A method and system for controlling dynamic splicing of LED display screen

Abstract

The present invention discloses a method and system for controlling dynamic splicing of an LED display screen, and relates to the technical field of dynamic splicing of an LED display screen, and includes the following steps: for each frame image of a pixel splicing area, the edge pixels of adjacent screens are overlapped using an initial edge pixel overlap ratio to preliminarily eliminate visible splicing marks; after preliminary processing, the edge portion of the pixel splicing area is divided into a plurality of sub-areas; for each sub-area, its pixel splicing situation is obtained. The present invention uses the initial edge pixel overlap ratio and a deep learning model to finely analyze the splicing situation, intelligently divide the smooth transition and non-smooth transition areas, centrally optimize the problem areas, dynamically adjust the overlap ratio and re-fuse the pixels through an adaptive edge fusion algorithm, ensure smooth image transition, solve the problems of misalignment and color difference, and improve the display effect and operation efficiency of dynamic content and large-scale display systems.

Description

Technical Field

[0001] The present invention relates to the technical field of dynamic splicing of LED display screens, and in particular to a method and system for controlling dynamic splicing of LED display screens. Background Art

[0002] LED display screen dynamic splicing control refers to the use of an intelligent control system to achieve seamless splicing and dynamic adjustment of screen content in a large-screen system composed of multiple LED screens. This technology can automatically adjust the display area and ratio of each screen according to the display content, screen layout and user needs to ensure the complete presentation of the picture. Especially when playing videos or dynamic content, it can adjust the splicing effect in real time to avoid tearing or dislocation of the picture. At the same time, it can also dynamically allocate display resources according to external signals (such as user input or environmental changes), optimize parameters such as picture quality and brightness, and enhance the visual experience. This technology is often used in large commercial displays, stage background walls or public information display systems.

[0003] During the splicing process, the edges between screens are the most likely places for misalignment or color differences to occur. To this end, the system uses edge blending technology to adjust parameters such as brightness, contrast, and color balance to make the edge transition between adjacent screens more natural, thereby eliminating visible splicing marks. The edge between screens refers to the physical connection area or intersection of adjacent screens when multiple independent display screens are spliced ​​into a large screen. Specifically, it is the outer edge of the display area of ​​each screen, which is directly adjacent during the splicing process and together forms a large continuous display area.

[0004] The prior art has the following deficiencies:

[0005] In the prior art, the edge pixel overlap ratio of the spliced ​​area is usually pre-set, mainly concentrated in the area where misalignment or color difference is most likely to occur. The preset edge pixel overlap ratio helps to reduce the complexity of real-time calculation and enable the system to run stably in different application scenarios. In most standard scenarios, the preset edge pixel overlap ratio can ensure a smooth visual transition and avoid frequent adjustments during use. However, if the preset edge pixel overlap ratio is insufficient to effectively solve the misalignment problem, there will be an obvious sense of screen tearing at the edge of the spliced ​​screen. Especially when playing fast-moving content or dynamic videos, the audience will find that the image in the spliced ​​area cannot be aligned smoothly, resulting in misalignment or breakage of the picture, which greatly destroys the overall visual experience. Especially on large display screens, this tearing phenomenon makes the displayed content difficult to understand or recognize. Similarly, if the edge pixel overlap ratio is insufficient to eliminate the color difference problem, the failure of edge fusion will cause obvious color faults in the spliced ​​area. This phenomenon is manifested as the colors of adjacent screens are completely different, forming an abrupt color block difference. For example, one part of the picture presents warm tones, while the other part is cold tones, which seriously destroys the visual coherence of the overall display and brings a bad viewing experience to the audience.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for controlling dynamic splicing of LED display screens. By combining the initial edge pixel overlap ratio with a deep learning model, the splicing situation of each sub-area is finely analyzed, and the area is intelligently divided into smooth transition and non-smooth transition areas, so as to optimize the problem areas in a centralized manner. For non-smooth transition areas, an adaptive edge fusion algorithm is used to dynamically adjust the edge pixel overlap ratio and re-fuse the pixels to ensure the visual smooth transition and consistency of the image, effectively solve the problems of misalignment and color difference in splicing, and improve the display effect of the splicing screen, especially in dynamic content and large-scale display systems, to improve the visual experience and system operation efficiency, so as to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a method for controlling dynamic splicing of an LED display screen, comprising the following steps:

[0009] For each frame of the pixel splicing area, the edge pixels of adjacent screens are overlapped using the initial edge pixel overlap ratio to preliminarily eliminate visible splicing traces;

[0010] After preliminary processing, the edge portion of the pixel stitching area is divided into several sub-areas;

[0011] For each sub-region, obtain its pixel splicing situation, use the pre-trained deep learning model to predict the pixel splicing situation of each sub-region, and evaluate the pixel splicing situation of the current sub-region;

[0012] According to the prediction results of the deep learning model, the processed sub-areas are divided into smooth transition areas and non-smooth transition areas. The smooth transition areas indicate that the stitching effect is normal and no further processing is required. The non-smooth transition areas indicate that there is obvious misalignment or color difference and optimization processing is required.

[0013] Mark the sub-regions identified as non-smooth transition regions and record their locations and severity;

[0014] For the marked non-smooth transition area, an adaptive edge fusion algorithm is used to dynamically adjust the edge pixel overlap ratio according to the initial edge pixel overlap ratio and the pixel splicing problem of the sub-area, and reallocate the adjusted edge pixel overlap ratio to the non-smooth transition area;

[0015] By adjusting and optimizing the edge pixel overlap ratio of the non-smooth transition area, the pixels of the marked area are fused again according to the adjusted edge pixel overlap ratio to ensure that the image is visually continuous and consistent in all areas, and finally achieve a smooth transition of pixel stitching.

[0016] Preferably, the edge portion of the splicing area is divided into a plurality of sub-areas, and the division method is based on a grid of a fixed size.

[0017] Preferably, for each sub-region, the spatial displacement difference of pixels at adjacent screen edges and the color gradient of pixels at adjacent screen edges are obtained, and after analyzing the spatial displacement difference of pixels at adjacent screen edges of the sub-region, an alignment difference reference value is generated, and the pixel alignment at adjacent screen edges is quantified by the alignment difference reference value; after analyzing the color gradient of pixels at adjacent screen edges of the sub-region, a color gradient reference value is generated, and the pixel difference at adjacent screen edges is quantified by the color gradient reference value; the obtained alignment difference reference value and color gradient reference value are input into a pre-trained deep learning model to generate a smooth transition prediction index, and the pixel splicing of each sub-region is predicted by the smooth transition prediction index to evaluate the pixel splicing of the current sub-region.

[0018] Preferably, the smooth transition prediction index generated by predicting the pixel splicing situation of the sub-area is compared and analyzed with a pre-set smooth transition prediction index reference threshold, and the processed sub-area is divided. The specific division steps are as follows: if the smooth transition prediction index is greater than or equal to the smooth transition prediction index reference threshold, the processed sub-area is divided into a non-smooth transition area; if the smooth transition prediction index is less than the smooth transition prediction index reference threshold, the processed sub-area is divided into a smooth transition area.

[0019] Preferably, for each sub-region, the spatial displacement difference of pixels at adjacent screen edges is obtained, and the specific steps of analyzing the spatial displacement difference of pixels at adjacent screen edges of the sub-region to generate the alignment difference reference value are as follows:

[0020] On the adjacent screen edge images of the sub-area, an advanced feature extraction algorithm is used to extract the edge images of screen A and screen B. and Process, extract and match feature points, and generate corresponding feature descriptors. The extracted expression is:

[0021] , , where i represents the index of the feature point extracted on screen A, and j represents the index of the feature point extracted on screen B. and They represent the coordinates of the feature points detected at the i-th feature point on screen A and the j-th feature point on screen B, respectively. and Respectively represent the descriptor vectors at the i-th feature point of screen A and the j-th feature point of screen B, which are used to match the feature points of the same image content. It is the feature extraction algorithm function;

[0022] According to the similarity of feature descriptors, the feature points of screen A are matched with the feature points of screen B. The matching expression is:

[0023] , where Represents the i-th feature point on screen B and screen A The position coordinates of the matched feature points, is the similarity measure between feature descriptors, Representation and Matched The index is calculated based on the descriptor similarity. MatchPairs represents the matching pair. Each element in the MatchPairs collection Represents the i-th feature point on screen A and the first Feature points The matching relationship;

[0024] Through the matching feature point pairs, the spatial displacement difference vectors of the adjacent screen edges are calculated, that is, the spatial offset of each pair of feature points between the two screens. The spatial offset reflects the degree of misalignment of the adjacent screen images in physical position. The calculation expression is: , is the displacement difference vector between the u-th pair of feature points, indicating the spatial offset of the edge feature points of screen A and screen B;

[0025] The structural tensor is used to analyze the changing characteristics of all displacement difference vectors, so as to capture the overall alignment of the splicing area. The structural tensor can represent the intensity of displacement differences in different directions and reflects the alignment consistency of the local splicing area. The expression of the structural tensor is: , where J is the structural tensor, which represents the overall change of the local area displacement difference vector field, represents the outer product of vectors, generating a symmetric matrix, is the weight of each pair of feature points;

[0026] The eigenvalue decomposition of the structural tensor is performed to extract the displacement difference direction and amplitude of the area. The decomposition expression is: ,in, , , where and are the two eigenvalues ​​of the structure tensor, which respectively represent the intensity of change in the main direction and the secondary direction of the displacement difference vector field. and is the characteristic vector of the change intensity in the main direction and the secondary direction, indicating the direction of change, ;

[0027] The alignment difference reference value is calculated using the eigenvalue of the structure tensor. The expression for the calculation is:

[0028] , where Indicates the alignment difference reference value, is a small positive number that prevents the eigenvalue from being zero.

[0029] Preferably, for each sub-region, the color gradient of the pixels at the adjacent screen edge is obtained, and the specific steps of analyzing the color gradient of the pixels at the adjacent screen edge of the sub-region to generate the color gradient reference value are as follows:

[0030] The color difference of pixels at adjacent screen edges in each sub-area is calculated as the basis for color gradient analysis. The calculation expression is:

[0031] , where represents the color difference value of the pth and qth pixels in the sub-region, , , is the RGB value of the pixel at the edge of the adjacent screen, , , is the RGB value of the corresponding pixel at the adjacent screen edge;

[0032] The color gradient vector in the sub-region is generated through the color difference value to quantify the color change direction and intensity in each sub-region. The expression for generating the color gradient vector is:

[0033] , where is the color gradient vector of the pth, qth pixel in the sub-region, and is the pixel coordinate vector of the corresponding pixel on screen A and screen B, is the spatial distance between two pixels;

[0034] All color gradient vectors in the sub-region are synthesized to generate the overall color gradient reference value of the sub-region, which is used to quantify the color smoothness of the region. The calculation expression of the color gradient reference value is: , where Represents the color gradient reference value of the sub-region, which is used to quantify the color change of the sub-region. is the modulus of the color gradient vector of the pth, qth pixel, indicating the intensity of the color change of the pixel. To adjust the nonlinear amplification factor of the color gradient to highlight significant color changes, is the weighting factor for each pixel in the sub-region.

[0035] Preferably, the sub-region identified as the non-smooth transition region is marked, and this process is achieved in the following manner:

[0036] First, the problematic pixel stitching in the sub-area is identified, and then a unique identifier is assigned to each non-smooth transition area, and its coordinate position on the screen is recorded. Next, the stitching problem in the sub-area is analyzed and its severity is quantified. Marking and recording can help the system focus on processing sub-areas with poor stitching quality.

[0037] Preferably, for the marked non-smooth transition area, the edge pixel overlap ratio is dynamically adjusted, and the adjusted edge pixel overlap ratio is reallocated to the non-smooth transition area. The specific steps are as follows:

[0038] Determine the correlation between the initial edge pixel overlap ratio and the smooth transition coefficient, and the expression determined is: , where is the initial edge pixel overlap ratio, which refers to the overlap ratio of adjacent screen edge pixels in the initial stage of splicing. is the adjustment coefficient, which is used to set the correlation strength between the initial overlap ratio and the smooth transition coefficient. It is a smooth transition coefficient, generated by a deep learning model, and is used to quantify the stitching effect of the current sub-region;

[0039] When a non-smooth transition area is identified, the original overlap ratio is adjusted by a dynamically calculated correction amount to further deal with the splicing problem. The calculation expression of the adaptive edge overlap correction amount is:

[0040] , where It is the adaptive edge overlap correction amount, which is used to correct the initial overlap ratio and increase the overlap ratio. is the correction coefficient, which controls the size of the adjustment. , is the preset smooth transition prediction index reference threshold, is the nonlinear adjustment coefficient, which is used to amplify the correction response when the splicing problem is serious;

[0041] Dynamically adjust the edge pixel overlap ratio. The expression for dynamic adjustment is:

[0042] , where is the adjusted edge pixel overlap ratio, which indicates the edge pixel overlap ratio finally applied in the non-smooth transition area. is the exponential attenuation coefficient, which is used to control the attenuation effect of the adaptive edge overlap correction amount, ensuring that the correction amount no longer excessively affects the overlap ratio after reaching a certain level. e is the natural exponent, which is used to introduce attenuation control;

[0043] Update the non-smooth transition area and perform edge blending. The updated expression is:

[0044] , , where Indicates the edge pixel position in the stitching area The pixel value after fusion is and Respectively indicate that screen A and screen B are at the edge The pixel value at , is a weight coefficient, used to represent the weighted ratio of adjacent screen edge pixels, and the weight coefficient is allocated based on the adjusted overlap ratio.

[0045] A dynamic splicing control system for an LED display screen, comprising an initial overlap processing module, a sub-region division module, a pixel splicing analysis module, a region classification module, a non-smooth transition marking module, an adaptive edge fusion module and a smooth optimization fusion module;

[0046] The initial overlap processing module overlaps edge pixels of adjacent screens using an initial edge pixel overlap ratio for each frame of the pixel splicing area, so as to preliminarily eliminate visible splicing traces;

[0047] The sub-region division module divides the edge portion of the pixel stitching region into several sub-regions after preliminary processing;

[0048] The pixel splicing analysis module obtains the pixel splicing status of each sub-region, uses a pre-trained deep learning model to predict the pixel splicing status of each sub-region, and evaluates the pixel splicing status of the current sub-region;

[0049] The region classification module divides the processed sub-regions into smooth transition regions and non-smooth transition regions according to the prediction results of the deep learning model. The smooth transition regions indicate that the stitching effect is normal and no further processing is required, while the non-smooth transition regions indicate that there is obvious misalignment or color difference and optimization processing is required;

[0050] A non-smooth transition marking module marks the sub-regions identified as non-smooth transition regions and records their locations and severity;

[0051] Adaptive edge fusion module, for the marked non-smooth transition area, adopts adaptive edge fusion algorithm, dynamically adjusts the edge pixel overlap ratio according to the initial edge pixel overlap ratio and the pixel splicing problem of the sub-area, and reallocates the adjusted edge pixel overlap ratio to the non-smooth transition area;

[0052] The smooth optimization fusion module adjusts and optimizes the edge pixel overlap ratio of the non-smooth transition area, and re-fuse the pixels of the marked area according to the adjusted edge pixel overlap ratio to ensure that the image is visually continuous and consistent in all areas, and finally achieves a smooth transition of pixel splicing.

[0053] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0054] The present invention uses the initial edge pixel overlap ratio to perform preliminary processing on the pixels of adjacent screens, and combines the prediction of the deep learning model to perform a refined pixel splicing analysis on each sub-region. According to the splicing situation, the system can intelligently divide the sub-region into a smooth transition region and a non-smooth transition region, ensuring that resources are concentrated on processing the most serious parts. For the marked non-smooth transition area, the system uses an adaptive edge fusion algorithm to dynamically adjust the edge pixel overlap ratio and re-fuse the pixels according to the initial edge pixel overlap ratio and splicing problem, ensuring that the image achieves visually smooth transition and consistency in all areas. Through this solution, the splicing problems caused by misalignment or color differences can be effectively solved, and the display effect of the spliced ​​screen can be improved. In particular, in dynamic content or large-scale display systems, precise adaptive splicing control is achieved, which greatly improves the overall visual experience and system operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0056] Figure 1 The present invention is a method flow chart of a method for dynamically splicing an LED display screen.

[0057] Figure 2 The present invention is a schematic diagram of a module of a dynamic splicing control system for an LED display screen. DETAILED DESCRIPTION

[0058] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0059] The present invention provides Figure 1 A method for controlling dynamic splicing of an LED display screen is shown, comprising the following steps:

[0060] For each frame of the pixel splicing area, the edge pixels of adjacent screens are overlapped using the initial edge pixel overlap ratio to preliminarily eliminate visible splicing traces;

[0061] This step uses existing edge fusion technology to achieve basic image transition by simply weighted averaging or linear blending pixels at adjacent screen edges. The preset overlap ratio (such as 5% or 10%) can effectively smooth the stitching area in most standard scenarios and reduce visual discontinuities caused by physical misalignment or slight color differences. This processing has low computational complexity and is suitable for systems with high real-time requirements.

[0062] After the preliminary processing, in order to achieve more refined optimization, the edge part of the pixel stitching area is divided into several sub-areas;

[0063] This division can be based on a fixed-size grid (such as 8×8 or 16×16 pixel blocks). Image block processing is a common method that facilitates parallel computing and local optimization. By refining to sub-regions, specific problems in different regions (such as large color differences or misalignment in some areas) can be accurately analyzed and processed.

[0064] Fixed-size grid division refers to dividing the entire stitching area into several rectangular or square areas of the same size according to uniform pixel blocks, usually 8×8 pixel blocks or 16×16 pixel blocks. The advantages of this division method are simplicity and efficiency, because the size and shape of each sub-region are uniform, which is convenient for parallel processing and calculation. The image processing system can analyze each sub-region independently without dynamically adjusting the shape of the sub-region. This method is particularly suitable for preliminary analysis in image stitching because it provides a basic block framework for subsequent image processing (such as alignment, color correction, etc.). In existing image processing technologies, fixed-size grid division is widely used in video encoding (such as H.264) and image compression to improve processing efficiency.

[0065] For each sub-region, obtain its pixel splicing situation, use the pre-trained deep learning model to predict the pixel splicing situation of each sub-region, and evaluate the pixel splicing situation of the current sub-region;

[0066] Deep learning models such as convolutional neural networks (CNN), recurrent neural networks (RNN) or integrated learning methods (such as random forests, XGBoost). These models learn the features under different splicing conditions through a large amount of training data, and can accurately predict whether the current sub-region has problems such as tearing and color difference. Deep learning models have mature applications in image recognition, quality assessment and anomaly detection, and can greatly improve the accuracy and efficiency of predictions.

[0067] For each sub-region, the spatial displacement difference of pixels at adjacent screen edges and the color gradient of pixels at adjacent screen edges are obtained. After analyzing the spatial displacement difference of pixels at adjacent screen edges of the sub-region, an alignment difference reference value is generated. The alignment of pixels at adjacent screen edges is quantified by the alignment difference reference value. After analyzing the color gradient of pixels at adjacent screen edges of the sub-region, a color gradient reference value is generated. The pixel difference at adjacent screen edges is quantified by the color gradient reference value. The obtained alignment difference reference value and color gradient reference value are input into a pre-trained deep learning model to generate a smooth transition prediction index. The pixel splicing of each sub-region is predicted by the smooth transition prediction index, and the pixel splicing of the current sub-region is evaluated.

[0068] A pre-trained deep learning model is one that is pre-trained using a large amount of image data sets, so that it can automatically learn and identify the potential features and patterns in the pixel stitching process. The model can be a convolutional neural network (CNN) or other image processing model. After training, it can generate a smooth transition prediction index based on the input alignment difference reference value and color gradient reference value, thereby predicting the smoothness of pixel stitching. During the training process, the model will learn how to identify alignment problems and color differences in image stitching through a large amount of annotated data, and form the ability to automatically evaluate the stitching effect of sub-regions. The trained model does not require human intervention and can make decisions adaptively based on the input data, improving the real-time processing efficiency of stitching quality.

[0069] When the spatial displacement difference of pixels at the edges of adjacent screens is large, it usually means that the pixel stitching effect is poor. This is because the spatial displacement difference reflects whether the pixel positions of two adjacent screens in the stitching area are aligned. When the displacement difference is large, it means that the edge pixels of the adjacent screens have been significantly offset in physical position, resulting in the image not being able to smoothly transition at the stitching point. This misalignment will cause screen tearing, image breakage or discontinuity, especially when playing dynamic content. The large spatial displacement difference makes it impossible to accurately connect the images in the stitching area, causing the audience to feel obvious visual discomfort when watching. For example, a moving object may appear split or incomplete when crossing the stitching point, which will destroy the smoothness and consistency of the overall picture. Therefore, a large spatial displacement difference is usually a direct manifestation of poor stitching effect.

[0070] For each sub-region, the spatial displacement difference of pixels at the adjacent screen edges is obtained, and the specific steps of analyzing the spatial displacement difference of pixels at the adjacent screen edges of the sub-region to generate an alignment difference reference value are as follows:

[0071] On the adjacent screen edge images of the sub-region, use an advanced feature extraction algorithm (such as SIFT or SURF) to extract the edge images of screen A and screen B. and Processing, extracting and matching feature points, these algorithms can capture the key points in adjacent screen images and generate corresponding feature descriptors. The extracted expressions are: , , where i represents the index of the feature point extracted on screen A, and j represents the index of the feature point extracted on screen B. and They represent the coordinates of the feature points detected at the i-th feature point on screen A and the j-th feature point on screen B, respectively. and Respectively represent the descriptor vectors at the i-th feature point of screen A and the j-th feature point of screen B, which are used to match the feature points of the same image content. It is a feature extraction algorithm function (such as SIFT, SURF);

[0072] According to the similarity of feature descriptors (such as cosine similarity), the feature points of screen A are matched with the feature points of screen B. The matching expression is:

[0073] , where Represents the i-th feature point on screen B and screen A The position coordinates of the matched feature points, is the similarity measure between feature descriptors, Representation and Matched The index is calculated based on the descriptor similarity. MatchPairs represents the matching pair. Each element in the MatchPairs collection Represents the i-th feature point on screen A and the first Feature points The matching relationship;

[0074] Through the matching feature point pairs, the spatial displacement difference vectors of the adjacent screen edges are calculated, that is, the spatial offset of each pair of feature points between the two screens. The spatial offset reflects the degree of misalignment of the adjacent screen images in physical position. The calculation expression is: , is the displacement difference vector between the u-th pair of feature points, indicating the spatial offset of the edge feature points of screen A and screen B;

[0075] The structural tensor is used to analyze the changing characteristics of all displacement difference vectors, so as to capture the overall alignment of the splicing area. The structural tensor can represent the intensity of displacement differences in different directions and reflects the alignment consistency of the local splicing area. The expression of the structural tensor is: , where J is the structural tensor, which represents the overall change of the local area displacement difference vector field, represents the outer product of vectors, generating a symmetric matrix, is the weight of each pair of feature points, which can be adjusted according to the confidence or stability of the feature points;

[0076] The eigenvalue decomposition of the structural tensor is performed to extract the displacement difference direction and amplitude of the area. The decomposition expression is: ,in, , , where and are the two eigenvalues ​​of the structure tensor, which respectively represent the intensity of change in the main direction and the secondary direction of the displacement difference vector field. and is the characteristic vector of the change intensity in the main direction and the secondary direction, indicating the direction of change, ;

[0077] The alignment difference reference value is calculated using the eigenvalue of the structure tensor. The expression for the calculation is:

[0078] , where Indicates the alignment difference reference value, It is a small positive number that prevents the eigenvalue from being zero. It is used to ensure the stability of the formula calculation. The preferred value is 0.001-0.01.

[0079] It can be seen from the alignment difference reference value that the larger the performance value of the alignment difference reference value generated after analyzing the spatial displacement difference of pixels at the adjacent screen edges of the sub-area, the larger the spatial displacement difference of adjacent screen edges, indicating that the splicing effect of the sub-area is poor, there is a more obvious misalignment or tearing phenomenon, and the image cannot transition smoothly. On the contrary, when the alignment difference reference value is smaller, it means that the spatial displacement difference is smaller, indicating that the pixels at the edge of the screen are better aligned, and the splicing effect is smoother and more natural. Therefore, the larger the alignment difference reference value, the worse the splicing effect; the smaller the alignment difference reference value, the better the splicing effect.

[0080] When the pixel color gradient at the edge of adjacent screens is large, it usually indicates that the pixel stitching effect is poor, because the color gradient reflects the degree of color change between adjacent pixels. If the color gradient is large, it means that the color transition between the two screens in the stitching area is not smooth, and the color change is too sudden or strong, resulting in a visual color fault or discontinuity. In a spliced ​​screen, the ideal splicing effect is that the colors of adjacent screens can blend naturally and gradually, and the audience cannot perceive the dividing line between the screens. On the contrary, when the color gradient is large, the colors of the splicing area appear inconsistent, and obvious color block differences may appear, especially when one screen displays warm tones and the adjacent screen displays cold tones. This color difference will appear very abrupt and destroy the overall visual coherence of the picture. Therefore, a large color gradient usually indicates that the colors of adjacent screens are not well matched or blended, resulting in poor splicing effects, which in turn affects the viewing experience.

[0081] For each sub-region, the color gradient of the pixels at the adjacent screen edge is obtained, and the specific steps of analyzing the color gradient of the pixels at the adjacent screen edge of the sub-region to generate the color gradient reference value are as follows:

[0082] The color difference of pixels at adjacent screen edges in each sub-area is calculated as the basis for color gradient analysis. The calculation expression is:

[0083] , where represents the color difference value of the pth and qth pixels in the sub-region, , , is the RGB value of the pixel at the edge of the adjacent screen (screen A), , , is the RGB value of the corresponding pixel at the adjacent screen edge (screen B);

[0084] This step calculates the color difference of two screen edge pixels based on the Euclidean distance, which directly reflects the color difference of adjacent pixels. The larger the color difference, the less smooth the color transition of the sub-area.

[0085] The color gradient vector in the sub-region is generated through the color difference value to quantify the color change direction and intensity in each sub-region. The expression for generating the color gradient vector is: , where is the color gradient vector of the pth, qth pixel in the sub-region, and is the pixel coordinate vector of the corresponding pixel on screen A and screen B, is the spatial distance between two pixels;

[0086] This step combines the color difference with the geometric information of the pixel position to generate a color gradient vector, which not only quantifies the magnitude of color change but also obtains the direction of color change, which helps to analyze the color transition characteristics of different regions.

[0087] All color gradient vectors in the sub-region are synthesized to generate the overall color gradient reference value of the sub-region, which is used to quantify the color smoothness of the region. The calculation expression of the color gradient reference value is: , where Represents the color gradient reference value of the sub-region, which is used to quantify the color change of the sub-region. is the modulus of the color gradient vector of the pth, qth pixel, indicating the intensity of the color change of the pixel. To adjust the nonlinear amplification factor of the color gradient, it is used to highlight significant color changes (for example, can amplify large color variations). A weighting factor for each pixel in the sub-region, assigning weights according to the importance of the region or changes in dynamic content;

[0088] Through this formula, the overall color change of the sub-region is quantified to generate a color gradient reference value. A high color gradient reference value indicates that the color change in the region is drastic and the stitching quality is poor; a low color gradient reference value indicates a smooth color transition.

[0089] It can be seen from the color gradient reference value that the larger the performance value of the color gradient reference value generated after analyzing the color gradient of pixels at the edges of adjacent screens in the sub-area, the worse the pixel stitching effect at the edges of adjacent screens in the sub-area. This is because the color gradient reference value measures the degree of color change between pixels at the edges of adjacent screens. The larger the performance value of the color gradient reference, the less smooth the color transition between adjacent screens, the more obvious the color difference, and the possible color faults or abrupt color block differences. Therefore, a larger color gradient reference value indicates that the visual incoherence of the stitching area is stronger and the stitching quality is poor. On the contrary, if the color gradient reference value is smaller, it means that the color transition in the area is smoother, the colors of adjacent screens are more consistent, the stitching effect is better, and the overall visual presentation is more coherent and natural.

[0090] The deep learning model is not specifically limited here, and can realize the alignment difference reference value generated in the sub-region and color gradient reference values Comprehensive analysis to generate smooth transition prediction index In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; the smooth transition prediction index The generated expression is:

[0091] , where , Alignment difference reference values and color gradient reference values The preset scaling factor of , Both are greater than 0.

[0092] It can be seen from the smooth transition prediction index that the larger the performance value of the alignment difference reference value generated after analyzing the spatial displacement difference of pixels at adjacent screen edges of the sub-area, the larger the performance value of the color gradient reference value generated after analyzing the color gradient of pixels at adjacent screen edges of the sub-area, that is, the larger the performance value of the smooth transition prediction index generated by predicting the pixel splicing situation of the sub-area, the worse the pixel splicing effect at adjacent screen edges of the sub-area, and vice versa.

[0093] According to the prediction results of the deep learning model, the processed sub-areas are divided into smooth transition areas and non-smooth transition areas. The smooth transition areas indicate that the stitching effect is normal and no further processing is required. The non-smooth transition areas indicate that there is obvious misalignment or color difference and optimization processing is required.

[0094] The smooth transition prediction index generated by predicting the pixel splicing of the sub-region is compared and analyzed with the preset smooth transition prediction index reference threshold, and the processed sub-region is divided. The specific division steps are as follows: if the smooth transition prediction index is greater than or equal to the smooth transition prediction index reference threshold, the processed sub-region is divided into a non-smooth transition region, and the non-smooth transition region indicates that there is obvious misalignment or color difference, and optimization processing is required; if the smooth transition prediction index is less than the smooth transition prediction index reference threshold, the processed sub-region is divided into a smooth transition region, and the smooth transition region indicates that the splicing effect is normal and no further processing is required. Such division enables the system to concentrate resources on processing the areas with the most serious problems and improve overall efficiency.

[0095] Mark the sub-regions identified as non-smooth transition regions and record their locations and severity;

[0096] For those sub-areas identified as non-smooth transition areas (i.e., areas with poor stitching effects), they need to be marked to record the specific positions of these sub-areas on the screen and the severity of the stitching problems.

[0097] This can be achieved by:

[0098] First, the pixel stitching problems in the sub-area (such as misalignment or color difference) are identified, and then a unique identifier is assigned to each non-smooth transition area, and its coordinate position on the screen is recorded. Then, by analyzing the stitching problem of the sub-area, its severity is quantified (such as the smooth transition prediction index), and this data is stored in the system for subsequent optimization processing. This marking and recording can help the system focus on processing sub-areas with poor stitching quality and improve the overall display effect.

[0099] For the marked non-smooth transition area, an adaptive edge fusion algorithm is used to dynamically adjust the edge pixel overlap ratio according to the initial edge pixel overlap ratio and the pixel splicing problem of the sub-area, and reallocate the adjusted edge pixel overlap ratio to the non-smooth transition area;

[0100] For the marked non-smooth transition area, the edge pixel overlap ratio is dynamically adjusted, and the adjusted edge pixel overlap ratio is redistributed to the non-smooth transition area. The specific steps are as follows:

[0101] Determine the correlation between the initial edge pixel overlap ratio and the smooth transition coefficient, and the expression determined is:

[0102] , where is the initial edge pixel overlap ratio, which refers to the overlap ratio of adjacent screen edge pixels in the initial stage of splicing. is the adjustment coefficient, which is used to set the correlation strength between the initial overlap ratio and the smooth transition coefficient. is the smooth transition coefficient, which is generated by the deep learning model and is used to quantify the splicing effect of the current sub-region;

[0103] The initial edge pixel overlap ratio is set to achieve basic edge transition effects during the stitching process, but its setting may not be able to handle complex non-smooth transition areas. Therefore, it is necessary to quantify the stitching quality of each sub-area through a smooth transition coefficient and associate it with the initial overlap ratio to help subsequent optimization processing.

[0104] When a non-smooth transition area is identified, the original overlap ratio is adjusted by a dynamically calculated correction to further deal with the splicing problem. The key to this step is to ensure that the poor splicing effect is significantly corrected by a nonlinear amplification response. The calculation expression of the adaptive edge overlap correction is: , where It is the adaptive edge overlap correction amount, which is used to correct the initial overlap ratio and increase the overlap ratio. It is the correction coefficient, which controls the size of the adjustment amount. It is usually used to adjust the influence of the smooth transition coefficient on the correction. , is the preset smooth transition prediction index reference threshold, It is a nonlinear adjustment coefficient, which is used to amplify the correction response when the splicing problem is serious, ensuring that the serious problem can be adequately corrected by overlap;

[0105] Dynamically adjust the edge pixel overlap ratio. The expression for dynamic adjustment is:

[0106] , where is the adjusted edge pixel overlap ratio, which indicates the edge pixel overlap ratio finally applied in the non-smooth transition area. is an exponential attenuation coefficient, which is used to control the attenuation effect of the adaptive edge overlap correction, ensuring that the correction no longer excessively affects the overlap ratio after reaching a certain level. e is a natural exponent, which is used to introduce attenuation control so that the correction effect gradually decreases as the value increases.

[0107] The overlap ratio of edge pixels is dynamically adjusted to ensure that the stitching effect can adapt to the actual problem.

[0108] Update the non-smooth transition area and perform edge blending. The updated expression is:

[0109] , , where Indicates the edge pixel position in the stitching area The pixel value after fusion at the position represents the final visual effect at that position. and Respectively indicate that screen A and screen B are at the edge The pixel value at , is a weight coefficient, used to represent the weighted ratio of adjacent screen edge pixels, and the weight coefficient is allocated based on the adjusted overlap ratio;

[0110] By redistributing the adjusted edge overlap ratio, the system can achieve a more natural edge transition in the splicing area. The edge fusion algorithm in this step can weightedly fuse the pixels of two adjacent screens to eliminate misalignment and color fault problems. Through this step, the system can fuse the edge pixels of two adjacent screens and use the adjusted overlap ratio to distribute the weights, making the pixel transition in the splicing area smoother and more natural. Ultimately, this step can effectively reduce the tearing and color fault problems at the splicing point and improve the overall visual coherence.

[0111] Specifically, the edge pixel overlap ratio is expanded according to the initial edge pixel overlap ratio and the pixel stitching problem of the sub-region; for example, if the initial overlap ratio is 5%, it can be increased to 10% or 15% for sub-regions with serious problems. Increasing the number of pixels in the overlapping area provides more space for pixel fusion, which can more effectively smooth out misalignment and color differences. Adaptive edge fusion methods are often used to dynamically adjust the overlap ratio to achieve more refined image transitions.

[0112] By adjusting and optimizing the edge pixel overlap ratio of the non-smooth transition area, the pixels of the marked area are re-fused according to the adjusted edge pixel overlap ratio to ensure that the image is visually continuous and consistent in all areas, and finally achieve a smooth transition of pixel splicing;

[0113] By dynamically adjusting the overlap ratio of edge pixels, the fusion effect of pixels can be controlled more accurately. In the area marked as problematic, the adjusted edge pixels will be re-fused, that is, the images of the two adjacent screens will be combined according to the new overlap ratio. The purpose of this is to ensure that the images of all sub-areas can achieve visual continuity and consistency, that is, both the alignment of the image and the color transition look more natural and harmonious. Finally, by processing these problematic areas, a smooth transition of the entire spliced ​​screen can be achieved. In this way, when watching the spliced ​​screen, the audience will not notice obvious boundaries or faults, and the entire image will look seamless and smooth in the spliced ​​area, which improves the overall visual experience. This adjustment and optimization process is particularly important for those large display screens or dynamic video playback application scenarios, because it ensures high-quality display effects in the spliced ​​area.

[0114] The present invention uses the initial edge pixel overlap ratio to perform preliminary processing on the pixels of adjacent screens, and combines the prediction of the deep learning model to perform a refined pixel splicing analysis on each sub-region. According to the splicing situation, the system can intelligently divide the sub-region into a smooth transition region and a non-smooth transition region, ensuring that resources are concentrated on processing the most serious parts. For the marked non-smooth transition area, the system uses an adaptive edge fusion algorithm to dynamically adjust the edge pixel overlap ratio and re-fuse the pixels according to the initial edge pixel overlap ratio and splicing problem, ensuring that the image achieves visually smooth transition and consistency in all areas. Through this solution, the splicing problems caused by misalignment or color differences can be effectively solved, and the display effect of the spliced ​​screen can be improved. In particular, in dynamic content or large-scale display systems, precise adaptive splicing control is achieved, which greatly improves the overall visual experience and system operation efficiency.

[0115] The present invention provides Figure 2A dynamic splicing control system for an LED display screen is shown, comprising an initial overlap processing module, a sub-region division module, a pixel splicing analysis module, a region classification module, a non-smooth transition marking module, an adaptive edge fusion module, and a smooth optimization fusion module;

[0116] The initial overlap processing module overlaps edge pixels of adjacent screens using an initial edge pixel overlap ratio for each frame of the pixel splicing area, so as to preliminarily eliminate visible splicing traces;

[0117] The sub-region division module divides the edge portion of the pixel stitching region into several sub-regions after preliminary processing;

[0118] The pixel splicing analysis module obtains the pixel splicing status of each sub-region, uses a pre-trained deep learning model to predict the pixel splicing status of each sub-region, and evaluates the pixel splicing status of the current sub-region;

[0119] The region classification module divides the processed sub-regions into smooth transition regions and non-smooth transition regions according to the prediction results of the deep learning model. The smooth transition regions indicate that the stitching effect is normal and no further processing is required, while the non-smooth transition regions indicate that there is obvious misalignment or color difference and optimization processing is required;

[0120] A non-smooth transition marking module marks the sub-regions identified as non-smooth transition regions and records their locations and severity;

[0121] Adaptive edge fusion module, for the marked non-smooth transition area, adopts adaptive edge fusion algorithm, dynamically adjusts the edge pixel overlap ratio according to the initial edge pixel overlap ratio and the pixel splicing problem of the sub-area, and reallocates the adjusted edge pixel overlap ratio to the non-smooth transition area;

[0122] The smooth optimization fusion module adjusts and optimizes the edge pixel overlap ratio of the non-smooth transition area, and re-fuse the pixels of the marked area according to the adjusted edge pixel overlap ratio to ensure that the image is visually continuous and consistent in all areas, and finally achieves a smooth transition of pixel splicing;

[0123] An LED display screen dynamic splicing control method provided by an embodiment of the present invention is implemented by the above-mentioned LED display screen dynamic splicing control system. The specific method and process of an LED display screen dynamic splicing control system are detailed in the embodiment of the above-mentioned LED display screen dynamic splicing control method, which will not be repeated here.

[0124] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0125] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0126] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for controlling dynamic splicing of an LED display screen, characterized in that: The following steps are involved: For each frame of the pixel splicing area, the edge pixels of adjacent screens are overlapped using the initial edge pixel overlap ratio to preliminarily eliminate visible splicing traces; After preliminary processing, the edge portion of the pixel stitching area is divided into several sub-areas; For each sub-region, obtain its pixel splicing situation, use the pre-trained deep learning model to predict the pixel splicing situation of each sub-region, and evaluate the pixel splicing situation of the current sub-region; According to the prediction results of the deep learning model, the processed sub-areas are divided into smooth transition areas and non-smooth transition areas. The smooth transition areas indicate that the stitching effect is normal and no further processing is required. The non-smooth transition areas indicate that there is obvious misalignment or color difference and optimization processing is required. Mark the sub-regions identified as non-smooth transition regions and record their locations and severity; For the marked non-smooth transition area, an adaptive edge fusion algorithm is used to dynamically adjust the edge pixel overlap ratio according to the initial edge pixel overlap ratio and the pixel splicing problem of the sub-area, and reallocate the adjusted edge pixel overlap ratio to the non-smooth transition area; Dynamically adjust the edge pixel overlap ratio. The expression for dynamic adjustment is: R adjusted =(R initial +ΔR)·e -λ·ΔR , where R initial is the initial edge pixel overlap ratio, which refers to the overlap ratio of adjacent screen edge pixels in the initial stage of splicing. ΔR is the adaptive edge overlap correction, which is used to correct the initial overlap ratio and increase the overlap ratio. adjusted is the adjusted edge pixel overlap ratio, which indicates the edge pixel overlap ratio finally applied in the non-smooth transition area, λ is the exponential attenuation coefficient, which is used to control the attenuation effect of the adaptive edge overlap correction amount, and e is the natural exponent, which is used to introduce attenuation control; Update the non-smooth transition area and perform edge fusion. The updated expression is: fusion (x, y)=w1·I1(x, y)+w2·I2(x, y), In the formula, I fusion (x, y) represents the fused pixel value at the edge pixel position (x, y) of the stitching area, I1(x, y) and I2(x, y) represent the pixel values ​​of screen A and screen B at the edge position (x, y), w1 and w2 are weight coefficients, which are used to represent the weighted ratio of the edge pixels of adjacent screens, and the weight coefficients are allocated based on the adjusted overlap ratio; By adjusting and optimizing the edge pixel overlap ratio of the non-smooth transition area, the pixels of the marked area are fused again according to the adjusted edge pixel overlap ratio to ensure that the image is visually continuous and consistent in all areas, and finally achieve a smooth transition of pixel stitching.

2. The LED display screen dynamic splicing control method according to claim 1 is characterized in that: The edge of the splicing area is divided into several sub-areas based on a fixed-size grid.

3. The LED display screen dynamic splicing control method according to claim 1 is characterized in that: For each sub-region, the spatial displacement difference of pixels at adjacent screen edges and the color gradient of pixels at adjacent screen edges are obtained. After analyzing the spatial displacement difference of pixels at adjacent screen edges of the sub-region, an alignment difference reference value is generated. The alignment of pixels at adjacent screen edges is quantified by the alignment difference reference value. After analyzing the color gradient of pixels at adjacent screen edges of the sub-region, a color gradient reference value is generated. The pixel difference at adjacent screen edges is quantified by the color gradient reference value. The obtained alignment difference reference value and color gradient reference value are input into a pre-trained deep learning model to generate a smooth transition prediction index. The pixel splicing of each sub-region is predicted by the smooth transition prediction index, and the pixel splicing of the current sub-region is evaluated.

4. The LED display screen dynamic splicing control method according to claim 1 is characterized in that: The smooth transition prediction index generated by predicting the pixel splicing situation of the sub-area is compared and analyzed with the pre-set smooth transition prediction index reference threshold, and the processed sub-area is divided. The specific division steps are as follows: if the smooth transition prediction index is greater than or equal to the smooth transition prediction index reference threshold, the processed sub-area is divided into a non-smooth transition area; if the smooth transition prediction index is less than the smooth transition prediction index reference threshold, the processed sub-area is divided into a smooth transition area.

5. The LED display screen dynamic splicing control method according to claim 1 is characterized in that: For each sub-region, the spatial displacement difference of pixels at the adjacent screen edges is obtained, and the specific steps of analyzing the spatial displacement difference of pixels at the adjacent screen edges of the sub-region to generate an alignment difference reference value are as follows: On the adjacent screen edge images of the sub-region, an advanced feature extraction algorithm is used to extract the edge images I of screen A and screen B. A and I B Process, extract and match feature points, and generate corresponding feature descriptors. The extracted expression is: Where i represents the index of the feature point extracted on screen A, j represents the index of the feature point extracted on screen B, and They represent the coordinates of the feature points detected at the i-th feature point on screen A and the j-th feature point on screen B, respectively. and They represent the descriptor vectors at the i-th feature point on screen A and the j-th feature point on screen B, respectively, and are used to match feature points with the same image content. FeatureExtraction is a feature extraction algorithm function. According to the similarity of feature descriptors, the feature points of screen A are matched with the feature points of screen B. The matching expression is: In the formula, Indicates the i-th feature point on screen B and screen A The position coordinates of the matched feature points, is the similarity measure between feature descriptors, π(i) represents the similarity with Matched The index is calculated based on the descriptor similarity. MatchPairs represents the matching pair. Each element in the MatchPairs collection Represents the i-th feature point on screen A and the π(i)th feature point on screen B The matching relationship; Through the matching feature point pairs, the spatial displacement difference vectors of the adjacent screen edges are calculated, that is, the spatial offset of each pair of feature points between the two screens. The spatial offset reflects the degree of misalignment of the adjacent screen images in physical position. The calculation expression is: Δd u is the displacement difference vector between the u-th pair of feature points, representing the spatial offset of the edge feature points of screen A and screen B; The structural tensor is used to analyze the changing characteristics of all displacement difference vectors, so as to capture the overall alignment of the splicing area. The structural tensor can represent the intensity of displacement differences in different directions and reflects the alignment consistency of the local splicing area. The expression of the structural tensor is: Where J is the structural tensor, which represents the overall change of the local area displacement difference vector field. represents the outer product of the vectors, generating a symmetric matrix, w u is the weight of each pair of feature points; The eigenvalue decomposition of the structure tensor is performed to extract the displacement difference direction and amplitude of the region. The decomposition expression is: Je k =λ k e k , where λ k is the eigenvalue of the structure tensor, expressed in the eigenvector e k The intensity of the change in the indicated direction, e k is the eigenvector, Je1=λ1e1, Je2=λ2e2, where λ1 and λ2 are the two eigenvalues ​​of the structure tensor, representing the change intensity in the main direction and the secondary direction of the displacement difference vector field, respectively, e1 and e2 are the eigenvectors of the change intensity in the main direction and the secondary direction, representing the direction of change, λ1>λ2; The alignment difference reference value is calculated using the eigenvalue of the structure tensor. The expression for the calculation is: In the formula, Alignment DI represents the alignment difference reference value, ∈ is a small positive number that prevents the eigenvalue from being zero.

6. The LED display screen dynamic splicing control method according to claim 5 is characterized in that: For each sub-region, the color gradient of the pixels at the adjacent screen edge is obtained, and the specific steps of analyzing the color gradient of the pixels at the adjacent screen edge of the sub-region to generate the color gradient reference value are as follows: Calculate the color difference of pixels at adjacent screen edges in each sub-area as the basis for color gradient analysis. The calculation expression is: Where, ΔC pq represents the color difference value of the pth and qth pixels in the sub-area, R1, G1, B1 are the RGB values ​​of the pixels at the adjacent screen edges, and R2, G2, B2 are the RGB values ​​of the corresponding pixels at the adjacent screen edges; The color gradient vector in the sub-region is generated through the color difference value to quantify the color change direction and intensity in each sub-region. The expression for generating the color gradient vector is: In the formula, is the color gradient vector of the pth, qth pixel in the sub-region, and is the pixel coordinate vector of the corresponding pixel on screen A and screen B, is the spatial distance between two pixels; All color gradient vectors in the sub-region are synthesized to generate the overall color gradient reference value of the sub-region, which is used to quantify the color smoothness of the region. The calculation expression of the color gradient reference value is: In the formula, I CG Represents the color gradient reference value of the sub-region, which is used to quantify the color change of the sub-region. is the modulus of the color gradient vector of the pth, qth pixel, indicating the color change intensity of the pixel, δ is the nonlinear amplification factor for adjusting the color gradient, used to highlight significant color changes, and W pg is the weighting factor for each pixel in the sub-region.

7. The LED display screen dynamic splicing control method according to claim 1, characterized in that: The sub-regions identified as non-smooth transition regions are marked. This is achieved in the following way: First, the problematic pixel stitching in the sub-area is identified, and then a unique identifier is assigned to each non-smooth transition area, and its coordinate position on the screen is recorded. Next, the stitching problem in the sub-area is analyzed and its severity is quantified. Marking and recording can help the system focus on processing sub-areas with poor stitching quality.

8. The LED display screen dynamic splicing control method according to claim 4, characterized in that: For the marked non-smooth transition area, the edge pixel overlap ratio is dynamically adjusted, and the adjusted edge pixel overlap ratio is redistributed to the non-smooth transition area. The specific steps are as follows: Determine the relationship between the initial edge pixel overlap ratio and the smooth transition coefficient, and the expression is: R initial =α·Smooth transition , where α is the adjustment coefficient, which is used to set the correlation strength between the initial overlap ratio and the smooth transition coefficient. transition It is a smooth transition coefficient, generated by a deep learning model, and is used to quantify the stitching effect of the current sub-region; When a non-smooth transition area is identified, the original overlap ratio is adjusted by a dynamically calculated correction amount to further deal with the splicing problem. The calculation expression of the adaptive edge overlap correction amount is: ΔR = γ·(Smooth transition β -Smooth ref β ), where γ is the correction coefficient, which controls the size of the adjustment, Smooth transition >Smooth ref , Smooth ref is the preset reference threshold of the smooth transition prediction index, and β is the nonlinear adjustment coefficient, which is used to amplify the correction response when the splicing problem is serious.

9. A LED display screen dynamic splicing control system, used to implement the LED display screen dynamic splicing control method described in any one of claims 1 to 8, characterized in that: It includes an initial overlap processing module, a sub-region division module, a pixel splicing analysis module, a region classification module, a non-smooth transition marking module, an adaptive edge fusion module, and a smooth optimization fusion module; The initial overlap processing module overlaps edge pixels of adjacent screens using an initial edge pixel overlap ratio for each frame of the pixel splicing area, so as to preliminarily eliminate visible splicing traces; The sub-region division module divides the edge portion of the pixel stitching region into several sub-regions after preliminary processing; The pixel splicing analysis module obtains the pixel splicing status of each sub-region, uses a pre-trained deep learning model to predict the pixel splicing status of each sub-region, and evaluates the pixel splicing status of the current sub-region; The region classification module divides the processed sub-regions into smooth transition regions and non-smooth transition regions according to the prediction results of the deep learning model. The smooth transition regions indicate that the stitching effect is normal and no further processing is required, while the non-smooth transition regions indicate that there is obvious misalignment or color difference and optimization processing is required; A non-smooth transition marking module marks the sub-regions identified as non-smooth transition regions and records their locations and severity; Adaptive edge fusion module, for the marked non-smooth transition area, adopts adaptive edge fusion algorithm, dynamically adjusts the edge pixel overlap ratio according to the initial edge pixel overlap ratio and the pixel splicing problem of the sub-area, and reallocates the adjusted edge pixel overlap ratio to the non-smooth transition area; The smooth optimization fusion module adjusts and optimizes the edge pixel overlap ratio of the non-smooth transition area, and re-fuse the pixels of the marked area according to the adjusted edge pixel overlap ratio to ensure that the image is visually continuous and consistent in all areas, and finally achieves a smooth transition of pixel splicing.

Images