Display optimization method and device for multi-screen spliced picture, equipment, medium and product

Abstract

The invention provides a display optimization method and device for a multi-screen spliced picture, equipment, a medium and a product, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining the dynamic feature data of a spliced screen according to the collection data of a sensor arranged in the spliced screen in advance; according to the image information of the spliced screen, identifying each sub-screen in the spliced screen, and obtaining static feature data corresponding to each sub-screen; performing prediction according to the dynamic feature data, the static feature data and the image information of the spliced screen to obtain a display category of the spliced screen; and adjusting the picture display of the spliced screen according to the display category. According to the embodiment of the invention, the method and the device have the advantages of higher recognition precision of spliced picture display and higher recognition efficiency of spliced picture display, effectively optimize the screen display effect and improve the experience feeling of audiences.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, medium, and product for optimizing the display of multi-screen spliced ​​images. Background Technology

[0002] In existing screen display technologies, the main challenge in optimizing and recognizing the display status of multi-screen splicing systems lies in identifying the display status of the spliced ​​areas. This includes issues such as uneven brightness, color deviation, and differences in splicing seams. These problems often lead to a poor viewing experience, especially in large-screen monitoring and display scenarios, potentially resulting in image distortion and information loss. Existing technologies mainly employ the following methods to identify and handle spliced ​​display conditions: methods based on static parameter analysis, methods based on edge detection and image processing, methods based on manual observation and experience-based judgment, and dynamic recognition methods based on sensor monitoring.

[0003] However, existing technologies have significant shortcomings in display optimization and recognition for multi-screen splicing systems. First, methods based on static parameter analysis rely too heavily on preset thresholds and rules, making it difficult to handle complex ambient lighting conditions and dynamic changes in displayed content in real-world applications, thus limiting recognition accuracy. Furthermore, while edge detection and image processing methods offer improved accuracy, their complex algorithms and high computational requirements make them inefficient for large-scale splicing systems. Methods based on manual observation and experience rely excessively on the expertise of technicians, which is not only inefficient in large splicing systems but also prone to inconsistencies in display areas. While sensor-based dynamic recognition methods can provide real-time adjustments, their high implementation costs and complex system architecture limit their widespread application. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, device, medium, and product for optimizing the display of multi-screen splicing images, in order to solve the problems of low recognition accuracy and efficiency of the display status of the splicing area in a multi-screen splicing system and poor display optimization effect.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a method for optimizing the display of multi-screen spliced ​​images, comprising:

[0006] Based on the data collected by sensors pre-placed within the splicing screen, dynamic characteristic data of the splicing screen is obtained;

[0007] Based on the image information of the splicing screen, each sub-screen within the splicing screen is identified to obtain static feature data corresponding to each sub-screen.

[0008] Based on the dynamic feature data, the static feature data, and the image information of the spliced ​​screen, a prediction is made to obtain the display category of the spliced ​​screen;

[0009] Adjust the display of the splicing screen according to the display category.

[0010] Optionally, the method further includes:

[0011] Obtain a first parameter for at least one target point on the splicing screen, wherein the first parameter includes the adjacent side length of the splicing seam of the sub-screen to which the target point belongs and the number of corners of the splicing seam within the target range corresponding to the target point;

[0012] The sensor arrangement density corresponding to the target point is calculated based on the adjacent side length and the number of corners; wherein, within the target range corresponding to the target point, multiple sensors are arranged according to the sensor arrangement density.

[0013] Optionally, the dynamic feature data includes brightness value, contrast value, edge halo difference value, splicing color difference value, and color temperature gradient value;

[0014] The step of obtaining dynamic feature data of the splicing screen based on data collected by sensors pre-positioned within the splicing screen includes:

[0015] Based on the data collected by sensors pre-placed within the splicing screen, the brightness, contrast, color temperature, and average color of the splicing screen are obtained in real time.

[0016] The edge halo difference value of the splicing screen is calculated based on the brightness values ​​of two adjacent sub-screens in the splicing screen.

[0017] The color difference value of the splicing screen is calculated based on the average color value of two adjacent sub-screens in the splicing screen.

[0018] The color temperature gradient value of the splicing screen is calculated based on the color temperature value and the color temperature gradient operator corresponding to the splicing screen, wherein the color temperature gradient operator is different for different splicing areas in the splicing screen.

[0019] Optionally, calculating the edge halo difference value of the spliced ​​screen based on the brightness values ​​of two adjacent sub-screens in the spliced ​​screen includes:

[0020] The edge halo difference value of the spliced ​​screen is calculated based on the brightness values ​​of the edges of two adjacent sub-screens and the brightness values ​​of the corners of the two sub-screens.

[0021] Optionally, calculating the splicing color difference value of the splicing screen based on the average color value of two adjacent sub-screens in the splicing screen includes:

[0022] The splicing color difference value of the splicing screen is calculated based on the average color value of the edges of two adjacent sub-screens and the average color value of the corners of the two sub-screens.

[0023] Optionally, calculating the color temperature gradient value of the splicing screen based on the color temperature value and the color temperature gradient operator corresponding to the splicing screen includes:

[0024] Based on the color temperature value, calculate the unidirectional horizontal color temperature gradient operator and the unidirectional vertical color temperature gradient operator corresponding to the splicing area of ​​the adjacent edge in the splicing screen;

[0025] Based on the color temperature value, calculate the unidirectional horizontal color temperature gradient operator, the unidirectional vertical color temperature gradient operator, the positive diagonal color temperature gradient operator, and the negative diagonal color temperature gradient operator corresponding to the splicing area of ​​the adjacent corner in the splicing screen.

[0026] The color temperature gradient value of the splicing screen is calculated based on the unidirectional horizontal and unidirectional vertical color temperature gradient operators corresponding to the splicing areas of adjacent edges in the splicing screen, as well as the unidirectional horizontal, unidirectional vertical, positive diagonal, and negative diagonal color temperature gradient operators corresponding to the splicing areas of adjacent corners in the splicing screen.

[0027] Optionally, the step of identifying each sub-screen within the spliced ​​screen based on the image information of the spliced ​​screen to obtain static feature data corresponding to each sub-screen includes:

[0028] For the target sub-screen currently being processed, extract the image information of the first region corresponding to the target sub-screen based on the image information of the spliced ​​screen;

[0029] Based on the image information of the first region corresponding to the target sub-screen, a convolutional neural network is used to predict and obtain the spatial type corresponding to the target sub-screen. The number of sub-screens spliced ​​together corresponds to different spatial types.

[0030] Based on the image information of the first region corresponding to the target sub-screen and the spatial type corresponding to the target sub-screen, prediction is performed to obtain the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively. The first sub-screen is a sub-screen that is spliced ​​together with the target sub-screen. The static feature data includes edge gap width, tilt angle and flatness.

[0031] Based on the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively, the parameters are adjusted using a pre-trained parameter correction model to obtain the final static feature data corresponding to the target sub-screen.

[0032] Optionally, the step of adjusting the initial static feature data corresponding to the target sub-screen and the first sub-screen using a pre-trained parameter correction model to obtain the final static feature data corresponding to the target sub-screen includes:

[0033] Based on the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively, the parameters are adjusted using a pre-trained multi-screen parameter correction model to obtain the adjusted static feature data corresponding to the target sub-screen.

[0034] Based on the adjusted static feature data corresponding to the target sub-screen, the model is adjusted using a pre-trained parameter correction model within a single screen to obtain the final static feature data corresponding to the target sub-screen.

[0035] Optionally, the step of predicting the display category of the spliced ​​screen based on the dynamic feature data, the static feature data, and the image information of the spliced ​​screen includes:

[0036] The static feature data is extracted using a static multilayer perceptron to obtain a static feature vector;

[0037] The dynamic feature data is extracted using a dynamic multilayer perceptron to obtain a dynamic feature vector;

[0038] Image feature vectors are obtained by extracting image information from the spliced ​​screen using a convolutional neural network.

[0039] Based on the static feature vector, the dynamic feature vector, and the image feature vector, a perceptual hierarchical network is used to predict and obtain the display category of the spliced ​​screen, wherein the display category includes a comfort level and a content display completeness level.

[0040] This invention also provides a display optimization device for multi-screen splicing, comprising:

[0041] The first acquisition module is used to obtain dynamic feature data of the splicing screen based on the data collected by sensors pre-arranged in the splicing screen.

[0042] The first calculation module is used to identify each sub-screen in the splicing screen according to the image information of the splicing screen, and obtain static feature data corresponding to each sub-screen.

[0043] The first prediction module is used to predict the display category of the splicing screen based on the dynamic feature data, the static feature data and the image information of the splicing screen.

[0044] The first adjustment module is used to adjust the display of the splicing screen according to the display category.

[0045] This invention also provides a network device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the multi-screen splicing display optimization method as described in any of the preceding embodiments.

[0046] This invention also provides a readable storage medium, comprising: a program stored on the readable storage medium, wherein when the program is executed by a processor, it implements the steps of display optimization of multi-screen splicing images as described in any of the preceding claims.

[0047] This invention also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps for optimizing the display of multi-screen splicing images as described in any of the preceding embodiments.

[0048] At least one of the above technical solutions of the present invention has the following beneficial effects:

[0049] In the above scheme, firstly, dynamic feature data of the splicing screen is obtained based on data collected by sensors pre-placed within the splicing screen. Then, each sub-screen within the splicing screen is identified based on the image information of the splicing screen, obtaining static feature data corresponding to each sub-screen. Finally, the display category of the splicing screen is predicted based on the dynamic feature data, static feature data, and image information of the splicing screen. The display of the splicing screen is then adjusted according to the display category. This embodiment of the invention predicts the comfort level and content display completeness level of the splicing screen based on the screen's dynamic feature data, static feature data, and image information, effectively improving the recognition accuracy of the splicing screen display and exhibiting high splicing screen display recognition efficiency. Furthermore, adjusting the display of the splicing screen according to its display category effectively optimizes the screen display effect and enhances the viewer experience. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating the method for optimizing the display of multi-screen spliced ​​images according to an embodiment of the present invention;

[0051] Figure 2 This is a schematic diagram of the structure of the splicing screen according to an embodiment of the present invention;

[0052] Figure 3A schematic diagram illustrating the process of optimizing and adjusting the display of spliced ​​screens using the multi-screen splicing display optimization provided by this invention;

[0053] Figure 4 This is a schematic diagram of the structure of the multi-screen splicing display optimization device according to an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0056] like Figure 1 As shown, this embodiment of the invention provides a method for optimizing the display of multi-screen spliced ​​images, including:

[0057] Step S101: Obtain dynamic feature data of the splicing screen based on the data collected by sensors pre-arranged within the splicing screen.

[0058] In step S101, the sensors include, but are not limited to, a brightness sensor and a color sensor; the dynamic feature data include, but are not limited to, brightness values, contrast values, edge halo difference values, splicing color difference values, and color temperature gradient values.

[0059] Step S102: Based on the image information of the splicing screen, identify each sub-screen within the splicing screen to obtain static feature data corresponding to each sub-screen.

[0060] In step S102, the splicing screen is composed of multiple sub-screens spliced ​​together. The static feature data includes, but is not limited to, the width of the splicing edge gap, the tilt angle and the flatness. In the multi-screen splicing system, the gap width, splicing angle and flatness between each sub-screen are interrelated feature attributes.

[0061] Step S103: Based on the dynamic feature data, the static feature data, and the image information of the splicing screen, a prediction is made to obtain the display category of the splicing screen;

[0062] In step S103, dynamic feature data, static feature data, and image information of the splicing screen are comprehensively considered to predict the display category of the entire splicing screen, which effectively improves the recognition accuracy of the splicing screen display and has high splicing screen display recognition efficiency. The display category is used to indicate the current display status of the splicing screen, such as the viewing comfort and display completeness of the splicing screen.

[0063] Step S104: Adjust the display of the splicing screen according to the display category.

[0064] In step S104, the display of the splicing screen is dynamically adjusted according to the display category corresponding to the splicing screen, and the display of the screen is continuously monitored and optimized.

[0065] In this embodiment of the invention, firstly, dynamic feature data of the splicing screen is obtained based on data collected by sensors pre-placed within the splicing screen. Then, based on the image information of the splicing screen, each sub-screen within the splicing screen is identified, and static feature data corresponding to each sub-screen is obtained. Finally, based on the dynamic feature data, static feature data, and image information of the splicing screen, a display category of the splicing screen is predicted. Based on the display category, the display of the splicing screen is adjusted. This embodiment of the invention predicts the comfort level and content display completeness level of the splicing screen based on the screen's dynamic feature data, static feature data, and image information, effectively improving the recognition accuracy of the splicing screen display and exhibiting high splicing screen display recognition efficiency. Furthermore, adjusting the display of the splicing screen based on its display category effectively optimizes the screen display effect and enhances the viewer experience.

[0066] Optionally, the method further includes:

[0067] Obtain a first parameter for at least one target point on the splicing screen, wherein the first parameter includes the adjacent side length of the splicing seam of the sub-screen to which the target point belongs and the number of corners of the splicing seam within the target range corresponding to the target point;

[0068] The sensor arrangement density corresponding to the target point is calculated based on the adjacent side length and the number of corners; wherein, within the target range corresponding to the target point, multiple sensors are arranged according to the sensor arrangement density.

[0069] In this embodiment of the invention, in the splicing gap area of ​​the splicing screen, slight differences in the display characteristics of the splicing screen or differences in time synchronization can easily lead to problems such as inconsistent screen color and brightness. Figure 2 As shown, this inconsistency may not be noticeable on a single screen, but it becomes very apparent in splicing areas, especially in corners and gaps. Furthermore, in corner areas, the edges of multiple screens intersect, meaning that even a slight color deviation on any one screen will be amplified in this area. Therefore, this invention proposes an adaptive method for sensor placement density. By adjusting the sensor placement density near the edges and corners of the splicing seams, it can more accurately capture and calibrate information output near the adjacent edges and corners of the splicing seams, reducing deviations. The sensor placement density formula is as follows:

[0070]

[0071] Where D(x,y) is the sensor density at any coordinate point (x,u) on the screen, D o Wc edge Wc is the weighting coefficient for the adjacent area of ​​the splicing seam in a video wall. corner This refers to the weighting coefficient for the corner areas of the splicing seams in a video wall. The edge influence function represents the effect of the relative length of the adjacent edges in the splicing seam region on the sensor density, Lc edge (x,y) is the length of the adjacent side of the splicing seam at point (x,y), L total It is the total length of the adjacent sides. Let N be the corner influence function, representing the effect of the relative number of corners in the splicing seam corner region on the sensor density. corners (x,y) represents the number of corners corresponding to the point (x,y), N total This represents the total number of corners, and h(x,y) is the position correction function, indicating the adjustment of sensor density based on the specific requirements of the location of point (x,y). For example... Figure 2 As shown, L is the length of the adjacent side of the splicing seam on the right side of sub-screen 1. Regarding the number of corners, there are 2 corners in area A and 4 corners in area B.

[0072] Specifically, Lc edge (x,y) refers to the actual length of the adjacent edge of the seam corresponding to the shortest vertical distance among the multiple seams (top / bottom / left / right) of the point (x,y) from its sub-screen; L total N refers to the sum of the lengths of the adjacent edges of the seams of the subscreen to which the point (x,y) belongs; corners(x,y) refers to the number of corners within the target area corresponding to point (x,y), for example: the number of corners with a distance ≤ R from point (x,y); N total It refers to the maximum number of corners corresponding to the seams in a spliced ​​screen. Figure 2 In, N total =4.

[0073] Therefore, it can be concluded that the closer (x,y) is to the seam, the better. The larger the value, the closer (x,y) is to the corner. The larger the value, the more likely it is that if the coordinate point (x, y) is located in the center of the screen, the following usually applies: At this point, the density of the coordinate point (x,y) is the base value D0h(x,y).

[0074] Optionally, the dynamic feature data includes brightness value, contrast value, edge halo difference value, splicing color difference value, and color temperature gradient value;

[0075] The step of obtaining dynamic feature data of the splicing screen based on data collected by sensors pre-positioned within the splicing screen includes:

[0076] Based on the data collected by sensors pre-placed within the splicing screen, the brightness, contrast, color temperature, and average color of the splicing screen are obtained in real time.

[0077] The edge halo difference value of the splicing screen is calculated based on the brightness values ​​of two adjacent sub-screens in the splicing screen.

[0078] The color difference value of the splicing screen is calculated based on the average color value of two adjacent sub-screens in the splicing screen.

[0079] The color temperature gradient value of the splicing screen is calculated based on the color temperature value and the color temperature gradient operator corresponding to the splicing screen, wherein the color temperature gradient operator is different for different splicing areas in the splicing screen.

[0080] In this embodiment of the invention, firstly, real-time luminance value Li and real-time contrast value Co are obtained through a pre-arranged luminance sensor, and the average color value (Re, Gr, Bl) detected by the sensor is obtained through a pre-arranged color sensor; then, based on the above average color value (Re, Gr, Bl), it is first converted to the CIE 1931 XYZ color space to calculate the chromaticity coordinates x, y, and then the color temperature value T is calculated based on the color temperature formula (McCamy). i ,i=1,…,N,N is an integer greater than 1,T i Represents the i-th measurement point (x) i ,y i The real-time color temperature value.

[0081] Finally, to prevent inconsistent edge halo effects between adjacent sub-screens, which could lead to unnatural brightness transitions at the splicing points, the halo difference value at the splicing edges of the splicing screens is extracted based on the aforementioned brightness value Li. To address the issue of color discontinuity in the splicing screens, the splicing color difference value is extracted based on the aforementioned color mean. To comprehensively measure color temperature differences at the splicing points, ensure color temperature consistency at the splicing seams, and avoid color mismatch issues at corners, the color temperature gradient value of the splicing screens is calculated based on the color temperature value and the corresponding color temperature gradient operator.

[0082] Optionally, calculating the edge halo difference value of the spliced ​​screen based on the brightness values ​​of two adjacent sub-screens in the spliced ​​screen includes:

[0083] The edge halo difference value of the spliced ​​screen is calculated based on the brightness values ​​of the edges of two adjacent sub-screens and the brightness values ​​of the corners of the two sub-screens.

[0084] In this embodiment of the invention, the formula for calculating the edge halo difference value of the spliced ​​screen is as follows: (The formula is based on the brightness values ​​at the edges of two adjacent sub-screens and the brightness values ​​at the corners of the two sub-screens.)

[0085]

[0086] Among them, L cs It is the edge halo difference value, |L screenel -L screener | Represents the average brightness difference at the edges of two adjacent sub-screens, used to measure the inconsistency of the halo at the boundary of the splicing seam between two adjacent sub-screens. Specifically: L screenel It is the brightness value of one of the adjacent sub-screens near the edge of the splicing seam. The splicing seam refers to the gap between two adjacent sub-screens. L screener It is the brightness value of another sub-screen adjacent to the splicing seam. The average of the pairwise brightness differences between the four corner measurement points is used to measure the dispersion of brightness in the corner areas. Specifically, L screenc,k and L screenc,j Each represents the corner brightness value of the adjacent seam between two sub-screens, where k, j = 1, 2, 3, 4, and are the corner numbers. In the specific calculation, the four corner measurement points L... corner,1～4 The calculation is performed according to a fixed sequence of numbers (e.g., top left, top right, bottom right, bottom left).

[0087] It should be noted that if two adjacent left and right sub-screens are spliced ​​together, and there is a splicing gap between the two adjacent spliced ​​sub-screens, then L screenel It is the brightness value of the edge of the left sub-screen near the splicing seam, L screener L is the brightness value of the edge of the adjacent splicing seam of the right sub-screen; if two adjacent upper and lower sub-screens are spliced ​​together, and there is a splicing seam between the two adjacent spliced ​​sub-screens, then L screenel It is the brightness value of the edge of the upper sub-screen near the splicing seam, L screener It is the brightness value of the edge of the lower sub-screen near the splicing seam.

[0088] Optionally, calculating the splicing color difference value of the splicing screen based on the average color value of two adjacent sub-screens in the splicing screen includes:

[0089] The splicing color difference value of the splicing screen is calculated based on the average color value of the edges of two adjacent sub-screens and the average color value of the corners of the two sub-screens.

[0090] In this embodiment of the invention, the formula for calculating the color difference value of the splicing screen based on the average color of the edges of two adjacent sub-screens and the average color of the corners of the two sub-screens is as follows:

[0091]

[0092] Among them, C a Re represents the color difference value for multi-screen splicing. screencel Gr screencel ,Bl screencel Re represents the average color value of the edge of the seam between two adjacent sub-screens in a splicing arrangement. screencer Gr screencer ,Bl screencer Re is the average color value of the edge of the seam of another adjacent sub-screen. screenc,i Re screenc,j Gr screenc,i Gr screenc,j ,Bl screenc,i ,Bl screenc,j The values ​​represent the average color of two adjacent sub-screens, i, j = 1, 2, 3, 4, which are the numbers of the four corners of the adjacent seam between the two sub-screens. In the specific calculation, the measurement points Re at the four corners are used... screenc1～4 Gr screenc,1～4 ,Bl screenc,1～4 The calculation is performed according to a fixed sequence of numbers (e.g., top left, top right, bottom right, bottom left).

[0093] Optionally, calculating the color temperature gradient value of the splicing screen based on the color temperature value and the color temperature gradient operator corresponding to the splicing screen includes:

[0094] Based on the color temperature value, calculate the unidirectional horizontal color temperature gradient operator, the unidirectional horizontal color temperature gradient operator, and the unidirectional vertical color temperature gradient operator corresponding to the splicing area of ​​the adjacent edge in the splicing screen.

[0095] Based on the color temperature value, calculate the unidirectional vertical color temperature gradient operator, the positive oblique color temperature gradient operator, and the negative oblique color temperature gradient operator corresponding to the splicing area of ​​the adjacent corner in the splicing screen.

[0096] The color temperature gradient value of the splicing screen is calculated based on the unidirectional horizontal and unidirectional vertical color temperature gradient operators corresponding to the splicing areas of adjacent edges in the splicing screen, as well as the unidirectional horizontal, unidirectional vertical, positive diagonal, and negative diagonal color temperature gradient operators corresponding to the splicing areas of adjacent corners in the splicing screen.

[0097] In this embodiment of the invention, firstly, the color temperature value T is... i This is transformed into a color temperature field (continuous or rasterized) C(x,y) on the coordinate plane, as shown in the formula:

[0098]

[0099] in This indicates the interpolation / resampling operator (commonly bilinear or Gaussian weighted interpolation).

[0100] Then, based on a single splicing gap, this embodiment of the invention proposes a horizontal gradient operator, which represents a unidirectional horizontal gradient in the splicing region of adjacent edges, and represents a unidirectional vertical gradient operator, a positive oblique gradient operator, and a negative oblique gradient operator in the splicing region of adjacent corners. The specific formula is as follows:

[0101] One-way horizontal gradient operator:

[0102] One-way vertical gradient operator:

[0103] Positive rhombic gradient operator:

[0104] Inverse oblique gradient operator:

[0105] Finally, in this embodiment of the invention, the global color temperature gradient operator of the spliced ​​screen is calculated based on the gradient operators corresponding to the adjacent edge splicing area and the adjacent corner splicing area, respectively. The specific formula is as follows:

[0106]

[0107] in, This represents the total color temperature gradient value of the spliced ​​screen. This represents the color temperature gradient along the horizontal gradient. This represents the color temperature gradient along the vertical gradient. Positive orthogonal color temperature gradient, For the reverse oblique color temperature gradient, δ(x,y) is an identifier used to distinguish the type of splicing region. When δ(x,y) is 1, it represents a splicing region at adjacent edges; when δ(x,y) is 2, it represents a splicing region at adjacent corners. α h α v β represents the weights of the horizontal and vertical gradients in the region where adjacent edges are joined. h β v β d1 β d2 The weights of gradients in each direction in the adjacent corner splicing region.

[0108] Optionally, the step of identifying each sub-screen within the spliced ​​screen based on the image information of the spliced ​​screen to obtain static feature data corresponding to each sub-screen includes:

[0109] For the target sub-screen currently being processed, extract the image information of the first region corresponding to the target sub-screen based on the image information of the spliced ​​screen;

[0110] Based on the image information of the first region corresponding to the target sub-screen, a convolutional neural network is used to predict and obtain the spatial type corresponding to the target sub-screen. The number of sub-screens spliced ​​together corresponds to different spatial types.

[0111] Based on the image information of the first region corresponding to the target sub-screen and the spatial type corresponding to the target sub-screen, prediction is performed to obtain the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively. The first sub-screen is a sub-screen that is spliced ​​together with the target sub-screen. The static feature data includes edge gap width, tilt angle and flatness.

[0112] Based on the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively, the parameters are adjusted using a pre-trained parameter correction model to obtain the final static feature data corresponding to the target sub-screen.

[0113] In this embodiment of the invention, static feature data includes, but is not limited to, edge gap width, tilt angle, and flatness. In a multi-screen splicing system, the gap width, splicing angle, and flatness between each sub-screen are interconnected feature attributes. This is especially true when a sub-screen is surrounded by multiple sub-screens, where the mutual influence is more significant. For example: Figure 2 As shown, a sub-screen with a space type of center position is surrounded by 8 sub-screens, a sub-screen with a space type of edge position is surrounded by 5 sub-screens, and a sub-screen with a space type of corner position is surrounded by 3 sub-screens.

[0114] Taking the gap width as an example, when the gap on one side increases, the gap width on the adjacent side is often affected. Furthermore, if the splicing angle of a sub-screen has an error (such as tilt), this error may be transmitted to adjacent sub-screens. It should be noted that the gap width in this embodiment refers to the maximum gap width. Similarly, when a sub-screen is uneven (the screen surface height is slightly off), the flatness of adjacent sub-screens will also affect each other.

[0115] Therefore, this embodiment of the invention first identifies the spatial type corresponding to each sub-screen based on a convolutional neural network to determine the position of each sub-screen in the spliced ​​screen, as shown in the following formula:

[0116] F1 = CNN(I) crop );

[0117]

[0118] in, For the prediction type, softmax(F1) is the prediction type for I. crop After convolution extraction, the result is converted into a probability distribution to determine the spatial type of the sub-screen (e.g., splicing 8 screens, splicing 5 screens, splicing 3 screens). The F1 score is I. crop Higher-order feature information, I crop This refers to the image information of the first area corresponding to the sub-screen. For screen classification, the cross-entropy classification loss function is y. j It is the actual type.

[0119] For I crop This invention also provides an acquisition method that involves cropping the captured image of the entire spliced ​​screen I to extract a first region I that is slightly larger than the target sub-screen being processed. crop This ensures that transition information at the edges / corners of the screen is captured, while also preventing too many irrelevant areas from entering subsequent networks, thus improving prediction accuracy and efficiency. The formula is as follows:

[0120] I crop =Crop(I,x start ,ystart ,hi,we).

[0121] Among them, I crop It refers to the cropped image, i.e., image information, Crop(I,x) start ,y start (,hi,we) represents the coordinates x start ,y start Start cropping to create an image with height hi and width we.

[0122] Furthermore, after obtaining the spatial type of the sub-screen, for the target sub-screen being processed, prediction parameters are established based on its spatial type to predict the static feature data of the target sub-screen and multiple first sub-screens that are concatenated with the target sub-screen. The formula is as follows:

[0123]

[0124] Among them, w center-init θ is the initial maximum gap width prediction value of the target sub-screen unit currently being processed. center-init It is the initial tilt angle prediction value of the target sub-screen unit currently being processed, Δh. center-init It is the initial flatness prediction value of the target sub-screen unit currently being processed, w j-init θ is the initial maximum gap width prediction value of the first sub-screen that is stitched together with the target sub-screen. j-init It is the initial tilt angle prediction value of the first sub-screen, Δh j-init is the initial flatness prediction value of the first sub-screen, and M is the number of the first sub-screens.

[0125] It should be noted that the predicted value of the maximum gap width w is the predicted value of the four sides of the screen, so it is a 1x4 vector. The tilt angle θ is the slope of the top of the sub-screen unit, and the flatness Δh is the highest distance of the screen from the panel.

[0126] Finally, in this embodiment of the invention, a parameter correction model is pre-trained to address the mutual influence between the target sub-screen and the first sub-screen, as well as the mutual influence of parameters within the target sub-screen. The initial static feature data is adjusted based on the parameter correction model corresponding to the spatial features of the target sub-screen to obtain the final static feature data of the target sub-screen. This effectively reduces the error caused by a single feature, thereby improving the accuracy of prediction for each sub-screen. The parameter correction model includes a first parameter correction model between multiple sub-screens and a second parameter correction model within the target sub-screen.

[0127] Optionally, the step of adjusting the initial static feature data corresponding to the target sub-screen and the first sub-screen using a pre-trained parameter correction model to obtain the final static feature data corresponding to the target sub-screen includes:

[0128] Based on the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively, the parameters are adjusted using a pre-trained multi-screen parameter correction model to obtain the adjusted static feature data corresponding to the target sub-screen.

[0129] Based on the adjusted static feature data corresponding to the target sub-screen, the model is adjusted using a pre-trained parameter correction model within a single screen to obtain the final static feature data corresponding to the target sub-screen.

[0130] In this embodiment of the invention, firstly, the initial static feature data corresponding to the target sub-screen and the first sub-screen are adjusted using a pre-trained first parameter correction model among multiple sub-screens, as shown in the following formula:

[0131]

[0132] Among them, f w (·) is a function used to correct the gap width, w center-middle This is the adjusted maximum gap width prediction value for the currently processed target sub-screen, w. center-init , is the initial maximum gap width prediction value of the target sub-screen unit, parameter w j-init It is the initial maximum gap width prediction value for the first sub-screen.

[0133]

[0134] Among them, f θ (·) is a function used to correct the tilt angle, θ center-middle It is the adjusted tilt angle prediction value of the target sub-screen being processed, θ. center-init It is the initial tilt angle prediction value of the target sub-screen unit, θ. j-init It is the initial tilt angle prediction value for the first sub-screen.

[0135]

[0136] Among them, f h (·) is a function used to correct for flatness, Δh center-middle It is the adjusted flatness prediction value of the target sub-screen being processed, Δh center-init It is the initial flatness prediction value of the target sub-screen unit, Δh j-init It is the initial flatness prediction value for the first sub-screen.

[0137]

[0138] in, The loss function of the model is adjusted for the first parameter across multiple sub-screens, w true θ true Δh true These represent the actual values ​​of the maximum gap width, tilt angle, and flatness of the target sub-screen.

[0139] Then, the adjusted static feature data corresponding to the target sub-screen is adjusted using the pre-trained second parameter correction model inside the target sub-screen. The specific formula is as follows:

[0140] w center-final =g w (w center-middle ,θ center-middle ,Δh center-middle ).

[0141] θ center-final =g θ (w center-middle ,θ center-middle ,Δh center-middle ).

[0142] Δh center-final =g Δh (w center-middle ,θ center-middle ,Δh center-middle ).

[0143] Among them, w center-final θ is the predicted maximum gap width of the target sub-screen. center-final Δh is the predicted tilt angle of the target subscreen. center-final For the final predicted flatness value of the target subscreen, g w It is a function that modifies the maximum screen gap width of the target sub-screen, g θ It is a function that corrects the tilt angle of the target subscreen, g Δh It is a function that corrects the flatness of the target subscreen.

[0144]

[0145] in, The loss function is used to correct the second parameter of the model inside the target sub-screen.

[0146] Optionally, the step of predicting the display category of the spliced ​​screen based on the dynamic feature data, the static feature data, and the image information of the spliced ​​screen includes:

[0147] The static feature data is extracted using a static multilayer perceptron to obtain a static feature vector;

[0148] The dynamic feature data is extracted using a dynamic multilayer perceptron to obtain a dynamic feature vector;

[0149] Image feature vectors are obtained by extracting image information from the spliced ​​screen using a convolutional neural network.

[0150] Based on the static feature vector, the dynamic feature vector, and the image feature vector, a perceptual hierarchical network is used to predict and obtain the display category of the spliced ​​screen, wherein the display category includes a comfort level and a content display completeness level.

[0151] In this embodiment of the invention, before predicting the comfort level and content display integrity level of the spliced ​​screen, the data needs to be processed. First, the brightness value L collected by the sensor is processed. i Contrast ratio Co, screen color mean (Re, Gr, Bl), edge halo difference L cs Multi-screen splicing color difference value C a and total color temperature gradient value The features are concatenated to form the initial feature set D. start : Then, for D start Data cleaning, missing value imputation, and data normalization are performed to form the model training dataset D. train Finally, D train The dataset is concatenated with static feature data to obtain the final model training dataset. θ center-final ,Δh center-final ].

[0152] Subsequently, a multi-screen static multi-branch coordination network was used to train the model on the dataset. The process involves processing and predicting the comfort level and content display integrity level of the spliced ​​screens. The multi-screen static multi-branch coordination network includes a screen static feature layer (static branch network), a real-time display dynamic feature layer (dynamic feature branch network), and a monitoring image input layer (convolutional neural network). The specific operations are as follows:

[0153] The real-time monitoring image of the entire splicing screen is obtained, which is the image information Vi of the splicing screen in this article;

[0154] from Extracting static feature data: Rock = [w center-final ,θ center-final ,Δh center-final ];

[0155] from The dynamic feature data is extracted:

[0156] Extracting higher-order functions from Rock vectors using a static branching network:

[0157] H static =MLP static (Rock)

[0158] Among them, H static For static feature vectors, MLP static It is a static multilayer sensing network;

[0159] High-order function extraction from Lumi vectors based on dynamic feature branching networks:

[0160] H dynamic =MLP dynamic (Lumi)

[0161] Among them, H dynamic For dynamic feature vectors, MLP dynamic It is a dynamic multilayer sensing network;

[0162] The image feature vector H is generated by convolutional extraction of the Vi vector using a convolutional neural network. image :

[0163] H image =CNN(Vi).

[0164] The outputs of the three branch networks are merged:

[0165] H fused =Fusion(H static H dynamic H image )

[0166] Among them, H fused The feature vectors are multi-screen real-time fusion vectors, and Fusion is the fusion layer;

[0167] Furthermore, classification prediction is performed based on a perceptual hierarchical network:

[0168] y comfort ,y integrity =Softmax(W·H) fused +b)

[0169] Among them, y comfort Comfort prediction level of splicing screen, y integrity Predicted level of content display completeness on spliced ​​screens.

[0170] Furthermore, the loss function for a multi-screen static multi-branch coordination network is:

[0171]

[0172] in, For the loss value of a multi-screen static multi-branch coordination network, y com With y int These represent the actual values ​​for viewing comfort and content display completeness, respectively. com , λ int These are the weighting coefficients used to balance the two tasks.

[0173] Example 1: The display category of the video wall includes comfort level and content display completeness level. In step S104, the specific scheme for adjusting the display of the video wall according to the display category is shown in Table 1 below:

[0174] Table 1. Display Adjustment Scheme for Video Walls

[0175]

[0176]

[0177]

[0178] The comfort level of the splicing screen aims to evaluate the overall sensory experience of the audience when viewing the splicing screen, including the impact of splicing gaps, uneven brightness, color difference, etc. on visual comfort. In this embodiment of the invention, the comfort level is divided into high comfort, medium comfort, and low comfort, but the invention does not limit this.

[0179] The content display integrity level of a spliced ​​screen is used to evaluate the impact of splicing features (such as splicing gaps, color differences, uneven brightness, etc.) on the integrity and loss of content display. It involves whether the content is distorted, information is missing, or visually fragmented due to splicing. In this embodiment of the invention, it is divided into three levels: content integrity, slight loss, and significant loss, but the invention does not limit this.

[0180] The visual effect of the splicing screen is evaluated by considering the comfort level and content display completeness level of the splicing screen, which is more comprehensive. In addition, when predicting the comfort level and content display completeness level of the splicing screen, the embodiments of the present invention simultaneously consider the dynamic feature data, static feature data and image information of the splicing screen, which effectively improves the recognition accuracy of the splicing screen display.

[0181] Example 2: Figure 3 As shown, the process for optimizing and adjusting the display of spliced ​​screens using the multi-screen splicing display optimization method provided by this invention is as follows:

[0182] Step S301: Collect data from sensor measurement points using a pre-set brightness sensor and color sensor;

[0183] Step S302: The data collected by the sensor is cleaned, missing values ​​are supplemented, and data is normalized. Dynamic feature data is extracted according to the method provided in the embodiment of the present invention. The dynamic feature data includes, but is not limited to, brightness value, contrast value, edge halo difference value, splicing color difference value and color temperature gradient value.

[0184] Step S303: According to the method provided in the embodiment of the present invention, the static feature data of each sub-screen in the splicing screen is predicted by using a pre-built parameter correction model. The static feature data includes, but is not limited to, edge gap width, tilt angle and flatness.

[0185] Step S304: Based on the image information of the splicing screen acquired in real time, the dynamic feature data acquired in step S302, and the static feature data acquired in step S303, a multi-screen static multi-branch coordination network is used to make predictions to obtain the comfort level and content display completeness of the splicing screen.

[0186] Step S305: Based on the comfort level and content display completeness of the splicing screen, generate optimization measures to dynamically adjust the display screen of the splicing screen. For details, please refer to Embodiment 1.

[0187] In step S306, during the adjustment process in step S305, both manual and system personnel continuously monitor and adjust the display parameters;

[0188] Step S307: Generate a display optimization process report for the splicing screen implemented in the above steps, and proceed to step S301 to continue optimizing the display.

[0189] like Figure 4 As shown, this embodiment of the invention also provides a display optimization device for multi-screen splicing, comprising:

[0190] The first acquisition module 401 is used to obtain dynamic feature data of the splicing screen based on the data collected by sensors pre-arranged in the splicing screen.

[0191] The first calculation module 402 is used to identify each sub-screen in the splicing screen according to the image information of the splicing screen, and obtain static feature data corresponding to each sub-screen.

[0192] The first prediction module 403 is used to predict the display category of the splicing screen based on the dynamic feature data, the static feature data and the image information of the splicing screen.

[0193] The first adjustment module 404 is used to adjust the display of the splicing screen according to the display category.

[0194] Optionally, the device further includes:

[0195] The second acquisition module is used to acquire first parameters of at least one target point on the splicing screen, wherein the first parameters include the adjacent side length of the splicing seam of the sub-screen to which the target point belongs and the number of corners of the splicing seam within the target range corresponding to the target point.

[0196] The second calculation module is used to calculate the sensor arrangement density corresponding to the target point based on the adjacent side length and the number of corners; wherein, within the target range corresponding to the target point, multiple sensors are set according to the sensor arrangement density.

[0197] Optionally, the dynamic feature data in the first acquisition module 401 includes brightness value, contrast value, edge halo difference value, splicing color difference value, and color temperature gradient value;

[0198] The first acquisition module 401 includes:

[0199] The first acquisition submodule is used to acquire the brightness value, contrast value, color temperature value and color average value of the splicing screen in real time based on the data collected by the sensors pre-arranged in the splicing screen.

[0200] The first calculation submodule is used to calculate the edge halo difference value of the splicing screen based on the brightness values ​​of two adjacent splicing subscreens in the splicing screen;

[0201] The second calculation submodule is used to calculate the splicing color difference value of the splicing screen based on the average color value of two adjacent spliced ​​subscreens in the splicing screen.

[0202] The third calculation submodule is used to calculate the color temperature gradient value of the splicing screen based on the color temperature value and the color temperature gradient operator corresponding to the splicing screen, wherein the color temperature gradient operator corresponding to different splicing areas in the splicing screen is different.

[0203] Optionally, the first computing submodule includes:

[0204] The first calculation unit is used to calculate the edge halo difference value of the splicing screen based on the brightness values ​​of the edges of two adjacent sub-screens in the splicing screen and the brightness values ​​of the corners of the two sub-screens.

[0205] Optionally, the second calculation submodule includes:

[0206] The second calculation unit is used to calculate the splicing color difference value of the splicing screen based on the average color value of the edges of two adjacent spliced ​​sub-screens and the average color value of the corners of the two sub-screens.

[0207] Optionally, the third computing submodule includes:

[0208] The sixth calculation unit is used to calculate the unidirectional horizontal color temperature gradient operator and the unidirectional vertical color temperature gradient operator corresponding to the splicing area of ​​the adjacent edge in the splicing screen based on the color temperature value.

[0209] The seventh calculation unit is used to calculate, based on the color temperature value, the unidirectional horizontal color temperature gradient operator, the unidirectional vertical color temperature gradient operator, the positive diagonal color temperature gradient operator, and the negative diagonal color temperature gradient operator corresponding to the adjacent corner splicing area in the splicing screen;

[0210] The eighth calculation unit is used to calculate the color temperature gradient value of the splicing screen based on the unidirectional horizontal color temperature gradient operator and the unidirectional vertical color temperature gradient operator corresponding to the splicing area of ​​the adjacent edge in the splicing screen, as well as the unidirectional horizontal color temperature gradient operator, the unidirectional vertical color temperature gradient operator, the positive diagonal color temperature gradient operator and the negative diagonal color temperature gradient operator corresponding to the splicing area of ​​the adjacent corner in the splicing screen.

[0211] Optionally, the first computing module 402 includes:

[0212] The first extraction submodule is used to extract the image information of the first region corresponding to the target subscreen based on the image information of the spliced ​​screen for the currently processed target subscreen.

[0213] The first prediction submodule is used to predict the spatial type of the target subscreen by using a convolutional neural network based on the image information of the first region corresponding to the target subscreen, wherein the number of subscreens spliced ​​is different for subscreens of different spatial types.

[0214] The second prediction submodule is used to make predictions based on the image information of the first region corresponding to the target subscreen and the spatial type corresponding to the target subscreen, and to obtain the initial static feature data corresponding to the target subscreen and the first subscreen respectively. The first subscreen is a subscreen that is spliced ​​together with the target subscreen. The static feature data includes edge gap width, tilt angle and flatness.

[0215] The first correction submodule is used to adjust the target subscreen and the first subscreen based on the initial static feature data respectively, using a pre-trained parameter correction model, to obtain the final static feature data corresponding to the target subscreen.

[0216] Optionally, the first correction submodule includes:

[0217] The first correction unit is used to adjust the target sub-screen based on the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively, using a pre-trained parameter correction model between multiple screens, to obtain the adjusted static feature data corresponding to the target sub-screen.

[0218] The second correction unit is used to adjust the target sub-screen based on the adjusted static feature data corresponding to the target sub-screen using a pre-trained parameter correction model within a single screen, so as to obtain the final static feature data corresponding to the target sub-screen.

[0219] Optionally, the first prediction module 403 includes:

[0220] The first extraction submodule is used to extract the static feature data through a static multilayer perceptron to obtain a static feature vector;

[0221] The second extraction submodule is used to extract the dynamic feature data through a dynamic multilayer perceptron to obtain a dynamic feature vector;

[0222] The third extraction submodule is used to extract image information from the spliced ​​screen using a convolutional neural network to obtain image feature vectors;

[0223] The third prediction submodule is used to predict the display category of the spliced ​​screen using a perceptual hierarchical network based on the static feature vector, the dynamic feature vector, and the image feature vector. The display category includes a comfort level and a content display completeness level.

[0224] It should be noted that the embodiments of this device are devices corresponding to the embodiments of the above methods. All implementations in the embodiments of the above methods are applicable to the embodiments of this device and can achieve the same technical effect.

[0225] This invention also provides a network device, including: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the multi-screen splicing display optimization method as described in any of the preceding claims and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0226] This invention also provides a readable storage medium, comprising: a program stored on the readable storage medium, wherein when the program is executed by a processor, it implements the display optimization steps for multi-screen splicing as described in any of the preceding claims, and achieves the same technical effect; to avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0227] This invention also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the display optimization steps of the multi-screen splicing screen as described in any of the preceding claims, and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0228] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0229] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for optimizing the display of multi-screen spliced ​​images, characterized in that, include: Based on the data collected by sensors pre-placed within the splicing screen, dynamic characteristic data of the splicing screen is obtained; Based on the image information of the splicing screen, each sub-screen within the splicing screen is identified to obtain static feature data corresponding to each sub-screen. Based on the dynamic feature data, the static feature data, and the image information of the spliced ​​screen, a prediction is made to obtain the display category of the spliced ​​screen; Adjust the display of the splicing screen according to the display category.

2. The method for optimizing the display of multi-screen spliced ​​images according to claim 1, characterized in that, The method further includes: Obtain a first parameter for at least one target point on the splicing screen, wherein the first parameter includes the adjacent side length of the splicing seam of the sub-screen to which the target point belongs and the number of corners of the splicing seam within the target range corresponding to the target point; The sensor arrangement density corresponding to the target point is calculated based on the adjacent side length and the number of corners; wherein, within the target range corresponding to the target point, multiple sensors are arranged according to the sensor arrangement density.

3. The method for optimizing the display of multi-screen spliced ​​images according to claim 1, characterized in that, The dynamic feature data includes brightness value, contrast value, edge halo difference value, splicing color difference value, and color temperature gradient value; The step of obtaining dynamic feature data of the splicing screen based on data collected by sensors pre-positioned within the splicing screen includes: Based on the data collected by sensors pre-placed within the splicing screen, the brightness, contrast, color temperature, and average color of the splicing screen are obtained in real time. The edge halo difference value of the splicing screen is calculated based on the brightness values ​​of two adjacent sub-screens in the splicing screen. The color difference value of the splicing screen is calculated based on the average color value of two adjacent sub-screens in the splicing screen. The color temperature gradient value of the splicing screen is calculated based on the color temperature value and the color temperature gradient operator corresponding to the splicing screen, wherein the color temperature gradient operator is different for different splicing areas in the splicing screen.

4. The method for optimizing the display of multi-screen spliced ​​images according to claim 3, characterized in that, The step of calculating the edge halo difference value of the spliced ​​screen based on the brightness values ​​of two adjacent sub-screens in the spliced ​​screen includes: The edge halo difference value of the spliced ​​screen is calculated based on the brightness values ​​of the edges of two adjacent sub-screens and the brightness values ​​of the corners of the two sub-screens.

5. The method for optimizing the display of multi-screen spliced ​​images according to claim 3, characterized in that, The step of calculating the splicing color difference value of the splicing screen based on the average color value of two adjacent sub-screens in the splicing screen includes: The splicing color difference value of the splicing screen is calculated based on the average color value of the edges of two adjacent sub-screens and the average color value of the corners of the two sub-screens.

6. The method for optimizing the display of multi-screen spliced ​​images according to claim 3, characterized in that, The step of calculating the color temperature gradient value of the splicing screen based on the color temperature value and the color temperature gradient operator corresponding to the splicing screen includes: Based on the color temperature value, calculate the unidirectional horizontal color temperature gradient operator and the unidirectional vertical color temperature gradient operator corresponding to the splicing area of ​​the adjacent edge in the splicing screen; Based on the color temperature value, calculate the unidirectional horizontal color temperature gradient operator, the unidirectional vertical color temperature gradient operator, the positive diagonal color temperature gradient operator, and the negative diagonal color temperature gradient operator corresponding to the splicing area of ​​the adjacent corner in the splicing screen. The color temperature gradient value of the splicing screen is calculated based on the unidirectional horizontal and unidirectional vertical color temperature gradient operators corresponding to the splicing areas of adjacent edges in the splicing screen, as well as the unidirectional horizontal, unidirectional vertical, positive diagonal, and negative diagonal color temperature gradient operators corresponding to the splicing areas of adjacent corners in the splicing screen.

7. The method for optimizing the display of multi-screen spliced ​​images according to claim 1, characterized in that, The step of identifying each sub-screen within the spliced ​​screen based on the image information of the spliced ​​screen, and obtaining static feature data corresponding to each sub-screen, includes: For the target sub-screen currently being processed, extract the image information of the first region corresponding to the target sub-screen based on the image information of the spliced ​​screen; Based on the image information of the first region corresponding to the target sub-screen, a convolutional neural network is used to predict and obtain the spatial type corresponding to the target sub-screen. The number of sub-screens spliced ​​together corresponds to different spatial types. Based on the image information of the first region corresponding to the target sub-screen and the spatial type corresponding to the target sub-screen, prediction is performed to obtain the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively. The first sub-screen is a sub-screen that is spliced ​​together with the target sub-screen. The static feature data includes edge gap width, tilt angle and flatness. Based on the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively, the parameters are adjusted using a pre-trained parameter correction model to obtain the final static feature data corresponding to the target sub-screen.

8. The method for optimizing the display of multi-screen spliced ​​images according to claim 7, characterized in that, The step of adjusting the initial static feature data corresponding to the target sub-screen and the first sub-screen using a pre-trained parameter correction model to obtain the final static feature data corresponding to the target sub-screen includes: Based on the initial static feature data corresponding to the target sub-screen and the first sub-screen respectively, the parameters are adjusted using a pre-trained multi-screen parameter correction model to obtain the adjusted static feature data corresponding to the target sub-screen. Based on the adjusted static feature data corresponding to the target sub-screen, the model is adjusted using a pre-trained parameter correction model within a single screen to obtain the final static feature data corresponding to the target sub-screen.

9. The method for optimizing the display of multi-screen spliced ​​images according to claim 1, characterized in that, The step of predicting the display category of the spliced ​​screen based on the dynamic feature data, the static feature data, and the image information of the spliced ​​screen includes: The static feature data is extracted using a static multilayer perceptron to obtain a static feature vector; The dynamic feature data is extracted using a dynamic multilayer perceptron to obtain a dynamic feature vector; Image feature vectors are obtained by extracting image information from the spliced ​​screen using a convolutional neural network. Based on the static feature vector, the dynamic feature vector, and the image feature vector, a perceptual hierarchical network is used to predict and obtain the display category of the spliced ​​screen, wherein the display category includes a comfort level and a content display completeness level.

10. A display optimization device for multi-screen splicing, characterized in that, include: The first acquisition module is used to obtain dynamic feature data of the splicing screen based on the data collected by sensors pre-arranged in the splicing screen. The first calculation module is used to identify each sub-screen in the splicing screen according to the image information of the splicing screen, and obtain static feature data corresponding to each sub-screen. The first prediction module is used to predict the display category of the splicing screen based on the dynamic feature data, the static feature data and the image information of the splicing screen. The first adjustment module is used to adjust the display of the splicing screen according to the display category.

11. A network device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the display optimization method for multi-screen splicing as described in any one of claims 1 to 9.

12. A readable storage medium, characterized in that, include: The readable storage medium stores a program that, when executed by a processor, implements the steps of display optimization of multi-screen splicing as described in any one of claims 1 to 9.

13. A computer program product, characterized in that, It includes computer instructions, which, when executed by a processor, implement the steps of display optimization of the multi-screen splicing image as described in any one of claims 1 to 9.

Images