] Figure 1 It is a flowchart of the method for detecting faults in the LCD system based on image analysis according to an embodiment of the present application.

[0011] Figure 2 It is a flowchart of step S3 in the method for detecting faults in the LCD system based on image analysis according to an embodiment of the present application.

[0012] Figure 3 It is a flowchart of step S31 in the method for detecting faults in the LCD system based on image analysis according to an embodiment of the present application.

[0013] Figure 4 It is a flowchart of step S32 in the method for detecting faults in the LCD system based on image analysis according to an embodiment of the present application.

[0014] Figure It is a flowchart of step S321 in the method for detecting faults in the LCD system based on image analysis according to an embodiment of the present application.

[0015] ​ It is a flowchart of step S321-3 in the method for detecting faults in the LCD system based on image analysis according to an embodiment of the present application.

[0016] ​It is a flowchart of step S4 in the method for detecting faults in an LCD system based on image analysis according to an embodiment of the present application.

[0017] ​ It is a system block diagram of a fault detection system for an LCD system based on image analysis according to an embodiment of the present application. Detailed implementation manners

[0018] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0019] Based on this, the technical concept of the present application is to construct a full-process detection system for Mura defects on the LCD panel through a multi-level image enhancement and multi-scale feature fusion framework. Specifically, after obtaining the original image using a high-resolution camera, the illumination component and the reflection component are separated by homomorphic filtering, and the local contrast is enhanced by combining the CLAHE algorithm, effectively overcoming the uneven background illumination and surface reflection interference; subsequently, the image is decomposed into different frequency bands and directional features, and defect features with a larger receptive field are extracted to enhance the sensitivity to low-contrast gradual defects; then, through multi-scale feature fusion and semantic segmentation, the boundary distribution of different morphological Mura defects is accurately located; finally, statistical features such as texture and brightness distribution of the defect area are extracted, and a defect quantification model is established to achieve an objective evaluation consistent with the subjective perception of the human eye. This solution suppresses background noise interference through frequency-domain to spatial-domain joint enhancement technology, improves the detection ability of tiny defects using multi-resolution analysis, and solves the problem that traditional algorithms lack a quantification standard for Mura defects by statistical feature modeling, significantly improving the adaptability of the system to complex industrial environments while ensuring detection sensitivity.

[0020] ​ It is a flowchart of the method for detecting faults in an LCD system based on image analysis according to an embodiment of the present application. As ​ shown, the method for detecting faults in an LCD system based on image analysis according to an embodiment of the present application includes: S1, obtaining an LCD panel image collected by a high-resolution industrial camera; S2, performing homomorphic filtering and CLAHE contrast enhancement on the LCD panel image to obtain an enhanced LCD panel image; S3, performing multi-scale feature enhancement fusion on the enhanced LCD panel image to obtain a multi-scale fusion feature map of the LCD panel image; S4, performing statistical analysis on the multi-scale fusion feature map of the LCD panel image based on the Mura segmentation region to obtain a Mura region statistical feature coding vector; S5, obtaining a Mura defect recognition result based on the Mura region statistical feature coding vector.

[0021] In step S1, an LCD panel image collected by a high-resolution industrial camera is obtained. It should be understood that the information in the LCD panel image is essentially a digital mapping of the panel's optical performance. In the spatial dimension, the gray value of each pixel accurately reflects the brightness intensity of the corresponding area, and this brightness distribution characteristic is directly related to the uniformity of the liquid crystal molecule arrangement and the optical diffusion efficiency of the backlight module. For example, when abnormal fluctuations occur in the gray values of a local area, it may imply changes in the light transmittance caused by uneven liquid crystal layer thickness or abnormal driving voltage, which are typical manifestations of Mura defects. At the level of texture features, the high-resolution image can clearly present the periodic arrangement structure of the TFT array. When this regularity is disrupted, it may mean pixel misalignment caused by packaging process defects or external force damage. Such texture anomalies often accompany specific forms of Mura defects. Generally speaking, by deeply analyzing the information in the obtained LCD panel image, Mura defects can be effectively identified.

[0022] In step S2, the LCD panel image is subjected to homomorphic filtering and CLAHE contrast enhancement to obtain an enhanced LCD panel image. Correspondingly, considering that the uneven illumination and surface reflection interference in the image acquisition stage will significantly affect the recognition rate of defects. Since it is difficult to achieve absolute uniformity in the light source distribution of the industrial detection environment, the images captured by the camera often contain low-frequency brightness fluctuations caused by the luminance difference of the backlight module or external environmental reflection. Such interference appears as a large-area gradual brightness shift in the spatial domain, and its frequency characteristics overlap with the low-frequency gradual characteristics of real Mura defects. In addition, the relative position change between the high-reflectivity material on the LCD surface and the detection device is likely to generate specular reflection, resulting in abnormal saturation of local pixel values or the formation of highlight artifacts. These interference factors will mask the weak contrast changes of real defects, making it difficult for traditional threshold segmentation algorithms to effectively distinguish the defect area from optical artifacts. Therefore, in order to enhance the local contrast of the image and eliminate the influence of adverse factors such as uneven illumination, noise, and reflection, this application performs homomorphic filtering and CLAHE contrast enhancement on the LCD panel image to obtain an enhanced LCD panel image. That is, homomorphic filtering decomposes the image into an illumination component and a reflection component through frequency-domain filtering operations, compresses the dynamic range of the low-frequency global brightness distribution, and at the same time enhances the high-frequency detail features, effectively suppressing the low-frequency brightness drift caused by uneven backlighting. The CLAHE algorithm performs contrast-limited histogram equalization in local sub-regions, breaking through the limitation of traditional histogram equalization for global contrast adjustment, enhancing the local contrast of Mura defects while avoiding information distortion caused by over-enhancement in the specular reflection area. The combination of the two forms a dual enhancement mechanism of "global brightness correction - local contrast optimization": homomorphic filtering eliminates large-area background interference, and CLAHE enhances the gradient change at the defect edge, jointly improving the signal-to-noise ratio of low-contrast defects and the background.

[0023] In step S3, multi-scale feature enhancement fusion is performed on the enhanced LCD panel image to obtain a multi-scale fusion feature map of the LCD panel image. Specifically, ​ It is a flowchart of step S3 in the LCD system fault detection method based on image analysis according to an embodiment of the present application. As ​ shown, the step S3 includes: S31, performing wavelet decomposition and feature extraction on the enhanced LCD panel image to obtain a low-frequency approximation subband feature map, a horizontal direction detail subband feature map, a vertical direction detail subband feature map, and a diagonal direction detail subband feature map; S32, performing feature enhancement fusion processing based on the spatial-semantic two dimensions on the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map to obtain the multi-scale fusion feature map of the LCD panel image.

[0024] In step S31, wavelet decomposition and feature extraction are performed on the enhanced LCD panel image to obtain a low-frequency approximation subband feature map, a horizontal direction detail subband feature map, a vertical direction detail subband feature map, and a diagonal direction detail subband feature map. Specifically, ​ It is a flowchart of step S31 in the LCD system fault detection method based on image analysis according to an embodiment of the present application. As ​ shown, the step S31 includes: S311, performing wavelet decomposition on the enhanced LCD panel image to obtain a low-frequency approximation subband matrix, a horizontal direction detail subband matrix, a vertical direction detail subband matrix, and a diagonal direction detail subband matrix; S312, performing feature extraction based on dilated convolution coding on the low-frequency approximation subband matrix, the horizontal direction detail subband matrix, the vertical direction detail subband matrix, and the diagonal direction detail subband matrix to obtain the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map.

[0025] In step S311, wavelet decomposition is performed on the enhanced LCD panel image to obtain a low-frequency approximation subband matrix, a horizontal direction detail subband matrix, a vertical direction detail subband matrix, and a diagonal direction detail subband matrix. Correspondingly, considering that the morphological features of defects often exhibit multi-scale distribution characteristics. When traditional image processing methods directly process full-frequency domain information, the coupling of high-frequency noise and low-frequency background interference will lead to feature confusion. Especially for the gradual Mura with a contrast lower than the human eye recognition threshold, its weak brightness gradient change is easily submerged in the background texture. More critically, the diffuse edges of Mura defects have different responses in different spatial directions. For example, a horizontally extending strip-shaped defect may exhibit a stronger feature response in the vertical direction detail subband, while traditional single-scale feature extraction methods cannot effectively separate this direction-sensitive feature. Therefore, in order to be able to analyze and process the detail information in different directions more carefully, the present application performs wavelet decomposition on the enhanced LCD panel image to obtain a low-frequency approximation subband matrix, a horizontal direction detail subband matrix, a vertical direction detail subband matrix, and a diagonal direction detail subband matrix. In this way, the image is decomposed into a low-frequency approximation subband and three high-frequency detail subbands in the horizontal, vertical, and diagonal directions through orthogonal wavelet basis functions. The low-frequency approximation subband focuses on the overall brightness distribution of the image and can effectively characterize the low-frequency brightness gradual change feature of Mura defects; the horizontal detail subband strengthens the mutation response of vertical edges in the image and is suitable for detecting linear defects with horizontal extension characteristics; the vertical detail subband enhances the horizontal edge feature and has higher sensitivity to the vertically distributed Mura region; the diagonal detail subband simultaneously captures oblique texture changes and is used to identify irregularly shaped cloud-like defects. This multi-directional and multi-scale decomposition method breaks through the limitation of the single-dimensionality of feature dimensions in traditional spatial domain processing and provides a feature expression basis with physical significance for subsequent feature coding.

[0026] In step S312, feature extraction based on dilated convolution coding is performed on the low-frequency approximation subband matrix, the horizontal direction detail subband matrix, the vertical direction detail subband matrix, and the diagonal direction detail subband matrix to obtain the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map. Accordingly, considering that the low-frequency approximation subband matrix, the horizontal direction detail subband matrix, the vertical direction detail subband matrix, and the diagonal direction detail subband matrix respectively contain information about different scales and directions of the image. When the fixed receptive field of the conventional convolution kernel processes the low-frequency approximation subband, it may not be able to capture the Mura region formed by the large-range brightness gradient due to insufficient coverage; while in the high-frequency detail subbands, the too-small convolution window is prone to over-focusing on local noise and ignoring the continuity features of the defect edges. In addition, the morphological differences of Mura defects in the horizontal, vertical, and diagonal directions require the feature extractor to have direction perception ability, but the traditional convolution operation lacks pertinence in the response to direction features, resulting in difficulty in effectively mining the key information in different direction subbands. Based on this, in the technical solution of the present application, feature extraction based on dilated convolution coding is performed on the low-frequency approximation subband matrix, the horizontal direction detail subband matrix, the vertical direction detail subband matrix, and the diagonal direction detail subband matrix to obtain the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map. Specifically, for the low-frequency approximation subband, a dilated convolution layer with a large dilation rate is used to expand the receptive field without increasing the number of parameters, so that it can cover a larger range of brightness gradient regions and effectively capture the overall distribution law of low-contrast Mura; in the three high-frequency detail subbands of horizontal, vertical, and diagonal, a group of directional dilated convolution kernels is used to align the dilation direction of the convolution kernel with the spatial characteristics of each subband. For example, the horizontal direction detail subband matches the horizontally extended dilated convolution pattern, thereby enhancing the response sensitivity to the edge features in a specific direction. This coding method breaks through the isotropic limitation of the traditional convolution operation, enabling the feature extraction process to adapt to the spatial-frequency characteristics of different frequency bands.

[0027] In step S32, feature enhancement and fusion processing based on the spatial-semantic two-dimensionality is performed on the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map to obtain the multi-scale fusion feature map of the LCD panel image. Specifically, ​ It is a flowchart of step S32 in the LCD system fault detection method based on image analysis according to an embodiment of the present application. As ​As shown in the figure, step S32 includes: S321, performing feature enhancement based on the spatial-semantic two dimensions on the low-frequency approximation sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain an enhanced low-frequency approximation sub-band feature map, an enhanced horizontal direction detail sub-band feature map, an enhanced vertical direction detail sub-band feature map, and an enhanced diagonal direction detail sub-band feature map; S322, fusing the enhanced low-frequency approximation sub-band feature map, the enhanced horizontal direction detail sub-band feature map, the enhanced vertical direction detail sub-band feature map, and the enhanced diagonal direction detail sub-band feature map to obtain the multi-scale fusion feature map of the LCD panel image.

[0028] In step S321, performing feature enhancement based on the spatial-semantic two dimensions on the low-frequency approximation sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain an enhanced low-frequency approximation sub-band feature map, an enhanced horizontal direction detail sub-band feature map, an enhanced vertical direction detail sub-band feature map, and an enhanced diagonal direction detail sub-band feature map. Specifically, ​ It is a flowchart of step S321 in the LCD system fault detection method based on image analysis according to an embodiment of the present application. As ​ shown, step S321 includes: S321-1, extracting the channel feature vector at the (i, j) pixel position from the low-frequency approximation sub-band feature map as the low-frequency approximation sub-band feature channel feature vector to be enhanced; S321-2, performing n random scans on the low-frequency approximation sub-band feature map to obtain n low-frequency approximation sub-band feature channel feature vectors as a sparse set of low-frequency approximation sub-band feature reference vectors; S321-3, performing implicit compensation enhancement based on the mask control component on the sparse set of low-frequency approximation sub-band feature reference vectors and the low-frequency approximation sub-band feature channel feature vector to be enhanced to obtain a low-frequency approximation sub-band feature enhancement component implicit coding vector; S321-4, fusing the low-frequency approximation sub-band feature enhancement component implicit coding vector and the low-frequency approximation sub-band feature channel feature vector to be enhanced to obtain an enhanced low-frequency approximation sub-band feature channel feature vector, where the enhanced low-frequency approximation sub-band feature channel feature vector is the channel feature vector at the pixel position (i, j) of the enhanced low-frequency approximation sub-band feature map.

[0029] It should be understood that although the sub-band feature maps of different frequency bands and directions achieve multi-scale expression of defect features through wavelet decomposition, there are still problems of inherent interference and limitations in feature expression in each sub-band. In the low-frequency approximation sub-band, there is spectral overlap between the global brightness gradient caused by uneven backlighting and the low-frequency characteristics of real defects. Traditional feature enhancement methods are difficult to suppress background interference while preserving the overall shape of the defect. In the high-frequency detail sub-bands, high-frequency artifacts formed by camera noise and surface reflection are easily confused with defect edge features. Moreover, the different responses of sub-bands in different directions to the shape of the defect (such as the horizontal sub-band being more sensitive to vertically extending defects) require that the enhancement process takes into account both spatial distribution and semantic association. More critically, the local gradual change characteristics of Mura defects may only appear as weak signals in a single sub-band, and effective characterization can only be formed by mining cross-region context information. However, conventional enhancement methods lack the ability to explicitly model the implicit semantic associations in the feature space, resulting in the enhancement process being unable to adapt to the physical characteristics of different sub-bands. Therefore, in this application, feature enhancement based on the spatial-semantic two-dimensionality is performed on the low-frequency approximation sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain an enhanced low-frequency approximation sub-band feature map, an enhanced horizontal direction detail sub-band feature map, an enhanced vertical direction detail sub-band feature map, and an enhanced diagonal direction detail sub-band feature map.

[0030] Specifically, the processing process of the low-frequency approximation sub-band feature map is taken as an example for detailed description here. Specifically, first, the low-frequency approximation sub-band feature map is randomly scanned to obtain a set of reference vectors covering different brightness distribution regions. The hierarchical association between the feature vector to be enhanced and each reference vector in the hyperbolic space is measured by the Poincaré distance, and a spatial modulation parameter matrix is constructed to quantify the intensity of spatial correlation. At the same time, the implicit semantic association between the feature to be enhanced and the reference vector is calculated, and the abstract similarity in the brightness gradual change pattern between the two is mined to generate a semantic association coding matrix. Subsequently, the reference vectors are screened at the semantic level through the semantic association coding matrix as the basic mask control component, and then the spatial correlation modulation is implemented with the spatial modulation degree matrix as the secondary mask control component to form an information compensation mechanism controlled by a double-layer mask. This process dynamically injects the reference vector information that conforms to the defect brightness distribution pattern into the feature to be enhanced. For example, when there is a weak brightness change suspected of being a defect in a certain area, the reference area with a similar gradual change pattern is screened out through semantic association, and then its contribution weight is strengthened according to the spatial correlation. Finally, the enhanced low-frequency approximation sub-band feature map is generated through feature fusion.

[0031] Specifically, in the embodiment of this application, step S321-1 includes: extracting the channel feature vector at the (i, j) pixel position from the low-frequency approximation sub-band feature map as the channel feature vector of the low-frequency approximation sub-band feature to be enhanced, which can be expressed by the following formula: ; among them, is the low-frequency approximation sub-band feature map, is the set of real numbers, and are respectively the height and width of each feature matrix along the channel dimension, is the number of channels, represents the set of real numbers composed of the number of channels of is the channel feature vector at the (i, j) pixel position in is the channel feature vector of the low-frequency approximation sub-band feature channel to be enhanced.

[0032] It should be understood that the channel feature vector at a specific position (i, j) is extracted from the low-frequency approximation sub-band feature map as the object to be enhanced, aiming to focus on the brightness distribution characteristics of this area. Specifically, since there is spectral overlap between the global backlight non-uniformity in the low-frequency sub-band and the low-frequency characteristics of real defects, directly processing the entire feature map may cause the background interference to be enhanced by mistake. Through local feature vector extraction, on the basis of retaining the overall shape of the defect (such as the gradual change characteristics of Mura defects), the local brightness gradual change pattern and background noise can be separated specifically. That is, through the localization operation, an accurate object to be enhanced is provided for the subsequent space-semantic modulation, avoiding the over-smoothing or information loss of the non-defect area by global operations, and at the same time maintaining the spatial continuity of the feature map, laying a foundation for cross-region context mining.

[0033] Specifically, in the embodiment of the present application, the step S321-2 includes: performing n random scans on the low-frequency approximation sub-band feature map to obtain n low-frequency approximation sub-band feature channel feature vectors as a sparse set of low-frequency approximation sub-band feature reference vectors, which can be expressed by the following formula: ; among them, is the sparse set of low-frequency approximation sub-band feature reference vectors, and are respectively the 1st, the 2nd, the th, and the th low-frequency approximation sub-band feature reference vectors in the sparse set of low-frequency approximation sub-band feature reference vectors, and .

[0034] It should be understood that by generating a sparse set of low-frequency approximate subband feature reference vectors through n random samplings of the low-frequency subband feature map, it is essentially constructing a dynamic feature memory bank that covers different brightness distribution patterns. Specifically, the randomness design breaks through the sensitivity of fixed sampling (such as sliding window) to local overfitting. By introducing spatial diversity, it ensures that the reference set can cover the typical brightness gradient patterns in the feature map (including potential defect areas and normal background areas). The sparse sampling strategy then balances computational efficiency and information coverage, simulating the "selective attention" mechanism of biological vision, screening representative local patterns from global features, and avoiding interference from noise or redundant features. That is, this step provides a multi-scale and multi-position context reference benchmark for subsequent modulation, enabling the enhancement process to adapt to the brightness distribution characteristics of different regions.

[0035] Specifically, ​ FIG. is a flowchart of step S321-3 in the LCD system fault detection method based on image analysis according to an embodiment of the present application. As ​ shown, the step S321-3 includes: S321-31, calculating the Poincaré distance between the channel feature vector of the low-frequency approximate subband feature to be enhanced and each low-frequency approximate subband feature reference vector in the sparse set of low-frequency approximate subband feature reference vectors to obtain a low-frequency approximate subband feature space modulation parameter matrix; S321-32, calculating the implicit semantic association between the channel feature vector of the low-frequency approximate subband feature to be enhanced and each low-frequency approximate subband feature reference vector in the sparse set of low-frequency approximate subband feature reference vectors to obtain a set of low-frequency approximate subband feature semantic association coding matrices; S321-33, using each low-frequency approximate subband feature semantic association coding matrix in the set of low-frequency approximate subband feature semantic association coding matrices as a basic mask control component and using the low-frequency approximate subband feature space modulation parameter matrix as a secondary mask control component, and performing explicit modeling control on the low-frequency approximate subband feature information compensation coding vector between each low-frequency approximate subband feature reference vector in the sparse set of low-frequency approximate subband feature reference vectors and the channel feature vector of the low-frequency approximate subband feature to be enhanced to obtain the low-frequency approximate subband feature enhancement component implicit coding vector.

[0036] More specifically, in the embodiment of the present application, the step S321-31 includes: calculating the Poincaré distance between the channel feature vector of the low-frequency approximate subband feature to be enhanced and each low-frequency approximate subband feature reference vector in the sparse set of low-frequency approximate subband feature reference vectors to obtain a low-frequency approximate subband feature space modulation parameter matrix, which can be expressed by the following formula: ; where is the square of the Euclidean norm of the vector, is the inverse hyperbolic cosine function, is and the Poincaré distance between, are the eigenvalues in the low-frequency approximate subband feature space modulation parameter matrix, is the low-frequency approximate subband feature space modulation parameter matrix.

[0037] It should be understood that using the Poincaré distance to measure the hierarchical correlation between the feature vectors of the low-frequency approximate subband features to be enhanced and the low-frequency approximate subband feature reference vectors in the hyperbolic space is more adept at capturing the characteristics of tree-like or hierarchical data distributions compared to the Euclidean distance. Specifically, the brightness gradient patterns in the low-frequency subbands (such as the intensity changes between defects and the background) often exhibit non-linear hierarchical relationships (e.g., strong gradient in the center to weak gradient at the edges). The Poincaré distance, through the geometric properties of the hyperbolic space, models such relationships as the strength of spatial correlation. The generated low-frequency approximate subband feature space modulation parameter matrix not only quantifies the spatial similarity between the region to be enhanced and the reference region but also implicitly encodes the brightness distribution patterns across regions (such as the local gradient trend of defects and the global change differences of the background), providing a basis for differential weight allocation in subsequent spatial modulation, thereby enhancing defects while suppressing interference from non-related regions.

[0038] More specifically, in the embodiment of the present application, the step S321-32 includes: calculating the implicit semantic association between the feature vectors of the low-frequency approximate subband features to be enhanced and each low-frequency approximate subband feature reference vector in the sparse set of the low-frequency approximate subband feature reference vectors to obtain a set of low-frequency approximate subband feature semantic association coding matrices, which can be expressed by the following formula: ; where is matrix multiplication, is the corresponding weight matrix, is the transpose operation, is and the length of the vector after multiplication, is function, is and the low-frequency approximate subband feature semantic association coding matrix between, that is, the th low-frequency approximate subband feature semantic association coding matrix in the set of low-frequency approximate subband feature semantic association coding matrices, is the set of low-frequency approximate subband feature semantic association coding matrices, and are the 1st, 2nd and respectively in the set of low-frequency approximate subband feature semantic association coding matricesA semantic association coding matrix for low-frequency approximate sub-band features.

[0039] It should be understood that by mining the abstract semantic similarity between the feature channel vectors of the low-frequency approximate sub-band features to be enhanced and the reference vectors of the low-frequency approximate sub-band features (such as the commonalities and differences in the brightness gradient patterns), a semantic association coding matrix for the low-frequency approximate sub-band features is generated. This step goes beyond the pixel-level surface feature matching and focuses on high-order semantic patterns (such as "center-edge gradient" and "multi-scale texture similarity"), which can effectively solve the mis-association problem caused by traditional methods relying on explicit feature similarity. For example, when there are weak defects in the area to be enhanced, semantic association can screen out reference areas with similar gradient directions but different intensities, even if their spatial distances are far. That is, the obtained semantic association coding matrix for the low-frequency approximate sub-band features serves as a basic mask, realizing the semantic-level screening of the reference vectors, ensuring that only the feature components strongly related to the defect pattern are injected into the subsequent information compensation, and avoiding the interference of irrelevant background information.

[0040] More specifically, in the embodiment of the present application, the steps S321-33 include: calculating the position-wise difference vectors between each of the reference vectors of the low-frequency approximate sub-band features and the feature channel vectors of the low-frequency approximate sub-band features to be enhanced to obtain a set of information compensation coding vectors for the low-frequency approximate sub-band features. This process can be expressed as: ; where is and the information compensation coding vector between

[0041] Based on the basic mask control component and the secondary mask control component, perform mask-compensation optimization based on dual-field collaborative modulation on each information compensation coding vector in the set of information compensation coding vectors for the low-frequency approximate sub-band features to obtain a set of optimized coding vectors for the information compensation of the low-frequency approximate sub-band features. This process can be expressed as: ; where is the regulated information compensation coding vector for the low-frequency approximate sub-band features, is the optimized information compensation coding vector for the low-frequency approximate sub-band features.

[0042] Based on the basic mask control component and the secondary mask control component, perform global explicit modeling modulation aggregation on the set of optimized coding vectors for the information compensation of the low-frequency approximate sub-band features to obtain the implicit coding vector for the enhanced component of the low-frequency approximate sub-band features. The above process can be expressed as a formula as a whole: ; where is the number of vectors in the sparse set of reference vectors for the low-frequency approximate sub-band features, is the hidden coding vector of the low-frequency approximate sub-band feature enhancement component.

[0043] It should be understood that a hierarchical modulation mechanism is formed with the low-frequency approximate sub-band feature semantic association coding matrix as the primary mask and the low-frequency approximate sub-band feature spatial modulation parameter matrix as the secondary mask. Specifically, the primary mask filters out the low-frequency approximate sub-band feature reference vectors (such as regions with similar gradient directions) that match the defect pattern based on semantic similarity, and the secondary mask then strengthens the contribution weight of the low-frequency approximate sub-band feature reference vectors within the local neighborhood according to spatial correlation. For example, for a suspected defect area, the semantic mask preferentially selects reference vectors with similar gradient patterns, and the spatial mask assigns higher weights to adjacent areas, thereby dynamically fusing cross-region context information. This explicit modeling strategy breaks through the dependence of traditional enhancement methods on implicit feature learning. Through an interpretable double-layer control mechanism, it accurately injects enhancement components that conform to the physical characteristics of the defect (such as supplementing weak gradient signals), while suppressing pseudo-enhancement caused by noise or uneven background.

[0044] In particular, in the information coding framework for low-frequency approximate sub-band feature enhancement, the implementation of the field density modulation configuration depends on the synergistic effect of the low-frequency approximate sub-band feature semantic association coding matrix and the low-frequency approximate sub-band feature spatial modulation parameter matrix. Specifically, the low-frequency approximate sub-band feature semantic association coding matrix serves as the standard field, encoding the hidden semantic associations between feature vectors (such as the abstract similarity of brightness gradient patterns), and it maintains the transformation invariance constraint condition. This operation essentially projects the information compensation component (i.e., the difference information between the feature to be enhanced and the reference vector) regulated by the semantic mask into the geometric measure space of the standard field, ensuring the robustness of the enhancement process to the global brightness gradient background. The low-frequency approximate sub-band feature spatial modulation parameter matrix serves as the covariant adjustment field, dynamically adjusting the local geometric structure of the standard field through the spatial correlation intensity (i.e., the hierarchical brightness distribution pattern) quantified by the Poincaré distance. Based on The covariant adjustment mechanism applies spatial modulation parameters to the curved spatial topology, enabling the derivative-level geometric measure of the standard field to adapt to the spatial distribution characteristics of the feature map. For example, at the edge of a defect, the covariant adjustment field corrects the deviation of the standard field in local curvature by strengthening the contribution weight of neighboring reference vectors, thereby improving the spatial continuity of the defect gradient signal. In other words, the intensity coupling of the field interaction is quantified using conventional partial differential forms. The core of this approach is to establish a differential manifold relationship between semantic association and spatial modulation. Specifically, the partial differential coupling term between the standard field and the covariant adjustment field reflects the joint gradient direction of semantic filtering and spatial enhancement, enabling the optimized coding vector compensated by low-frequency approximate subband feature information to simultaneously satisfy semantic consistency and spatial smoothness constraints. This coupling mechanism not only suppresses high-frequency noise interference (such as artifacts caused by camera noise) but also enhances the low-frequency subband feature representation of weak defects by mining cross-regional contextual information (such as the long-range correlation between the local gradient pattern of the defect and the reference vector).

[0045] Specifically, in an embodiment of the present application, step S321-4 is used to: fuse the low-frequency approximate sub-band feature enhancement component implicit coding vector and the low-frequency approximate sub-band feature channel feature vector to be enhanced to obtain an enhanced low-frequency approximate sub-band feature channel feature vector, wherein the enhanced low-frequency approximate sub-band feature channel feature vector is the channel feature vector of the pixel position (i, j) of the enhanced low-frequency approximate sub-band feature map, which can be expressed as the following formula: ;in, and is a weighted hyperparameter, yes The enhanced low-frequency approximate sub-band feature channel feature vector after enhancement is the channel feature vector of the (i, j)th pixel position of the enhanced low-frequency approximate sub-band feature map.

[0046] It should be understood that fusing the implicit coding vector of the low-frequency approximate subband feature enhancement component with the feature vector of the low-frequency approximate subband feature channel to be enhanced can achieve the coordinated optimization of local feature expression and global context information. Specifically, the original feature vector of the low-frequency approximate subband feature channel to be enhanced retains the initial morphology and low-frequency characteristics of the defect, while the implicit coding vector of the low-frequency approximate subband feature enhancement component supplements the weak signal lost due to background interference or noise suppression (such as the local gradient details of mura defects) through cross-regional semantic association. The fusion process is not a simple weighted superposition, but rather reconstructs the feature space distribution through nonlinear mapping, so that the enhanced features are both sensitive to small defects and robust to background changes. For example, in areas with uneven backlighting, the fused features can highlight the continuous gradient pattern of the defect while weakening the interference of global brightness changes, ultimately improving the low-frequency subband feature's ability to distinguish defects from the background.

[0047] In step S322, the enhanced low-frequency approximation sub-band feature map, the enhanced horizontal direction detail sub-band feature map, the enhanced vertical direction detail sub-band feature map, and the enhanced diagonal direction detail sub-band feature map are fused to obtain the multi-scale fusion feature map of the LCD panel image. Accordingly, considering that although the sub-band feature maps of different frequency bands and directions respectively enhance the defect features of specific dimensions through enhancement processing, there are still limitations of information fragmentation in their isolated analysis. Although the enhanced low-frequency approximation sub-band feature map can represent the overall brightness distribution of the defect, it loses the edge detail information; while the high-frequency detail enhanced sub-band feature maps in the horizontal, vertical, and diagonal directions capture the local gradient changes, but lack the context awareness of the global morphology of the defect. Therefore, in order to make full use of the advantages of each sub-band, integrate the global and local information of the image, and more comprehensively describe the features of the LCD panel image, this application fuses the enhanced low-frequency approximation sub-band feature map, the enhanced horizontal direction detail sub-band feature map, the enhanced vertical direction detail sub-band feature map, and the enhanced diagonal direction detail sub-band feature map to obtain the multi-scale fusion feature map of the LCD panel image. In this way, the obtained multi-scale fusion feature map of the LCD panel image can provide a comprehensive feature representation for subsequent defect detection and analysis. It contains both the overall structural information of the image and the detailed texture information, organically combines the features of different scales and directions, enables the model to analyze and understand the image from multiple angles, and thus can more accurately locate and describe the features such as the position, shape, size, and texture of the Mura defect, providing a more sufficient basis for the classification and evaluation of the defect.

[0048] In step S4, statistical analysis based on the Mura segmentation region is performed on the multi-scale fusion feature map of the LCD panel image to obtain the Mura region statistical feature coding vector. Specifically, ​ It is a flowchart of step S4 in the LCD system fault detection method based on image analysis according to an embodiment of the present application. As ​ shown, the step S4 includes: S41, performing image semantic segmentation on the multi-scale fusion feature map of the LCD panel image to obtain the Mura segmentation region image; S42, extracting the statistical features of the Mura segmentation region image to obtain the Mura region statistical feature coding vector.

[0049] In step S41, image semantic segmentation is performed on the multi-scale fusion feature map of the LCD panel image to obtain a Mura segmentation region image. Accordingly, considering that although the multi-scale fusion feature map of the LCD panel image integrates the global brightness distribution and multi-directional edge details, the problem of accurate boundary division of the defect region still needs to be solved. Image semantic segmentation can classify each pixel point in the image according to the information in the feature map to determine whether it belongs to the Mura defect region, thereby accurately outlining the boundary and scope of the Mura defect, and providing accurate position information for subsequent defect analysis and processing. Based on this, in the technical solution of this application, image semantic segmentation is performed on the multi-scale fusion feature map of the LCD panel image to obtain a Mura segmentation region image. That is, the Mura segmentation region image obtained through image semantic segmentation can accurately identify the Mura defect region on the LCD panel. Classify each pixel point in the image, mark the pixel points belonging to the Mura defect, and form a clear segmentation region, making the Mura defect clearly visible in the image, facilitating direct positioning and observation of the defect by operators or subsequent analysis algorithms, and providing a basis for defect evaluation and repair.

[0050] Specifically, first, the multi-scale fusion feature map of the LCD panel image is used as input data, and this feature map has integrated multi-scale information of low-frequency brightness distribution and high-frequency edge details through previous steps. To ensure that the feature values of each channel are in the same dimension, it is necessary to normalize the input feature map, for example, using the Min-Max scaling method, to avoid subsequent segmentation deviation caused by differences in numerical ranges.

[0051] In the semantic segmentation network architecture design stage, a network model with an encoder-decoder structure is usually adopted, such as U-Net or DeepLabv3+, to balance the fusion of spatial details and high-level semantic information. The encoder part gradually downsamples through convolutional and pooling operations to extract high-level semantic features, including the texture pattern and brightness distribution law of the Mura defect. The decoder part restores the spatial resolution through transposed convolution or interpolation upsampling and combines the skip connection features of the encoder to refine the segmentation boundary. To further improve the segmentation accuracy, a spatial or channel attention mechanism, such as the CBAM module, can be introduced in the decoding stage to dynamically weight the key regions related to the Mura defect in the multi-scale feature map while suppressing the interference of background noise.

[0052] During the training phase, labeled data needs to be prepared as the supervision signal. Usually, manually labeled or simulated synthetic binary masks of Mura defects are used, where the defect areas are marked as 1 and the background is marked as 0. When designing the loss function, the class imbalance problem where Mura defects usually account for a relatively small proportion needs to be considered. Therefore, the main loss function can adopt Dice Loss or Focal Loss to enhance the model's attention to the defect areas. In addition, a boundary-aware loss, such as Boundary Loss, can be introduced to strengthen the network's ability to recognize the gradually changing transition areas at the edges of Mura defects.

[0053] During the inference phase, the preprocessed multi-scale fusion feature maps of the LCD panel images are input into the trained semantic segmentation network, and a probability map indicating the probability of each pixel belonging to a Mura defect is output, with its value range from 0 to 1. Subsequently, the probability map is converted into a binary segmentation mask through thresholding. An adaptive thresholding method (such as the Otsu algorithm) or a fixed threshold (such as 0.5) can be used for binarization to generate a preliminary Mura segmentation region image. To optimize the segmentation result, morphological post-processing is required, including using opening operations to eliminate isolated noise points and retaining regions with an area larger than a preset threshold through connected component analysis to filter out small pseudo-defects.

[0054] Finally, the optimized binary mask is aligned with the original image to ensure that the boundaries of the defect areas match precisely, and a Mura segmentation region image is output, where the defect areas are represented in white and the background is represented in black. To verify the segmentation accuracy, metrics such as IoU (Intersection over Union) or pixel-level accuracy can be used to evaluate the consistency between the segmentation result and the ground truth annotation, ensuring that it meets the accuracy requirements of industrial inspection.

[0055] In step S42, statistical features of the Mura segmentation region image are extracted to obtain the Mura region statistical feature coding vector. Correspondingly, although semantic segmentation can accurately locate the defect region, relying solely on the segmentation result cannot meet the requirements of defect classification and quantitative evaluation. When directly evaluating defects based on pixel-level brightness values or simple morphological parameters, it is difficult to characterize the complex visual characteristics of defects. For example, the human eye's perception of Mura defects depends not only on the absolute value of the brightness difference but also on abstract features such as distribution uniformity and texture coherence. In addition, there are significant differences in the distribution laws of different forms of Mura defects (such as dot-shaped, line-shaped, cloud-shaped) in the statistical feature space, but the existing methods lack systematic modeling of multi-dimensional features, resulting in difficulty in establishing the mapping relationship between defect forms and quality levels for the classification model. Based on this, the present application extracts the statistical features of the Mura segmentation region image to obtain the Mura region statistical feature coding vector. In particular, in a specific example of the present application, for the segmented binary mask region, first, the brightness distribution features in the original enhanced image within its coverage are extracted, including indicators such as gray mean, variance, and gradient histogram entropy, to characterize the overall contrast and local fluctuation characteristics of the defects; at the same time, the morphological features of the defect region, such as area perimeter ratio, compactness, and Fourier descriptors, are calculated to quantify its geometric distribution law; further, the contrast and correlation parameters of the gray-level co-occurrence matrix are extracted by combining texture analysis algorithms to capture the microscopic texture characteristics of the defects. After normalization and feature selection, these features are fused into a statistical feature coding vector of a fixed dimension according to preset weights, forming a multi-dimensional defect fingerprint that takes into account brightness, morphology, and texture.

[0056] In step S5, based on the Mura region statistical feature coding vector, a Mura defect recognition result is obtained. Specifically, in the embodiment of the present application, step S5 includes: inputting the Mura region statistical feature coding vector into a defect classifier based on an SVM model to obtain the Mura defect recognition result. In particular, the Mura defect recognition result is divided into levels such as "no defect", "slight defect", "moderate defect", and "severe defect". It should be understood that the support vector machine (SVM) is a supervised machine learning algorithm that performs well in pattern recognition and classification tasks. It has good performance in dealing with a limited number of high-dimensional data, and the Mura region statistical feature coding vector is exactly a high-dimensional data representation obtained after feature extraction from the Mura segmented region image. The SVM can find an optimal hyperplane in the high-dimensional space to separate data points of different categories, and has effective processing methods for both linearly separable and non-linearly separable data (mapping the data to a higher-dimensional space through a kernel function to make it linearly separable). In the Mura defect classification task, the statistical feature coding vectors corresponding to different types of Mura defects can be regarded as data points of different categories, and the SVM can use the features of these data points to construct a classification model. By analyzing and judging the input feature vector through the SVM model, according to the pre-trained classification model, the Mura defects are classified into corresponding categories, such as different types of Mura defects corresponding to different shapes, sizes, gray-scale distributions, etc. This helps production personnel understand the specific situation of the defects, so as to take targeted measures for repair or improvement of the production process.

[0057] In summary, the image analysis-based LCD system fault detection method according to the embodiment of the present application is clarified. First, an LCD panel image collected by a high-resolution industrial camera is obtained, then the image is subjected to homomorphic filtering and CLAHE contrast enhancement processing to obtain an enhanced LCD panel image, then multi-scale feature enhancement fusion is performed on the enhanced image to obtain a multi-scale fusion feature map of the LCD panel image, then statistical analysis is performed on the feature map based on the Mura segmentation region to obtain a Mura region statistical feature coding vector, and finally, based on this coding vector, a Mura defect recognition result is obtained. In this way, the features of Mura defects can be captured more accurately, reducing false positives and missed detections caused by interference factors such as light sources and noise, improving the detection accuracy of Mura defects, and at the same time, the detection system can adapt to different production conditions and ensure the stability of the detection results.

[0058] ​ FIG. is a system block diagram of an image analysis-based LCD system fault detection system according to an embodiment of the present application. As ​As shown, the LCD system fault detection system 100 based on image analysis according to an embodiment of the present application includes: an image acquisition module 110 for acquiring an LCD panel image collected by a high-resolution industrial camera; an image preprocessing module 120 for performing homomorphic filtering and CLAHE contrast enhancement on the LCD panel image to obtain an enhanced LCD panel image; an image encoding module 130 for performing multi-scale feature enhancement fusion on the enhanced LCD panel image to obtain a multi-scale fusion feature map of the LCD panel image; a Mura region analysis module 140 for performing statistical analysis based on the Mura segmentation region on the multi-scale fusion feature map of the LCD panel image to obtain a Mura region statistical feature coding vector; and a defect recognition module 150 for obtaining a Mura defect recognition result based on the Mura region statistical feature coding vector.

[0059] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned LCD system fault detection system 100 based on image analysis have been described in detail above with reference to ​ the description of the LCD system fault detection method based on image analysis, and therefore, the repeated description thereof will be omitted.

[0060] In summary, the LCD system fault detection system 100 based on image analysis according to an embodiment of the present application is elucidated. It first acquires an LCD panel image collected by a high-resolution industrial camera, then performs homomorphic filtering and CLAHE contrast enhancement processing on the image to obtain an enhanced LCD panel image, then conducts multi-scale feature enhancement fusion on the enhanced image to obtain a multi-scale fusion feature map of the LCD panel image, then performs statistical analysis based on the Mura segmentation region on this feature map to obtain a Mura region statistical feature coding vector, and finally, based on this coding vector, obtains a Mura defect recognition result. In this way, it can more accurately capture the characteristics of Mura defects, reduce false positives and missed detections caused by interference factors such as light sources and noise, improve the detection accuracy of Mura defects, and at the same time, the detection system can adapt to different production conditions to ensure the stability of the detection results.

Claims

1. A method for detecting faults in an LCD system based on image analysis, characterized in that, Including: Obtain an LCD panel image collected by a high-resolution industrial camera; Perform homomorphic filtering and CLAHE contrast enhancement on the LCD panel image to obtain an enhanced LCD panel image; Perform multi-scale feature enhancement fusion on the enhanced LCD panel image to obtain a multi-scale fusion feature map of the LCD panel image, including: perform wavelet decomposition and feature extraction on the enhanced LCD panel image to obtain a low-frequency approximation subband feature map, a horizontal direction detail subband feature map, a vertical direction detail subband feature map, and a diagonal direction detail subband feature map; perform feature enhancement fusion processing based on the spatial-semantic two-dimensionality on the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map to obtain the multi-scale fusion feature map of the LCD panel image, wherein, randomly scan the low-frequency approximation subband feature map, obtain a set of reference vectors covering different brightness distribution regions, measure the hierarchical association between the feature vector to be enhanced and each reference vector in the hyperbolic space through the Poincaré distance, and construct a spatial modulation parameter matrix to quantify the spatial correlation intensity; calculate the implicit semantic association between the feature to be enhanced and the reference vector, generate a semantic association coding matrix, use the semantic association coding matrix as a basic mask control component to screen the reference vector at the semantic level, and then use the spatial modulation degree matrix as a secondary mask control component to implement spatial correlation modulation; Perform statistical analysis based on the Mura segmentation region on the multi-scale fusion feature map of the LCD panel image to obtain a Mura region statistical feature coding vector; Based on the Mura region statistical feature coding vector, obtain a Mura defect recognition result.

2. The method for detecting faults in an LCD system based on image analysis according to claim 1, wherein, Perform wavelet decomposition and feature extraction on the enhanced LCD panel image to obtain a low-frequency approximation subband feature map, a horizontal direction detail subband feature map, a vertical direction detail subband feature map, and a diagonal direction detail subband feature map, including: Perform wavelet decomposition on the enhanced LCD panel image to obtain a low-frequency approximation subband matrix, a horizontal direction detail subband matrix, a vertical direction detail subband matrix, and a diagonal direction detail subband matrix; Perform feature extraction based on dilated convolution coding on the low-frequency approximation subband matrix, the horizontal direction detail subband matrix, the vertical direction detail subband matrix, and the diagonal direction detail subband matrix to obtain the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map.

3. The method for detecting faults in an LCD system based on image analysis according to claim 1, characterized in that, Perform feature enhancement fusion processing based on the spatial-semantic two-dimensionality on the low-frequency approximation subband feature map, the horizontal direction detail subband feature map, the vertical direction detail subband feature map, and the diagonal direction detail subband feature map to obtain the multi-scale fusion feature map of the LCD panel image, including: Perform feature enhancement based on the spatial-semantic two-dimensionality on the low-frequency approximate sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain an enhanced low-frequency approximate sub-band feature map, an enhanced horizontal direction detail sub-band feature map, an enhanced vertical direction detail sub-band feature map, and an enhanced diagonal direction detail sub-band feature map; Fuse the enhanced low-frequency approximate sub-band feature map, the enhanced horizontal direction detail sub-band feature map, the enhanced vertical direction detail sub-band feature map, and the enhanced diagonal direction detail sub-band feature map to obtain the multi-scale fusion feature map of the LCD panel image.

4. The method for detecting faults in an LCD system based on image analysis according to claim 3, characterized in that, Perform feature enhancement based on the spatial-semantic two-dimensionality on the low-frequency approximate sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain an enhanced low-frequency approximate sub-band feature map, an enhanced horizontal direction detail sub-band feature map, an enhanced vertical direction detail sub-band feature map, and an enhanced diagonal direction detail sub-band feature map, including: Extract the channel feature vector at the (i,j) pixel position from the low-frequency approximate sub-band feature map as the low-frequency approximate sub-band feature channel feature vector to be enhanced; Perform n random scans on the low-frequency approximate sub-band feature map to obtain n low-frequency approximate sub-band feature channel feature vectors as a sparse set of low-frequency approximate sub-band feature reference vectors; Perform implicit compensation enhancement based on the mask control component on the sparse set of low-frequency approximate sub-band feature reference vectors and the low-frequency approximate sub-band feature channel feature vector to be enhanced to obtain the implicit coding vector of the low-frequency approximate sub-band feature enhancement component; Fuse the implicit coding vector of the low-frequency approximate sub-band feature enhancement component and the low-frequency approximate sub-band feature channel feature vector to be enhanced to obtain the enhanced low-frequency approximate sub-band feature channel feature vector, where the enhanced low-frequency approximate sub-band feature channel feature vector is the channel feature vector at the pixel position (i,j) of the enhanced low-frequency approximate sub-band feature map.

5. The method for detecting faults of the LCD system based on image analysis according to claim 4, wherein, Perform implicit compensation enhancement based on the mask control component on the sparse set of low-frequency approximate sub-band feature reference vectors and the low-frequency approximate sub-band feature channel feature vector to be enhanced to obtain the implicit coding vector of the low-frequency approximate sub-band feature enhancement component, including: Calculate the Poincaré distance between the low-frequency approximate sub-band feature channel feature vector to be enhanced and each low-frequency approximate sub-band feature reference vector in the sparse set of low-frequency approximate sub-band feature reference vectors to obtain the low-frequency approximate sub-band feature space modulation parameter matrix; Calculate the implicit semantic association between the low-frequency approximate sub-band feature channel feature vector to be enhanced and each low-frequency approximate sub-band feature reference vector in the sparse set of low-frequency approximate sub-band feature reference vectors to obtain a set of low-frequency approximate sub-band feature semantic association coding matrices; Using each low-frequency approximation sub-band feature semantic association coding matrix in the set of low-frequency approximation sub-band feature semantic association coding matrices as a basic mask regulation component and using the low-frequency approximation sub-band feature space modulation parameter matrix as a secondary mask regulation component, perform explicit modeling regulation on the low-frequency approximation sub-band feature information compensation coding vectors between each low-frequency approximation sub-band feature reference vector in the sparse set of low-frequency approximation sub-band feature reference vectors and the low-frequency approximation sub-band feature channel feature vector to obtain the low-frequency approximation sub-band feature enhancement component implicit coding vector.

6. The method for detecting faults in an LCD system based on image analysis according to claim 5, wherein, Using each low-frequency approximation sub-band feature semantic association coding matrix in the set of low-frequency approximation sub-band feature semantic association coding matrices as a basic mask regulation component and using the low-frequency approximation sub-band feature space modulation parameter matrix as a secondary mask regulation component, perform explicit modeling regulation on the low-frequency approximation sub-band feature information compensation coding vectors between each low-frequency approximation sub-band feature reference vector in the sparse set of low-frequency approximation sub-band feature reference vectors and the low-frequency approximation sub-band feature channel feature vector to obtain the low-frequency approximation sub-band feature enhancement component implicit coding vector, including: Calculate the position-wise difference vectors between each low-frequency approximation sub-band feature reference vector and the low-frequency approximation sub-band feature channel feature vector to be enhanced to obtain a set of low-frequency approximation sub-band feature information compensation coding vectors; Based on the basic mask regulation component and the secondary mask regulation component, perform mask-compensation optimization based on dual-field collaborative modulation on each low-frequency approximation sub-band feature information compensation coding vector in the set of low-frequency approximation sub-band feature information compensation coding vectors to obtain a set of low-frequency approximation sub-band feature information compensation optimized coding vectors; Based on the basic mask regulation component and the secondary mask regulation component, perform global explicit modeling modulation aggregation on the set of low-frequency approximation sub-band feature information compensation optimized coding vectors to obtain the low-frequency approximation sub-band feature enhancement component implicit coding vector.

7. The method for detecting faults of an LCD system based on image analysis according to claim 6, wherein Perform statistical analysis based on the Mura segmentation region on the LCD panel image multi-scale fusion feature map to obtain the Mura region statistical feature coding vector, including: Perform image semantic segmentation on the LCD panel image multi-scale fusion feature map to obtain the Mura segmentation region image; Extract the statistical features of the Mura segmentation region image to obtain the Mura region statistical feature coding vector.

8. The method for detecting faults in an LCD system based on image analysis according to claim 7, wherein Based on the Mura region statistical feature coding vector, obtain the Mura defect recognition result, including: inputting the Mura region statistical feature coding vector into a defect classifier based on the SVM model to obtain the Mura defect recognition result.

9. An LCD system fault detection system based on image analysis, characterized in that, Including: An image acquisition module for acquiring an LCD panel image collected by a high-resolution industrial camera; An image preprocessing module for performing homomorphic filtering and CLAHE contrast enhancement on the LCD panel image to obtain an enhanced LCD panel image; An image encoding module, configured to perform multi-scale feature enhancement fusion on the enhanced LCD panel image to obtain a multi-scale fusion feature map of the LCD panel image. Specifically, the image encoding module is configured to: perform wavelet decomposition and feature extraction on the enhanced LCD panel image to obtain a low-frequency approximation sub-band feature map, a horizontal direction detail sub-band feature map, a vertical direction detail sub-band feature map, and a diagonal direction detail sub-band feature map; perform feature enhancement fusion processing based on the spatial-semantic two-dimensionality on the low-frequency approximation sub-band feature map, the horizontal direction detail sub-band feature map, the vertical direction detail sub-band feature map, and the diagonal direction detail sub-band feature map to obtain the multi-scale fusion feature map of the LCD panel image. Among them, randomly scan the low-frequency approximation sub-band feature map to obtain a set of reference vectors covering different brightness distribution regions, measure the hierarchical association between the feature vector to be enhanced and each reference vector in the hyperbolic space through the Poincaré distance, and construct a spatial modulation parameter matrix to quantify the spatial correlation intensity; calculate the implicit semantic association between the feature to be enhanced and the reference vector, generate a semantic association coding matrix, use the semantic association coding matrix as a basic mask control component to perform semantic-level screening on the reference vector, and then use the spatial modulation degree matrix as a secondary mask control component to implement spatial correlation modulation. A Mura region analysis module, configured to perform statistical analysis based on the Mura segmentation region on the multi-scale fusion feature map of the LCD panel image to obtain a Mura region statistical feature coding vector. A defect recognition module, configured to obtain a Mura defect recognition result based on the Mura region statistical feature coding vector.

10. The LCD system fault detection system based on image analysis according to claim 9, characterized in that, The defect recognition module is configured to: input the Mura region statistical feature coding vector into a defect classifier based on an SVM model to obtain the Mura defect recognition result.

Images