Special-shaped LED pixel mapping method and system based on video content analysis

Abstract

The application relates to the technical field of image processing, in particular to a special-shaped LED pixel mapping method and system based on video content analysis, which comprises the following steps: performing scene classification on to-be-mapped video frame data to obtain scene category information; analyzing the spatial distribution of the video frame based on the scene category to extract edge contour features of a salient target area; constructing a non-uniform grid topology structure according to the spatial position coordinates of the pixel units of a special-shaped LED display screen; performing spatial position registration on the edge contour features and the non-uniform grid topology structure, and according to the spatial mapping relationship obtained after the registration, mapping each pixel point in the to-be-mapped video frame data to the corresponding grid node in the non-uniform grid topology structure to generate an initial mapping pixel matrix; correcting the pixel value according to the spatial neighborhood relationship of the grid nodes and outputting to the special-shaped LED screen for display; and the application can improve the efficiency of pixel mapping.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and system for mapping irregularly shaped LED pixels based on video content analysis. Background Technology

[0002] Irregularly shaped LED displays, with their flexible form and strong visual impact, are widely used in commercial exhibitions, cultural tourism performances, and urban landscapes. Current mainstream pixel mapping solutions are mostly based on regular rectangular grid models to construct mapping relationships, directly stretching and adapting standard video images to the display area proportionally. Because the pixel units of irregularly shaped LEDs are non-uniformly arranged and the display substrate outline is irregular, this mapping method easily causes spatial misalignment between the image content and the physical pixels, resulting in jagged edges and shape distortion of the main content. Furthermore, a uniform mapping logic cannot adapt to the visual weight distribution of different content, easily leading to the loss of local details or uneven brightness, making it difficult to achieve high-quality image presentation.

[0003] Existing pixel mapping optimization solutions for irregularly shaped screens mostly adjust mapping coordinates only from the perspective of hardware layout, without adaptive adjustments based on the semantic features of video content, resulting in insufficient adaptability to salient areas of the image. Some mapping methods that incorporate content analysis use general saliency detection logic without differentiated analysis for different scene categories, leading to limited accuracy in extracting target edge contours and large spatial registration errors. Furthermore, the pixel correction stage often employs a fixed neighborhood weight strategy, failing to fully consider the actual spatial distance differences between pixel units on the irregularly shaped substrate, resulting in overly harsh images and poor color uniformity. Overall, the mapping processing efficiency and display effect have significant shortcomings, failing to meet the high-precision and high-real-time requirements of irregularly shaped displays. Summary of the Invention

[0004] This invention provides a method and system for mapping irregularly shaped LED pixels based on video content analysis, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for mapping irregularly shaped LED pixels based on video content analysis, comprising:

[0006] Q1. Perform scene classification on the video frame data to be mapped to obtain the scene category information of the video frame data to be mapped;

[0007] Q2. Based on the scene category information, perform spatial distribution analysis on the video frame data to be mapped to obtain the salient target region and the corresponding spatial distribution information of the video frame data to be mapped, and determine the edge contour features of the salient target region according to the spatial distribution information.

[0008] Q3. Based on the spatial coordinates of each pixel unit on the irregular substrate in the pixel arrangement structure data of the irregular LED display, construct the non-uniform grid topology structure of the irregular LED display.

[0009] Q4. Spatial registration is performed between the edge contour features and the non-uniform grid topology, and according to the spatial mapping relationship obtained after registration, each pixel in the video frame data to be mapped is mapped to the corresponding grid node in the non-uniform grid topology to generate the initial mapping pixel matrix of the video frame data to be mapped.

[0010] Q5. Based on the actual spatial neighborhood relationship of each grid node in the initial mapped pixel matrix, perform pixel value correction on the initial mapped pixel matrix, and output the corrected mapped pixel matrix to the irregular LED display screen for display.

[0011] In a preferred embodiment, the step of classifying the video frame data to be mapped into scenes to obtain scene category information of the video frame data to be mapped includes:

[0012] The video frame data to be mapped is obtained, and the color distribution features, texture roughness features and edge density features of the video frame data to be mapped are fused at the feature level to obtain the comprehensive feature vector of the video frame data to be mapped.

[0013] The comprehensive feature vector is matched with the standard feature vectors corresponding to each scene category in the preset scene category library to determine the matching priority between the comprehensive feature vector and the standard feature vector;

[0014] The scene category corresponding to the standard feature vector with the highest matching priority is selected as the scene category information of the video frame data to be mapped.

[0015] In a preferred embodiment, the step of performing spatial distribution analysis on the video frame data to be mapped based on the scene category information to obtain the salient target region and corresponding spatial distribution information of the video frame data to be mapped includes:

[0016] The region of interest corresponding to the scene category information is delineated from the video frame data to be mapped, and the spatial resolution scale range of the video frame data to be mapped is determined based on the horizontal and vertical spans of the region of interest in the video frame data to be mapped.

[0017] Based on the spatial resolution scale range, spatial saliency response evaluation is performed on each pixel region of the video frame data to be mapped, and a saliency spatial distribution map of the video frame data to be mapped is constructed.

[0018] Connectivity regions are labeled in the saliency spatial distribution map to identify high-response connected regions in the saliency spatial distribution map as saliency target regions of the video frame data to be mapped;

[0019] Based on the position boundary of the salient target region in the salient spatial distribution map, the spatial position coordinates and spatial distribution range of the salient target region in the video frame data to be mapped are obtained as the spatial distribution information corresponding to the salient target region.

[0020] In a preferred embodiment, determining the edge contour features of the salient target region based on the spatial distribution information includes:

[0021] Based on the spatial coordinates, locate the starting boundary point of the salient target region from the video frame data to be mapped;

[0022] Based on the spatial distribution range, the contour scanning span of the salient target region is determined, and according to the contour scanning span, starting from the starting boundary point, the boundary trajectory is traced point by point along the boundary direction of the salient target region to obtain the boundary trajectory point sequence of the salient target region.

[0023] The boundary trajectory point sequence is fitted with a contour line to obtain the edge contour features of the salient target region.

[0024] In a preferred embodiment, constructing the non-uniform grid topology of the irregularly shaped LED display screen based on the spatial position coordinates of each pixel unit on the irregularly shaped substrate in the pixel arrangement structure data of the irregularly shaped LED display screen includes:

[0025] Obtain the pixel arrangement structure data of the irregular LED display screen, and extract the spatial position coordinates of each pixel unit on the irregular substrate from the pixel arrangement structure data;

[0026] Based on the spatial coordinates, each pixel unit is mapped to a corresponding coordinate point in a two-dimensional coordinate plane to generate the pixel coordinate matrix of the irregular LED display screen.

[0027] Based on the actual spatial distance between adjacent coordinate points in the pixel coordinate matrix, the neighborhood association relationship between each coordinate point in the pixel coordinate matrix is ​​determined.

[0028] Based on the neighborhood association relationship, connect the coordinate points in the pixel coordinate matrix that have the neighborhood association relationship to construct the non-uniform grid topology of the irregular LED display screen.

[0029] In a preferred embodiment, the step of spatially registering the edge contour features with the non-uniform mesh topology, and mapping each pixel in the video frame data to be mapped to the corresponding mesh node in the non-uniform mesh topology according to the spatial mapping relationship obtained after registration, to generate an initial mapped pixel matrix of the video frame data to be mapped, includes:

[0030] Based on the direction change of the edge contour features, the edge contour features are divided into multiple contour segments;

[0031] Based on the local clustering density of each grid node in the non-uniform grid topology, the non-uniform grid topology is divided into multiple grid sub-regions;

[0032] The spatial overlap between the spatial distribution range of the contour segments and the spatial distribution range of the grid sub-regions is compared, and the correspondence between the contour segments and the grid sub-regions is established based on the comparison results.

[0033] According to the correspondence, each contour point on each contour segment is bound to each grid node in the corresponding grid sub-region in order of spatial position.

[0034] The pixel values ​​at each contour point in the video frame data to be mapped are assigned to the corresponding grid nodes to generate the initial mapped pixel matrix of the video frame data to be mapped.

[0035] In a preferred embodiment, the step of correcting the pixel values ​​of the initial mapped pixel matrix based on the actual spatial neighborhood relationships of each grid node in the initial mapped pixel matrix, and outputting the corrected mapped pixel matrix to the irregularly shaped LED display screen for display, includes:

[0036] Obtain the actual spatial neighborhood relationship of the grid nodes in the initial mapped pixel matrix on the irregular substrate;

[0037] Based on the actual spatial neighborhood relationship, the current pixel value of the grid node and the neighboring pixel values ​​of the adjacent grid nodes are extracted from the initial mapped pixel matrix;

[0038] Using the actual spatial distance between the grid node and the adjacent grid node as a constraint, the neighboring pixel values ​​are weighted and allocated to obtain the corrected pixel values ​​of the grid node.

[0039] The current pixel value of the grid node in the initial mapped pixel matrix is ​​replaced with the corrected pixel value. After traversing all grid nodes in the initial mapped pixel matrix, the corrected mapped pixel matrix of the initial mapped pixel matrix is ​​generated.

[0040] The corrected mapped pixel matrix is ​​subjected to display frame encapsulation processing to obtain the display data stream of the irregularly shaped LED display screen;

[0041] The display data stream is sent to the irregularly shaped LED display screen for display.

[0042] To address the aforementioned problems, this invention also provides a non-standard LED pixel mapping system based on video content analysis, the system comprising:

[0043] The scene classification module is used to classify the video frame data to be mapped into scenes, and obtain the scene category information of the video frame data to be mapped.

[0044] The spatial distribution parsing module is used to perform spatial distribution parsing on the video frame data to be mapped based on the scene category information, to obtain the salient target region of the video frame data to be mapped and the corresponding spatial distribution information, and to determine the edge contour features of the salient target region based on the spatial distribution information.

[0045] The non-uniform mesh construction module is used to construct the non-uniform mesh topology of the irregular LED display based on the spatial position coordinates of each pixel unit on the irregular substrate in the pixel arrangement structure data of the irregular LED display.

[0046] The registration and mapping module is used to spatially register the edge contour features with the non-uniform grid topology, and according to the spatial mapping relationship obtained after registration, map each pixel in the video frame data to be mapped to the corresponding grid node in the non-uniform grid topology to generate the initial mapping pixel matrix of the video frame data to be mapped.

[0047] The correction output module is used to correct the pixel values ​​of the initial mapped pixel matrix according to the actual spatial neighborhood relationship of each grid node in the initial mapped pixel matrix, and output the corrected mapped pixel matrix to the irregular LED display screen for display.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. This invention, through scene classification and spatial distribution analysis of video frame data, can accurately locate salient target areas in the image, extract clear and accurate edge contour features, and achieve precise spatial registration of video content features with irregular LED pixel structures. The semantic-based mapping logic ensures a high degree of fit between the core content of the image and the physical pixel spatial position, fully adapting to the irregular contours of irregular display substrates, guaranteeing the morphological integrity of the main subject and the accuracy of edge rendering, effectively improving the image fidelity and visual expressiveness of video content on irregular LED screens, and adapting to the content presentation needs of various irregular display scenarios.

[0050] 2. This invention constructs a non-uniform grid topology based on the actual spatial positions of irregularly shaped LED pixels, establishing a spatial mapping relationship that fits the hardware layout. Combined with a pixel value weighting correction mechanism based on actual spatial neighborhood relationships, it can smoothly transition the color and brightness of different pixel spacing areas, improving the uniformity and image depth of the entire screen display. The entire mapping process is clear and the computational logic is efficient, compatible with various pixel arrangement forms of irregularly shaped LED displays. While ensuring high-quality image output, it improves the processing efficiency of pixel mapping, reduces the adjustment cost for adapting to different irregularly shaped screen structures, and provides stable, high-quality image output support for irregularly shaped LED display systems. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating a method for mapping irregularly shaped LED pixels based on video content analysis, provided in an embodiment of the present invention.

[0052] Figure 2 A functional block diagram of an irregularly shaped LED pixel mapping system based on video content analysis provided in an embodiment of the present invention;

[0053] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0054] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0055] This application provides a method for mapping irregularly shaped LED pixels based on video content analysis. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for mapping irregularly shaped LED pixels based on video content analysis can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0056] Reference Figure 1The diagram shown is a flowchart illustrating a method for mapping irregularly shaped LED pixels based on video content analysis according to an embodiment of the present invention. In this embodiment, the method for mapping irregularly shaped LED pixels based on video content analysis includes:

[0057] Q1. Perform scene classification on the video frame data to be mapped to obtain the scene category information of the video frame data to be mapped;

[0058] In this embodiment of the invention, the step of classifying the video frame data to be mapped into scenes to obtain scene category information of the video frame data to be mapped includes:

[0059] The video frame data to be mapped is obtained, and the color distribution features, texture roughness features and edge density features of the video frame data to be mapped are fused at the feature level to obtain the comprehensive feature vector of the video frame data to be mapped.

[0060] The comprehensive feature vector is matched with the standard feature vectors corresponding to each scene category in the preset scene category library to determine the matching priority between the comprehensive feature vector and the standard feature vector;

[0061] The scene category corresponding to the standard feature vector with the highest matching priority is selected as the scene category information of the video frame data to be mapped.

[0062] The process involves retrieving complete single-frame video data to be mapped, extracting pixel-by-pixel color values ​​from all pixels, and integrating the red, green, and blue channel values ​​of all pixels to form a raw dataset for color distribution features. Simultaneously, it involves statistically analyzing the grayscale differences in the neighborhood of each pixel within the image, integrating these results to form a raw dataset for texture roughness features. Furthermore, it involves performing a complete statistical analysis on boundary points where pixel grayscale changes abruptly, integrating the coordinate count and arrangement of all boundary points to form a raw dataset for edge density features. Finally, it involves dimensional concatenation and merging the raw datasets for color distribution, texture roughness, and edge density according to a fixed dimensional sorting rule: color distribution dimension first, texture roughness dimension in the middle, and edge density dimension last. After concatenation and merging, a comprehensive feature vector corresponding to the video frame data to be mapped is directly generated. This comprehensive feature vector fully carries all the image feature information contained in the three raw datasets.

[0063] The system reads all scene classification files stored in the local preset scene category library. Each scene classification file stores a standard feature vector corresponding to a scene category. The standard feature vectors are constructed according to the same dimensional sorting rules as the comprehensive feature vector. The system sequentially retrieves each standard feature vector in the scene category library and performs a dimensional comparison operation with the currently generated comprehensive feature vector. The comparison operation is performed one by one according to the vector dimension order. After each set of dimensional comparisons is completed, a set of overlapping dimensional content entries is recorded. The total number of overlapping dimensional content entries obtained after comparing all dimensions of the two sets of vectors is counted. The total number of overlapping dimensional content entries is used as the basis for determining the matching priority. Each standard feature vector in the scene category library will generate a set of exclusive matching priority values.

[0064] The matching priority values ​​corresponding to all standard feature vectors in the scene category library are retrieved. All matching priority values ​​are placed into a preset value comparison queue. The value comparison queue is automatically sorted in a fixed order from largest to smallest. The matching priority value corresponding to the first position of the sorted queue is extracted. The standard feature vector bound to the matching priority value is located. The scene category name recorded in the scene classification file to which the standard feature vector belongs is retrieved. The scene category name is directly set as the scene category information corresponding to the video frame data to be mapped. The scene category information is fully used to delineate the area of ​​interest in the subsequent video frame spatial distribution analysis stage.

[0065] The beneficial effects are that by simultaneously extracting three types of fixed image features from video frames and fusing them according to a unified dimensional rule to generate a comprehensive feature vector, coupled with a matching judgment method that compares and counts the number of overlapping dimensions one by one, and relying on a unified vector dimensional benchmark to complete the priority sorting and filtering of standard vectors for all scenes, it can stably output unique and definite scene category information. The unified data processing standard throughout the process can achieve complete process reproduction, avoiding the problem of missing image information caused by single feature extraction. At the same time, the judgment method of multi-dimensional overlapping items improves the matching degree between scene classification results and actual video image content, providing a precise basis for the delineation of the region range for subsequent analysis of salient target regions.

[0066] Q2. Based on the scene category information, perform spatial distribution analysis on the video frame data to be mapped to obtain the salient target region and the corresponding spatial distribution information of the video frame data to be mapped, and determine the edge contour features of the salient target region according to the spatial distribution information.

[0067] In this embodiment of the invention, the step of performing spatial distribution analysis on the video frame data to be mapped based on the scene category information to obtain the salient target region and corresponding spatial distribution information of the video frame data to be mapped includes:

[0068] The region of interest corresponding to the scene category information is delineated from the video frame data to be mapped, and the spatial resolution scale range of the video frame data to be mapped is determined based on the horizontal and vertical spans of the region of interest in the video frame data to be mapped.

[0069] Based on the spatial resolution scale range, spatial saliency response evaluation is performed on each pixel region of the video frame data to be mapped, and a saliency spatial distribution map of the video frame data to be mapped is constructed.

[0070] Connectivity regions are labeled in the saliency spatial distribution map to identify high-response connected regions in the saliency spatial distribution map as saliency target regions of the video frame data to be mapped;

[0071] Based on the position boundary of the salient target region in the salient spatial distribution map, the spatial position coordinates and spatial distribution range of the salient target region in the video frame data to be mapped are obtained as the spatial distribution information corresponding to the salient target region.

[0072] Determining the edge contour features of the salient target region based on the spatial distribution information includes:

[0073] Based on the spatial coordinates, locate the starting boundary point of the salient target region from the video frame data to be mapped;

[0074] Based on the spatial distribution range, the contour scanning span of the salient target region is determined, and according to the contour scanning span, starting from the starting boundary point, the boundary trajectory is traced point by point along the boundary direction of the salient target region to obtain the boundary trajectory point sequence of the salient target region.

[0075] The boundary trajectory point sequence is fitted with a contour line to obtain the edge contour features of the salient target region.

[0076] The scene category information of the determined video frame data to be mapped is retrieved, and matched with a pre-stored attention region coordinate template that corresponds one-to-one with that scene category. This coordinate template is derived from the statistical results of human visual attention hotspots under the corresponding scene category. It is formed by conducting visual gaze point collection experiments on a large number of video frames of the same scene, statistically analyzing the unified coordinate range of high-frequency gaze regions, and then solidifying it. Each scene category corresponds to an independent set of standard templates, which are pre-stored in the scene feature library. The origin of the coordinate template is aligned with the pixel coordinate origin of the video frame data to be mapped. The entire screen area covered by the coordinate template is delineated within the complete pixel frame of the video frame data to be mapped. This area is the attention region range corresponding to the scene category information. The attention region range is statistically analyzed along the horizontal pixel axis. The difference in the number of pixels between the leftmost and rightmost coordinates along the direction gives the horizontal span of the region of interest. The difference in the number of pixels between the topmost and bottommost coordinates along the vertical pixel axis gives the vertical span of the region of interest. One percent of the horizontal span is used as the minimum sampling unit width for spatial resolution, and one-tenth of the vertical span is used as the maximum sampling unit height for spatial resolution. This ratio is derived from the general scale division standard for spatial saliency analysis in the field of image processing. This ratio range can control the overall computational load while ensuring the accuracy of regional detail resolution. It is a conventional scale setting standard that has been verified in multiple scenarios. The minimum sampling unit width and the maximum sampling unit height together constitute the spatial resolution scale range of the video frame data to be mapped.

[0077] Based on the determined spatial resolution scale range, the complete pixel image of the video frame data to be mapped is divided into multiple pixel resolution units of uniform size. For each pixel resolution unit, the sum of the differences between the mean gray value of all pixels within the unit and the mean gray value of the eight neighboring resolution units is calculated. This sum of differences is used as the spatial saliency response value of the pixel resolution unit. After traversing all pixel resolution units, the spatial saliency response values ​​corresponding to all units are obtained. All spatial saliency response values ​​are filled into a blank coordinate map with the same size as the video frame data to be mapped according to the coordinate position of the corresponding pixel resolution unit, thus generating the saliency spatial distribution map of the video frame data to be mapped.

[0078] A preset high-response threshold is retrieved. This threshold is derived from the statistical mean of the spatial saliency response values ​​of video frames at the same resolution level. After calculating the full saliency response of a large number of video frame samples of the same specification, the arithmetic mean of all response values ​​is taken as a fixed threshold and stored in the judgment rule base in advance. The spatial saliency response values ​​of each coordinate point in the saliency spatial distribution map are compared coordinate by coordinate. Coordinate points with spatial saliency response values ​​higher than the high-response threshold are marked as high-response points. Any point is selected from all high-response points as the starting calibration point. The remaining high-response points directly adjacent to this point in the four directions of up, down, left, and right are sequentially retrieved. All high-response points that can be connected by adjacency are classified into the same connected region. After completing the connectivity classification of all high-response points, all high-response connected regions in the saliency spatial distribution map are obtained. All high-response connected regions are jointly determined as the saliency target region of the video frame data to be mapped.

[0079] Extract the leftmost, rightmost, topmost, and bottommost boundary coordinates of each high-response connected region in the saliency spatial distribution map. Convert the four boundary coordinates into corresponding coordinates in the pixel coordinate system of the video frame data to be mapped to obtain the spatial location coordinates of the saliency target region. Use the pixel difference between the leftmost and rightmost coordinates as the horizontal distribution value, and the pixel difference between the topmost and bottommost coordinates as the vertical distribution value. The horizontal and vertical distribution values ​​together constitute the spatial distribution range of the saliency target region. The spatial location coordinates and spatial distribution range are integrated to obtain the spatial distribution information corresponding to the saliency target region.

[0080] Retrieve the spatial coordinates contained in the spatial distribution information, select the leftmost and uppermost boundary coordinate point of the salient target area as the reference positioning point, and map the reference positioning point to the corresponding pixel position of the video frame data to be mapped. This pixel position is the starting boundary point of the salient target area.

[0081] The spatial distribution information includes the spatial range of the contour. The smaller value between the horizontal and vertical spread values ​​is set as the maximum moving distance for a single contour scan. This maximum moving distance is the contour scan span. Starting from the located starting boundary point, the system sequentially searches for adjacent pixel points in eight directions around the current point in a clockwise direction. The system selects adjacent points that belong to the boundary of the salient target area and have not been recorded as the next trajectory point. After recording each point, the system moves to that point to continue searching for the next point until it returns to the starting boundary point and stops tracking. All the boundary points recorded in sequence together constitute the boundary trajectory point sequence of the salient target area.

[0082] Read the coordinates of all trajectory points in the boundary trajectory point sequence in sequence. Take two adjacent trajectory points as a group and generate a smooth line with continuous transition between the two points in each group. All smooth lines are connected end to end to form a closed complete contour line. This closed complete contour line is the edge contour feature of the salient target area.

[0083] The beneficial effects include: matching the corresponding region of interest with scene category can narrow the processing range of spatial resolution; relying on the spatial resolution scale range set by a fixed ratio can ensure a balance between resolution accuracy and processing efficiency; the saliency response evaluation based on the gray-scale difference of neighboring areas can stably quantify the visual prominence of each pixel region; the four-neighbor connectivity calibration method can accurately divide the complete high-response region; the clockwise point-by-point tracking boundary extraction method can obtain a continuous and complete sequence of boundary points; the edge contour features generated by smooth fitting can accurately restore the shape boundary of salient targets; the entire process has a unified operation benchmark, clear judgment rules, and complete reproducibility, which can provide accurate contour basis for subsequent spatial registration and pixel mapping.

[0084] Q3. Based on the spatial coordinates of each pixel unit on the irregular substrate in the pixel arrangement structure data of the irregular LED display, construct the non-uniform grid topology structure of the irregular LED display.

[0085] In this embodiment of the invention, the step of constructing a non-uniform grid topology structure for the irregularly shaped LED display screen based on the spatial position coordinates of each pixel unit on the irregularly shaped substrate in the pixel arrangement structure data of the irregularly shaped LED display screen includes:

[0086] Obtain the pixel arrangement structure data of the irregular LED display screen, and extract the spatial position coordinates of each pixel unit on the irregular substrate from the pixel arrangement structure data;

[0087] Based on the spatial coordinates, each pixel unit is mapped to a corresponding coordinate point in a two-dimensional coordinate plane to generate the pixel coordinate matrix of the irregular LED display screen.

[0088] Based on the actual spatial distance between adjacent coordinate points in the pixel coordinate matrix, the neighborhood association relationship between each coordinate point in the pixel coordinate matrix is ​​determined.

[0089] Based on the neighborhood association relationship, connect the coordinate points in the pixel coordinate matrix that have the neighborhood association relationship to construct the non-uniform grid topology of the irregular LED display screen.

[0090] The system retrieves pre-archived pixel arrangement structure data of irregularly shaped LED displays. This data is generated by collecting the light-emitting center points of all pixel units on the surface of the irregularly shaped substrate point by point using an industrial-grade three-dimensional coordinate measuring instrument. Each set of data corresponds to a pixel unit and is accompanied by a unique identification number. The system reads the data entries of all pixel units line by line, extracts the spatial position coordinates of each pixel unit on the irregularly shaped substrate, and stores them in numerical order.

[0091] The mounting reference surface of the irregularly shaped LED display is selected as the projection reference surface of the two-dimensional coordinate plane. The spatial coordinates of each extracted pixel unit are projected into the two-dimensional coordinate plane by the vertical orthographic projection method. Each pixel unit corresponds to a unique coordinate point in the plane and retains the original numbering relationship. All coordinate points are arranged and combined according to the relative positions after spatial projection to generate the pixel coordinate matrix of the irregularly shaped LED display.

[0092] The preset neighborhood determination distance threshold is retrieved. This threshold is 1.2 times the standard physical distance between pixel units of this irregularly shaped LED display. For each coordinate point in the pixel coordinate matrix, the actual spatial distance between the point and all other coordinate points is calculated one by one. Two points whose actual spatial distance is less than or equal to the neighborhood determination distance threshold are determined to have a neighborhood relationship. After traversing all coordinate points, the neighborhood relationship between each coordinate point in the pixel coordinate matrix is ​​determined.

[0093] Based on the established neighborhood relationships, connecting line segments are drawn between each pair of coordinate points with neighborhood relationships in the pixel coordinate matrix. All coordinate points serve as grid nodes, and all connecting line segments serve as grid edges to form a complete mesh structure. This structure exhibits non-uniform characteristics due to the uneven density of pixel units on the irregular substrate, resulting in a non-uniform mesh topology structure for the completed irregular LED display.

[0094] The beneficial effects are that the pixel arrangement data based on industrial measurements ensures that the topology structure and the pixel position of the physical screen are completely matched; the vertical orthographic projection mapping method ensures that there is no positional distortion in the transformation from spatial coordinates to the two-dimensional plane; the rules for determining the neighborhood relationship based on the fixed ratio spacing threshold are unified, clear and reproducible; and the non-uniform grid topology structure constructed in the end can accurately restore the actual arrangement shape and adjacent connection relationship of the pixels of the irregular LED screen, providing an accurate physical topology benchmark for subsequent contour registration and pixel mapping.

[0095] Q4. Spatial registration is performed between the edge contour features and the non-uniform grid topology, and according to the spatial mapping relationship obtained after registration, each pixel in the video frame data to be mapped is mapped to the corresponding grid node in the non-uniform grid topology to generate the initial mapping pixel matrix of the video frame data to be mapped.

[0096] In this embodiment of the invention, the step of spatially registering the edge contour features with the non-uniform mesh topology, and mapping each pixel in the video frame data to be mapped to the corresponding mesh node in the non-uniform mesh topology according to the spatial mapping relationship obtained after registration, to generate an initial mapped pixel matrix of the video frame data to be mapped, includes:

[0097] Based on the direction change of the edge contour features, the edge contour features are divided into multiple contour segments;

[0098] Based on the local clustering density of each grid node in the non-uniform grid topology, the non-uniform grid topology is divided into multiple grid sub-regions;

[0099] The spatial overlap between the spatial distribution range of the contour segments and the spatial distribution range of the grid sub-regions is compared, and the correspondence between the contour segments and the grid sub-regions is established based on the comparison results.

[0100] According to the correspondence, each contour point on each contour segment is bound to each grid node in the corresponding grid sub-region in order of spatial position.

[0101] The pixel values ​​at each contour point in the video frame data to be mapped are assigned to the corresponding grid nodes to generate the initial mapped pixel matrix of the video frame data to be mapped.

[0102] Along the boundary trajectory of the edge contour feature, consecutive contour sampling points are selected sequentially at fixed sampling intervals. The tangent direction of the contour at each sampling point is determined one by one. The difference between the tangent direction of the current sampling point and the tangent direction of the previous sampling point is compared. When the difference between the two directions reaches the preset direction change judgment threshold, the current sampling point is marked as a contour segmentation point. The preset direction change judgment threshold is set to 15 degrees. This threshold is set according to the turning recognition of the edge of the conventional visual target. After traversing all sampling points of the complete edge contour feature, all marked contour segmentation points divide the edge contour feature into multiple continuous boundary segments. Each boundary segment is a contour segment.

[0103] Traverse each grid node in the non-uniform grid topology, delineate a circular statistical region with the coordinate position of a single grid node as the center and a preset statistical radius as the radius, count the total number of grid nodes contained within the circular statistical region, and use this number as the local cluster density of the corresponding grid node. The preset statistical radius is set to three times the standard center-to-center spacing of the pixel unit of the irregular LED display screen. Based on the numerical distribution of the local cluster density, perform region clustering on all grid nodes, and group grid nodes that are adjacent in position and whose local cluster density values ​​are in the same range into the same continuous region. All the divided continuous regions are multiple grid sub-regions.

[0104] For each contour segment, extract the maximum and minimum horizontal coordinates, maximum and minimum vertical coordinates of all contour points. The rectangular area enclosed by these four extreme values ​​represents the spatial distribution range of that contour segment. Similarly, extract the maximum and minimum horizontal coordinates, maximum and minimum vertical coordinates of all grid nodes in each grid sub-region. The rectangular area enclosed by these four extreme values ​​represents the spatial distribution range of that grid sub-region. Overlay the spatial distribution range of a single contour segment with the spatial distribution ranges of all grid sub-regions to determine the area of ​​the overlapping portion of the two rectangular regions. Select the grid sub-region with the largest overlapping area with the contour segment to establish a unique correspondence between the contour segment and the corresponding grid sub-region. After traversing all contour segments, the complete correspondence between contour segments and grid sub-regions is obtained.

[0105] For each pair of corresponding contour segments and grid sub-regions, sequentially number all contour points within the segment along the boundary trajectory of the contour segment. Sequentially number all grid nodes within the grid sub-region along the same arrangement direction as the contour segment. Bind each contour point corresponding to a number to a grid node corresponding to the same number in ascending order, ensuring that the spatial arrangement order of contour points and grid nodes is completely matched. After completing the binding operation of all pairs of corresponding relationships, the binding relationship between all contour points and grid nodes is obtained.

[0106] The pixel value of each contour point at the corresponding coordinate position in the video frame data to be mapped is read one by one. Based on the established binding relationship between the contour point and the grid node, the pixel value is directly assigned to the corresponding grid node. After all grid nodes have completed the pixel value assignment, all grid nodes with pixel values ​​are integrated and arranged according to the node arrangement relationship of the non-uniform grid topology to form a two-dimensional data set containing the pixel information of all grid nodes. This data set is the initial mapping pixel matrix of the video frame data to be mapped.

[0107] The beneficial effects are that by segmenting the contour based on the direction change threshold and dividing the grid sub-region based on the local aggregation density, it is possible to achieve precise matching between the edge of the video content and the grid of the irregular screen. Binding the contour points and grid nodes according to the spatial arrangement order can ensure the spatial position correspondence before and after mapping. The final generated initial mapping pixel matrix can accurately fit the non-uniform pixel arrangement characteristics of the irregular LED display, improving the spatial matching accuracy of pixel mapping and the visual restoration effect of the image edge.

[0108] Q5. Based on the actual spatial neighborhood relationship of each grid node in the initial mapped pixel matrix, perform pixel value correction on the initial mapped pixel matrix, and output the corrected mapped pixel matrix to the irregular LED display screen for display.

[0109] In this embodiment of the invention, the step of correcting the pixel values ​​of the initial mapped pixel matrix based on the actual spatial neighborhood relationships of each grid node in the initial mapped pixel matrix, and outputting the corrected mapped pixel matrix to the irregularly shaped LED display screen for display, includes:

[0110] Obtain the actual spatial neighborhood relationship of the grid nodes in the initial mapped pixel matrix on the irregular substrate;

[0111] Based on the actual spatial neighborhood relationship, the current pixel value of the grid node and the neighboring pixel values ​​of the adjacent grid nodes are extracted from the initial mapped pixel matrix;

[0112] Using the actual spatial distance between the grid node and the adjacent grid node as a constraint, the neighboring pixel values ​​are weighted and allocated to obtain the corrected pixel values ​​of the grid node.

[0113] The current pixel value of the grid node in the initial mapped pixel matrix is ​​replaced with the corrected pixel value. After traversing all grid nodes in the initial mapped pixel matrix, the corrected mapped pixel matrix of the initial mapped pixel matrix is ​​generated.

[0114] The corrected mapped pixel matrix is ​​subjected to display frame encapsulation processing to obtain the display data stream of the irregularly shaped LED display screen;

[0115] The display data stream is sent to the irregularly shaped LED display screen for display.

[0116] From the pre-stored non-uniform grid topology association data, each grid node in the initial mapped pixel matrix is ​​matched and retrieved one by one according to its unique identifier. The correspondence between all adjacent grid nodes of each grid node in the physical arrangement of the irregular substrate is pre-determined based on the actual physical spacing of the pixel units. The final retrieved correspondence between all adjacent nodes is the actual spatial neighborhood relationship of the grid nodes in the initial mapped pixel matrix.

[0117] Taking a single grid node to be processed as the processing object, first read the pixel value stored in the node itself from the initial mapped pixel matrix as the current pixel value. Then, based on the identifiers of all adjacent grid nodes corresponding to the node recorded in the actual spatial neighborhood relationship, read the pixel value stored in each adjacent grid node from the initial mapped pixel matrix in turn. The pixel values ​​of all adjacent nodes read are used as the neighborhood pixel values ​​of the grid node to be processed.

[0118] First, obtain the actual spatial distance value from the grid node to be processed to each adjacent grid node. Based on the display imaging rule that the closer the spatial distance, the higher the visual correlation of pixels, the neighboring pixel values ​​are weighted and allocated using an inverse distance weighting method. The formula for calculating the corrected pixel value is as follows:

[0119] ;

[0120] In the formula, For the corrected pixel value, The current pixel value, This represents the total number of adjacent grid nodes included in the actual spatial neighborhood relationship. The preset distance weight benchmark value, For the first grid node The neighboring pixel values ​​of each adjacent grid node. For the grid node to the first The actual spatial distance between adjacent grid nodes.

[0121] The corrected pixel value is the output of the pixel value correction step. It is calculated by fusing the pixel of the node to be processed with the pixels of its surrounding neighbors according to the distance weight. After the calculation is completed, it directly replaces the original value of the corresponding node in the initial mapped pixel matrix. Finally, the calculation results of all nodes together constitute the corrected mapped pixel matrix.

[0122] The current pixel value is taken from the original stored data of the grid node to be processed in the initial mapped pixel matrix. This data is directly generated by the binding mapping process between the video frame contour points and grid nodes, and is one of the basic data for correction calculation.

[0123] The total number of adjacent grid nodes is obtained by statistical analysis of the actual spatial neighborhood relationships. These neighborhood relationships are established during the non-uniform grid topology construction stage by determining the spatial spacing of pixel units. The statistics are obtained by traversing all neighborhood association entries of the node to be processed.

[0124] The distance weight benchmark value is a pre-set fixed reference value, which is based on the standard center-to-center spacing of the pixel units of the irregularly shaped LED display screen. This value is used to unify the benchmark scale for weight calculation, avoid large fluctuations in weight allocation due to differences in the number of neighboring nodes, and ensure the consistency and stability of the correction process.

[0125] The neighboring pixel values ​​are taken from the original stored data of each adjacent grid node in the initial mapped pixel matrix. When reading, the values ​​are extracted one by one from the corresponding positions in the matrix based on the node identifiers recorded in the actual spatial neighborhood relationship.

[0126] The actual spatial spacing is calculated and generated during the construction stage of the non-uniform grid topology. The calculation is based on the spatial position coordinates of the pixel unit on the irregular substrate. The physical straight-line distance is calculated by the difference between the horizontal and vertical coordinates of two points. The calculation results are stored together with the neighborhood association relationship and can be directly retrieved and used in the correction stage.

[0127] This formula corresponds to the weighted allocation process in the pixel value correction step. It uses the inverse ratio of spatial distance as the core rule for weight allocation, so that neighboring pixels with closer spacing occupy a higher proportion in the correction result. It adapts to the inconsistent pixel spacing caused by the non-uniform arrangement of irregular LED displays, eliminates the problem of excessive and abrupt color and brightness in local images after initial mapping, and directly uses the calculated correction result to update the node values ​​of the initial mapped pixel matrix, providing smoothed pixel data support for subsequent display frame encapsulation and screen output.

[0128] The calculated corrected pixel value is written to the storage location of the corresponding grid node in the initial mapped pixel matrix, overwriting the original current pixel value of the node. The calculation and replacement of the corrected pixel value are performed on all grid nodes in the initial mapped pixel matrix in a pre-set node traversal order. After all grid nodes have completed the pixel value replacement, the resulting complete pixel matrix is ​​the corrected mapped pixel matrix of the initial mapped pixel matrix.

[0129] According to the standard data transmission protocol of the irregular LED display driver chip, the pixel values ​​of all grid nodes in the corrected mapped pixel matrix are arranged according to the row scanning order corresponding to the display hardware. A frame start identifier, frame number, data length information and header verification information are added to the header of the arranged pixel data. A whole frame data check code is added to the tail of the data. The integrated single frame data is serialized according to a fixed transmission bit width, and finally a serial data sequence that conforms to the display hardware receiving specifications is generated. This serial data sequence is the display data stream of the irregular LED display.

[0130] Through the wired data transmission interface corresponding to the irregularly shaped LED display, the generated display data stream is sent frame by frame to the main control receiving unit of the display according to the fixed frame frequency preset by the display driver. After parsing each frame of data, the main control receiving unit distributes the pixel data to the driving circuit corresponding to each pixel unit, driving each pixel unit to light up according to the corresponding pixel value, and finally presents a complete video frame picture on the irregularly shaped LED display.

[0131] The beneficial effects are that pixel value correction is carried out based on the actual spatial neighborhood relationship and physical spacing of the irregular substrate, which can adapt to the pixel spacing difference caused by the non-uniform arrangement of irregular LED displays, eliminate the problem of excessive and abrupt local color and brightness in the image after initial mapping, improve the smoothness of the display and the overall visual consistency. The display data stream, which has been standardized and transmitted in a standardized frame, can be directly matched with the display hardware driving logic, ensuring the stability of the display output and the accuracy of the image reproduction.

[0132] like Figure 2 The diagram shown is a functional block diagram of an irregularly shaped LED pixel mapping system based on video content analysis provided in an embodiment of the present invention.

[0133] The irregular LED pixel mapping system based on video content analysis described in this invention can be installed in electronic devices. Depending on the functions implemented, the irregular LED pixel mapping system based on video content analysis may include a scene classification module, a spatial distribution analysis module, a non-uniform mesh construction module, a registration and mapping module, and a correction output module. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and is stored in the memory of the electronic device.

[0134] In this embodiment, the functions of each module / unit are as follows:

[0135] The scene classification module is used to classify the video frame data to be mapped into scenes, and obtain the scene category information of the video frame data to be mapped.

[0136] The spatial distribution analysis module is used to perform spatial distribution analysis on the video frame data to be mapped based on the scene category information, to obtain the salient target region of the video frame data to be mapped and the corresponding spatial distribution information, and to determine the edge contour features of the salient target region based on the spatial distribution information.

[0137] The non-uniform grid construction module is used to construct the non-uniform grid topology of the irregular LED display screen based on the spatial position coordinates of each pixel unit on the irregular substrate in the pixel arrangement structure data of the irregular LED display screen.

[0138] The registration and mapping module is used to spatially register the edge contour features with the non-uniform grid topology, and according to the spatial mapping relationship obtained after registration, map each pixel in the video frame data to be mapped to the corresponding grid node in the non-uniform grid topology to generate the initial mapping pixel matrix of the video frame data to be mapped.

[0139] The correction output module is used to correct the pixel values ​​of the initial mapped pixel matrix according to the actual spatial neighborhood relationship of each grid node in the initial mapped pixel matrix, and output the corrected mapped pixel matrix to the irregular LED display screen for display.

[0140] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0141] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0142] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0143] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0144] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for mapping irregularly shaped LED pixels based on video content analysis, characterized in that, The method includes: Q1. Perform scene classification on the video frame data to be mapped to obtain the scene category information of the video frame data to be mapped; Q2. Based on the scene category information, perform spatial distribution analysis on the video frame data to be mapped to obtain the salient target region and the corresponding spatial distribution information of the video frame data to be mapped, and determine the edge contour features of the salient target region according to the spatial distribution information. Q3. Based on the spatial coordinates of each pixel unit on the irregular substrate in the pixel arrangement structure data of the irregular LED display, construct the non-uniform grid topology structure of the irregular LED display. Q4. Spatial registration is performed between the edge contour features and the non-uniform grid topology, and according to the spatial mapping relationship obtained after registration, each pixel in the video frame data to be mapped is mapped to the corresponding grid node in the non-uniform grid topology to generate the initial mapping pixel matrix of the video frame data to be mapped. Q5. Based on the actual spatial neighborhood relationship of each grid node in the initial mapped pixel matrix, perform pixel value correction on the initial mapped pixel matrix, and output the corrected mapped pixel matrix to the irregular LED display screen for display.

2. The irregular LED pixel mapping method based on video content analysis as described in claim 1, characterized in that, The process of classifying the video frame data to be mapped into scenes to obtain scene category information for the video frame data to be mapped includes: The video frame data to be mapped is obtained, and the color distribution features, texture roughness features and edge density features of the video frame data to be mapped are fused at the feature level to obtain the comprehensive feature vector of the video frame data to be mapped. The comprehensive feature vector is matched with the standard feature vectors corresponding to each scene category in the preset scene category library to determine the matching priority between the comprehensive feature vector and the standard feature vector; The scene category corresponding to the standard feature vector with the highest matching priority is selected as the scene category information of the video frame data to be mapped.

3. The irregular LED pixel mapping method based on video content analysis as described in claim 1, characterized in that, The step of performing spatial distribution analysis on the video frame data to be mapped based on the scene category information to obtain the salient target region and corresponding spatial distribution information of the video frame data to be mapped includes: The region of interest corresponding to the scene category information is delineated from the video frame data to be mapped, and the spatial resolution scale range of the video frame data to be mapped is determined based on the horizontal and vertical spans of the region of interest in the video frame data to be mapped. Based on the spatial resolution scale range, spatial saliency response evaluation is performed on each pixel region of the video frame data to be mapped, and a saliency spatial distribution map of the video frame data to be mapped is constructed. Connectivity regions are labeled in the saliency spatial distribution map to identify high-response connected regions in the saliency spatial distribution map as saliency target regions of the video frame data to be mapped; Based on the position boundary of the salient target region in the salient spatial distribution map, the spatial position coordinates and spatial distribution range of the salient target region in the video frame data to be mapped are obtained as the spatial distribution information corresponding to the salient target region.

4. The irregular LED pixel mapping method based on video content analysis as described in claim 3, characterized in that, Determining the edge contour features of the salient target region based on the spatial distribution information includes: Based on the spatial coordinates, locate the starting boundary point of the salient target region from the video frame data to be mapped; Based on the spatial distribution range, the contour scanning span of the salient target region is determined, and according to the contour scanning span, starting from the starting boundary point, the boundary trajectory is traced point by point along the boundary direction of the salient target region to obtain the boundary trajectory point sequence of the salient target region. The boundary trajectory point sequence is fitted with a contour line to obtain the edge contour features of the salient target region.

5. The irregular LED pixel mapping method based on video content analysis as described in claim 1, characterized in that, The step of constructing a non-uniform grid topology structure for the irregularly shaped LED display screen based on the spatial coordinates of each pixel unit on the irregularly shaped substrate in the pixel arrangement structure data of the irregularly shaped LED display screen includes: Obtain the pixel arrangement structure data of the irregular LED display screen, and extract the spatial position coordinates of each pixel unit on the irregular substrate from the pixel arrangement structure data; Based on the spatial coordinates, each pixel unit is mapped to a corresponding coordinate point in a two-dimensional coordinate plane to generate the pixel coordinate matrix of the irregular LED display screen. Based on the actual spatial distance between adjacent coordinate points in the pixel coordinate matrix, the neighborhood association relationship between each coordinate point in the pixel coordinate matrix is ​​determined. Based on the neighborhood association relationship, connect the coordinate points in the pixel coordinate matrix that have the neighborhood association relationship to construct the non-uniform grid topology of the irregular LED display screen.

6. The irregular LED pixel mapping method based on video content analysis as described in claim 1, characterized in that, The step of spatially registering the edge contour features with the non-uniform mesh topology, and mapping each pixel in the video frame data to be mapped to the corresponding mesh node in the non-uniform mesh topology according to the spatial mapping relationship obtained after registration, to generate the initial mapped pixel matrix of the video frame data to be mapped, includes: Based on the direction change of the edge contour features, the edge contour features are divided into multiple contour segments; Based on the local clustering density of each grid node in the non-uniform grid topology, the non-uniform grid topology is divided into multiple grid sub-regions; The spatial overlap between the spatial distribution range of the contour segments and the spatial distribution range of the grid sub-regions is compared, and the correspondence between the contour segments and the grid sub-regions is established based on the comparison results. According to the correspondence, each contour point on each contour segment is bound to each grid node in the corresponding grid sub-region in the order of spatial position. The pixel values ​​at each contour point in the video frame data to be mapped are assigned to the corresponding grid nodes to generate the initial mapped pixel matrix of the video frame data to be mapped.

7. The irregular LED pixel mapping method based on video content analysis as described in claim 1, characterized in that, The step of correcting the pixel values ​​of the initial mapped pixel matrix based on the actual spatial neighborhood relationships of each grid node in the initial mapped pixel matrix, and outputting the corrected mapped pixel matrix to the irregularly shaped LED display screen for display, includes: Obtain the actual spatial neighborhood relationship of the grid nodes in the initial mapped pixel matrix on the irregular substrate; Based on the actual spatial neighborhood relationship, the current pixel value of the grid node and the neighboring pixel values ​​of the adjacent grid nodes are extracted from the initial mapped pixel matrix; Using the actual spatial distance between the grid node and the adjacent grid node as a constraint, the neighboring pixel values ​​are weighted and allocated to obtain the corrected pixel values ​​of the grid node. The current pixel value of the grid node in the initial mapped pixel matrix is ​​replaced with the corrected pixel value. After traversing all grid nodes in the initial mapped pixel matrix, the corrected mapped pixel matrix of the initial mapped pixel matrix is ​​generated. The corrected mapped pixel matrix is ​​subjected to display frame encapsulation processing to obtain the display data stream of the irregularly shaped LED display screen; The display data stream is sent to the irregularly shaped LED display screen for display.

8. A non-standard LED pixel mapping system based on video content analysis, characterized in that, The system for implementing the irregular LED pixel mapping method based on video content analysis as described in claim 1 includes: The scene classification module is used to classify the video frame data to be mapped into scenes, and obtain the scene category information of the video frame data to be mapped. The spatial distribution parsing module is used to perform spatial distribution parsing on the video frame data to be mapped based on the scene category information, to obtain the salient target region of the video frame data to be mapped and the corresponding spatial distribution information, and to determine the edge contour features of the salient target region based on the spatial distribution information. The non-uniform mesh construction module is used to construct the non-uniform mesh topology of the irregular LED display based on the spatial position coordinates of each pixel unit on the irregular substrate in the pixel arrangement structure data of the irregular LED display. The registration and mapping module is used to spatially register the edge contour features with the non-uniform grid topology, and according to the spatial mapping relationship obtained after registration, map each pixel in the video frame data to be mapped to the corresponding grid node in the non-uniform grid topology to generate the initial mapping pixel matrix of the video frame data to be mapped. The correction output module is used to correct the pixel values ​​of the initial mapped pixel matrix according to the actual spatial neighborhood relationship of each grid node in the initial mapped pixel matrix, and output the corrected mapped pixel matrix to the irregular LED display screen for display.

Images