Light guide plate dot distribution optimization design method and system based on deep learning

Abstract

The invention belongs to the technical field of light guide plate lattice point design, and provides a light guide plate lattice point distribution optimization design method and system based on deep learning, and the method comprises the steps: carrying out the type analysis of a plurality of lattice points on a light guide plate through a convolutional neural network, and recognizing light guide main lattice points and light guide auxiliary lattice points; according to the coordinate positions of the light guide main lattice points and the light guide auxiliary lattice points on the light guide plate, the light guide plate is subjected to regional division by using a convolutional neural network to obtain divided sub-regions, matching analysis on the light guide intensity is performed in combination with an actual light guide sub-region of the light guide plate, and non-anastomotic sub-regions and anastomotic sub-regions are screened out; therefore, the matching degree of the light guide intensities of different areas on the light guide plate in the specific environment is reflected, the areas with poor light guide performance of the light guide plate used in the specific environment can be quickly positioned in a targeted manner, and a direction is provided for subsequent optimization or adjustment.

Description

Technical Field

[0001] The present invention belongs to the technical field of light guide plate dot design, and specifically relates to an optimization design method and system for the dot distribution of a light guide plate based on deep learning. Background Art

[0002] In today's display technology and lighting field, as a key component, the light guide performance of the light guide plate directly affects the display effect and lighting quality of the product. During the design and manufacturing process of the light guide plate, accurately identifying and optimizing the dot distribution to improve the light guide performance has always been the focus of the industry. Currently, the identification of the types of dots on the light guide plate mainly relies on traditional algorithms or manual experience, making it difficult to accurately distinguish the main light guide dots and the auxiliary light guide dots, and unable to provide sufficiently targeted data for the subsequent optimization of the dot distribution design. As a result, the optimization work lacks a scientific basis and is inefficient.

[0003] In the prior art, when analyzing the light guide intensity matching degree in different regions of the light guide plate, there is a lack of effective means. It is impossible to reasonably divide the light guide plate into regions by combining the coordinate positions of the main light guide dots and the auxiliary light guide dots and conduct accurate light guide intensity matching analysis, making it difficult to quickly locate the regions with poor light guide performance, and resulting in a lack of a clear direction for subsequent optimization or adjustment. In addition, the accuracy of the light guide performance-related model constructed in the prior art is insufficient. When constructing the model, the complex relationship between the dot spacing and the light guide intensity is not fully considered, resulting in difficult-to-guarantee optimization accuracy when using the model to optimize the light guide performance of non-matching sub-regions, being unable to effectively solve the problem of insufficient light guide performance in individual sub-regions on the light guide plate under specific environments, and being difficult to reduce the light guide intensity and angle differences, thereby affecting the light guide uniformity and overall quality.

[0004] Therefore, the present invention provides an optimization design method and system for the dot distribution of a light guide plate based on deep learning. Summary of the Invention

[0005] In order to make up for the deficiencies of the prior art and solve at least one of the technical problems proposed in the background art.

[0006] The technical solution adopted by the present invention to solve its technical problems is as follows: In a first aspect, an optimization design method for the dot distribution of a light guide plate based on deep learning includes: Performing type analysis on the dots on the light guide plate through a convolutional neural network to identify the main light guide dots and the auxiliary light guide dots; According to the coordinate positions of the main light guide dots and the auxiliary light guide dots on the light guide plate, using a convolutional neural network to divide the light guide plate into sub-regions, and combining the required light guide intensity of the light guide sub-regions on the actual light guide plate to conduct light guide intensity matching analysis, and screening out non-matching sub-regions and matching sub-regions; Respectively obtain the dot pitch between the main light - guiding dots and the auxiliary light - guiding dots within the anastomosis sub - region, combine the light - guiding intensity within the anastomosis sub - region, and perform deep learning through a convolutional neural network to construct a dot - pitch - light - guiding intensity model; According to the constructed dot - pitch - light - guiding intensity model, obtain the non - anastomosis pitch optimization value, and formulate an optimization strategy based on the non - anastomosis pitch optimization value.

[0007] As a further solution of the present invention: Analyze the types of dots on the light - guiding plate through a convolutional neural network, and the process is as follows: Arbitrarily select a dot on the light - guiding plate as the target analysis dot, obtain the scattering intensity of the target analysis dot under light with different incident angles, and sort them from large to small according to the scattering intensity to obtain a scattering intensity convolution set; In the scattering intensity convolution set, combine the scattering intensities within adjacent convolution layers to obtain a convolution intensity combination; Input the convolution intensity combination into the Euclidean distance formula to output the target intensity value.

[0008] As a further solution of the present invention: The identification process of the main light - guiding dots and the auxiliary light - guiding dots is as follows: If the target intensity value is greater than the target intensity threshold, it is recorded as an auxiliary light - guiding dot; If the target intensity is less than or equal to the target intensity threshold, it is recorded as a main light - guiding dot.

[0009] As a further solution of the present invention: According to the coordinate positions of the main light - guiding dots and the auxiliary light - guiding dots on the light - guiding plate, use a convolutional neural network to divide the light - guiding plate into regions, and the process is as follows: Use a convolutional neural network to perform a spatial transformation on the light - guiding area of the light - guiding plate to obtain a two - dimensional space model; Within the two - dimensional space model, respectively obtain the coordinate points corresponding to the main light - guiding dots and the auxiliary light - guiding dots, and divide the two - dimensional space model to obtain sub - divided regions.

[0010] As a further solution of the present invention: Combine the actual light - guiding sub - regions of the light - guiding plate to perform a matching analysis of the light - guiding intensity to obtain a light - guiding angle matching ratio, and the process is as follows: Perform an averaging process on the scattering intensities of the main light - guiding dots within the sub - divided regions at different incident angles, and output the main light - guiding intensity; Extract the maximum scattering intensity and the minimum scattering intensity of the auxiliary light - guiding dots within the sub - divided regions at different incident angles respectively, and perform an averaging process to output the auxiliary light - guiding intensity; Sum the main light guiding intensity and the auxiliary light guiding intensity, calculate the output to obtain the light guiding intensity value, subtract it from the light guiding intensity corresponding to the actual light guiding sub-region, take the absolute value to obtain the light guiding intensity matching value, and obtain the proportion of the light guiding intensity corresponding to the actual light guiding sub-region, to obtain the light intensity matching ratio.

[0011] As a further solution of the present invention: The screening process of the non-matching sub-region and the matching sub-region is as follows: Extract the scattering angle of the main light guiding network points in the divided sub-region, subtract it from the light guiding angle corresponding to the actual light guiding sub-region, take the absolute value to obtain the light guiding angle matching value, and obtain the proportion of the light guiding angle corresponding to the actual light guiding sub-region, to obtain the light guiding angle matching ratio; Perform geometric product calculation on the light guiding intensity matching ratio and the light guiding angle matching ratio, and output to obtain the matching evaluation value; If the matching evaluation value is greater than the matching evaluation threshold, it is recorded as a non-matching sub-region; If the matching evaluation value is less than or equal to the matching evaluation threshold, it is recorded as a matching sub-region.

[0012] As a further solution of the present invention: Perform deep learning through a convolutional neural network to obtain a suspected correlation coefficient, and the process is as follows: In the two-dimensional space model, extract the coordinates of the main light guiding network points and the coordinates of the auxiliary light guiding network points in each matching sub-region, and input them into the coordinate point distance formula to obtain the network point spacing; Extract the matching evaluation value corresponding to each matching sub-region, combine it with the corresponding network point spacing, and incorporate it into the convolutional layer in the convolutional neural network to obtain multiple spacing matching analysis layers, and sort the multiple spacing matching analysis layers according to the size of the matching evaluation value to obtain a spacing matching sequence; Divide the spacing matching sequence into a deep training set and a deep validation set, and combine the adjacent spacing matching analysis layers in the deep training set to obtain multiple adjacent analysis groups; In the adjacent analysis groups, respectively perform difference processing on the matching evaluation values corresponding to the adjacent spacing matching analysis layers and the network point spacing to obtain the unit matching difference and the unit spacing difference; After performing ratio calculation on the unit matching difference and the unit spacing difference corresponding to each adjacent analysis group and then performing averaging processing, obtain the suspected correlation coefficient.

[0013] As a further solution of the present invention: The construction process of the spacing-light guiding intensity model is as follows: Based on the suspected correlation coefficient, combine the network point spacing corresponding to the spacing matching analysis layer in the deep validation set, perform product calculation, and output to obtain the suspected matching evaluation value; Combine all suspected matching evaluation values with the matching evaluation values in the spacing matching analysis layer within the corresponding depth verification set to obtain multiple model verification groups, and input them into the Euclidean calculation formula to output the verification difference value; If the verification difference value is less than or equal to the verification difference threshold, it is recorded as the model correlation coefficient. Based on the model correlation coefficient, with the X-axis being the dot pitch and the Y-axis being the matching evaluation value corresponding to the anastomotic sub-region, construct a dot pitch - light guiding intensity model, and the corresponding model equation is: , where is expressed as the correlation coefficient.

[0014] As a further solution of the present invention: Obtain the non-anastomotic pitch optimization value, and formulate an optimization strategy according to the non-anastomotic pitch optimization value. The process is as follows: Extract the model equation corresponding to the dot pitch - light guiding intensity model ; Subtract all the matching evaluation values from the matching evaluation threshold, take the absolute value, and output the matching evaluation difference; Input it into the model equation corresponding to the dot pitch - light guiding intensity model to output the non-anastomotic pitch optimization value; Compare the magnitudes of all the non-anastomotic pitch optimization values corresponding to the non-anastomotic sub-regions, and optimize the non-anastomotic sub-regions in descending order until all the non-anastomotic sub-regions on the light guide plate are optimized.

[0015] In a second aspect, a light guide plate dot distribution optimization design system based on deep learning includes: Dot type recognition module: Analyze the types of dots on the light guide plate through a convolutional neural network to identify the main light guiding dots and the auxiliary light guiding dots; Anastomosis analysis and evaluation module: According to the coordinate positions of the main light guiding dots and the auxiliary light guiding dots on the light guide plate, use a convolutional neural network to divide the light guide plate into sub-regions, and combine the required light guiding intensity of the light guiding sub-regions on the actual light guide plate to perform light guiding intensity matching analysis, and screen out the non-anastomotic sub-regions and the anastomotic sub-regions; Deep learning construction module: Respectively obtain the dot pitch between the main light guiding dots and the auxiliary light guiding dots within the anastomotic sub-region, combine the light guiding intensity within the anastomotic sub-region, and perform deep learning through a convolutional neural network to construct a dot pitch - light guiding intensity model; Model optimization design module: According to the constructed dot pitch - light guiding intensity model, obtain the non-anastomotic pitch optimization value, and formulate an optimization strategy according to the non-anastomotic pitch optimization value.

[0016] The beneficial effects of the present invention are as follows: 1. The present invention analyzes the types of multiple dots on a light guide plate through a convolutional neural network, identifies the main light guide dots and the auxiliary light guide dots, thereby identifying the types of dots on the light guide plate according to the target intensity value, providing targeted design optimization data for the subsequent optimization work on the dot distribution design of the light guide plate, and dividing the light guide plate into regions by using the convolutional neural network according to the coordinate positions of the main light guide dots and the auxiliary light guide dots on the light guide plate to obtain sub-regions, and combining the actual light guide sub-regions of the light guide plate to perform a matching analysis on the light guide intensity, screening out non-matching sub-regions and matching sub-regions, and then reflecting the matching degree of the light guide intensity in different regions of the light guide plate under a specific environment, which helps to quickly locate the regions with poor light guide performance of the light guide plate used in a specific environment and provides a direction for subsequent optimization or adjustment; 2. The present invention obtains the dot spacing between the main light guide dots and the auxiliary light guide dots in the matching sub-region, combines the light guide intensity in the matching sub-region, and performs deep learning through a convolutional neural network to construct a spacing-light guide intensity model, which not only improves the accuracy of the constructed model, but also improves the accuracy when optimizing the light guide performance of the non-matching sub-region according to the model. According to the constructed spacing-light guide intensity model, a non-matching spacing optimization value is obtained, and an optimization strategy is formulated according to the non-matching spacing optimization value, effectively solving the problem of insufficient light guide performance in individual sub-regions of the light guide plate under a specific environment, improving the light guide performance of the light guide plate, reducing the light guide intensity and angle difference, and improving the light guide uniformity and overall quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described below with reference to the drawings.

[0018] Figure 1 is a flowchart of the steps of a method for optimizing the dot distribution design of a light guide plate based on deep learning according to the present invention; Figure 2 is a schematic diagram of a system for optimizing the dot distribution design of a light guide plate based on deep learning according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0020] Embodiment 1 Please refer to Figure 1 As shown, a method for optimizing the dot distribution design of a light guide plate based on deep learning according to an embodiment of the present invention includes the following steps: Step 1: Analyze the types of multiple dots on the light guide plate through a convolutional neural network, and identify the main light guide dots and the auxiliary light guide dots; In some embodiments, any dot on the light guide plate is selected as the target analysis dot; Perform scattering intensity analysis on the target analysis dot to obtain the target intensity value. The process is as follows: Use the simulated light source to emit light at different incident angles onto the target analysis dot, record the corresponding scattering intensity after each emission, and input the recorded scattering intensity into the convolutional layer in the order of each emission of the simulated light source to obtain the scattering intensity convolution set; Among them, each convolutional layer only includes one scattering intensity; In the scattering intensity convolution set, combine the scattering intensities in adjacent convolutional layers to obtain the convolution intensity combination; Input the convolution emphasis combination into the Euclidean distance formula, and output to obtain the target intensity value; Specifically, the Euclidean distance formula is: , calculate to obtain the target intensity value , q represents the total number of convolution intensity combinations, represents the scattering intensity ranked in the q-th convolution intensity combination in the convolutional layer, represents the scattering intensity ranked in the q-th convolution intensity combination in the convolutional layer; It should be noted that the purpose of using the Euclidean distance formula is that: the Euclidean distance formula essentially quantifies the distance between two points in the spatial dimension. Now, taking the scattering intensities of the dots on the light guide plate at two different incident angles as two points in the spatial dimension, the Euclidean distance formula can quantify the difference between the scattering intensities of the dots on the light guide plate at different incident angles; Moreover, since the design function of the main dots on the light guide plate is to have a small difference in scattering intensity at different incident angles, however, the design function of the auxiliary dots on the light guide plate is to fill the dark areas of the main dots on the light guide plate. Therefore, by analyzing the difference in scattering intensity between the dots on the light guide plate at different incident angles through the Euclidean distance formula, the recognition accuracy of the dot types on the light guide plate is improved, providing targeted design optimization data for the subsequent distribution design optimization work of the dots on the light guide plate; It can be understood that the meaning represented by the target emphasis value is: the value calculated by the Euclidean distance formula, and its core meaning is to quantify the difference in scattering intensity of the target analysis dot at different incident angles. Specifically, if the target intensity value is large, it means that the change in scattering intensity of the dot at different incident angles is large. If the target intensity value is small, it means that the change in scattering intensity of the dot at different incident angles is small. Thus, the type of the dot on the light guide plate is identified according to the target intensity value, providing targeted design optimization data for the subsequent distribution design optimization work of the dots on the light guide plate; If the target intensity value is greater than the target intensity threshold, it indicates that the light received by the light guide plate dots being analyzed has a large deviation in scattering intensity after scattering at different incident angles, which is recorded as light guide auxiliary dots; If the target intensity is less than or equal to the target intensity threshold, it indicates that the light received by the light guide plate dots being analyzed has a small deviation in scattering intensity after scattering at different incident angles, which is recorded as light guide main dots; Step 2: According to the coordinate positions of the light guide main dots and light guide auxiliary dots on the light guide plate, use a convolutional neural network to divide the light guide plate into sub-regions, and combine the required light guide intensity of the light guide sub-regions on the actual light guide plate to perform light guide intensity matching analysis, and screen out non-matching sub-regions and matching sub-regions; In some embodiments, use a convolutional neural network to perform a spatial transformation on the light guide area of the light guide plate to obtain a two-dimensional space model; In the two-dimensional space model, respectively obtain the coordinate points corresponding to the light guide main dots and light guide auxiliary dots, and divide the two-dimensional space model to obtain divided sub-regions; It should be noted that there is only one light guide main dot and one light guide auxiliary dot in the divided sub-region; Obtain the light guide intensity value corresponding to the divided sub-region, and compare it with the actual light guide intensity value. The process is as follows: Average the scattering intensities of the light guide main dots in the divided sub-region at different incident angles, and output to obtain the light guide main intensity; Respectively extract the maximum scattering intensity and the minimum scattering intensity of the light guide auxiliary dots in the divided sub-region at different incident angles, and perform averaging processing to output to obtain the light guide auxiliary intensity; Perform a summation calculation on the light guide main intensity and the light guide auxiliary intensity, output to obtain the light guide intensity value, and take the absolute value of the difference between the light guide intensity value and the light guide intensity corresponding to the actual light guide sub-region to obtain the light guide intensity matching value; Perform a ratio calculation on the light guide intensity matching value and the light guide intensity corresponding to the actual light guide sub-region, and output to obtain the light guide intensity matching ratio; Extract the scattering angle of the light guide main dots in the divided sub-region, and take the absolute value of the difference between the scattering angle and the light guide angle corresponding to the actual light guide sub-region to obtain the light guide angle matching value; Perform a ratio calculation on the light guide angle matching value and the light guide angle corresponding to the actual light guide sub-region to obtain the light guide angle matching ratio; Perform a geometric product method calculation on the light guide intensity matching ratio and the light guide angle matching ratio, and output to obtain the matching evaluation value; It should be noted that the meaning represented by the matching evaluation value is as follows: comprehensively considering the difference in light guiding intensity between the main light guiding dots in the sub-region of the light guide plate and the light guiding intensity of the actual light guiding sub-region (light guiding intensity matching ratio) and the difference in light guiding angle between the light guiding angle and the actual light guiding sub-region. Specifically, the larger the matching evaluation value, the greater the differences in light guiding intensity and light guiding angle between the divided sub-region and the actual light guiding sub-region, and the lower the matching degree. If the matching evaluation value is smaller, it indicates that the differences in light guiding intensity and light guiding angle between the divided sub-region and the actual light guiding sub-region are smaller, and the matching degree is higher; Moreover, by multiplying the light guiding intensity matching ratio and the light guiding angle matching ratio through the geometric product formula, the matching degree of the light guiding performance in different regions of the light guide plate is reflected, which helps to quickly locate the regions with poor light guiding performance and provides a direction for subsequent optimization or adjustment; If the matching evaluation value is greater than the matching evaluation threshold, it indicates that the light guiding intensity matching degree between the divided sub-region and the actual light guiding sub-region is low, and it is recorded as a non-conforming sub-region; If the matching evaluation value is less than or equal to the matching evaluation threshold, it indicates that the light guiding intensity matching degree between the divided sub-region and the actual light guiding sub-region is relatively high, and it is recorded as a conforming sub-region; The specific solution of the embodiment of the present invention is as follows: through a convolutional neural network, type analysis is performed on multiple dots on the light guide plate to identify the main light guiding dots and the auxiliary light guiding dots, so as to identify the types of dots on the light guide plate according to the target intensity value, providing targeted design optimization data for subsequent optimization work on the dot distribution design of the light guide plate. And according to the coordinate positions of the main light guiding dots and the auxiliary light guiding dots on the light guide plate, the light guide plate is divided into regions by using a convolutional neural network to obtain divided sub-regions, and a matching analysis of the light guiding intensity is combined with the actual light guiding sub-region of the light guide plate to screen out non-conforming sub-regions and conforming sub-regions, thereby reflecting the matching degree of the light guiding intensity in different regions of the light guide plate under a specific environment, which helps to quickly locate the regions with poor light guiding performance of the light guide plate used under a specific environment and provides a direction for subsequent optimization or adjustment.

[0021] Embodiment 2 Please refer to Figure 1 As shown, a method for optimizing the dot distribution of a light guide plate based on deep learning according to an embodiment of the present invention further includes the following steps: Step 3: Respectively obtain the dot spacing between the main light guiding dots and the auxiliary light guiding dots in the conforming sub-region, and combine the light guiding intensity in the conforming sub-region to perform deep learning through a convolutional neural network to construct a spacing-light guiding intensity model; In some embodiments, in a two-dimensional space model, extract the coordinates of the main light guiding dots and the coordinates of the auxiliary light guiding dots in each conforming sub-region; Obtain the coordinate spacing between the main light - guiding dot coordinates and the auxiliary light - guiding dot coordinates through the coordinate - point distance formula, and calculate the ratio with the perimeter of the anastomosis sub - region to obtain the dot spacing; Extract the matching evaluation values corresponding to each anastomosis sub - region, combine with the corresponding dot spacings, and incorporate them into the convolutional layer in the convolutional neural network to obtain multiple spacing - matching analysis layers; Sort the multiple spacing - matching analysis layers according to the magnitudes of the matching evaluation values to obtain a spacing - matching sequence; Divide the spacing - matching sequence into a deep training set and a deep validation set; It should be noted that the number of spacing - matching analysis layers in the deep training set accounts for 70% of the total number of spacing - matching analysis layers in the spacing - matching sequence, and the spacing - matching analysis layers in the deep training set are selected according to the sorting of the spacing - matching analysis layers in the spacing - matching sequence; The number of spacing - matching analysis layers in the deep validation set accounts for 30% of the total number of spacing - matching analysis layers in the spacing - matching sequence, and the spacing - matching analysis layers in the deep validation set are also selected according to the sorting of the spacing - matching analysis layers in the spacing - matching sequence; Furthermore, the total number of spacing - matching analysis layers in the spacing - matching sequence is a multiple of ten. For example, if the total number of spacing - matching analysis layers in the spacing - matching sequence is 10, then the number of spacing - matching analysis layers in the deep training set is the first 7 spacing - matching analysis layers, and the spacing - matching analysis layers in the deep validation set are the last 3; Combine adjacent spacing - matching analysis layers in the deep training set to obtain multiple adjacent - analysis groups; Within the adjacent - analysis groups, perform difference processing on the matching evaluation values corresponding to adjacent spacing - matching analysis layers, and output to obtain the unit - matching difference; Similarly, perform difference processing on the dot spacings corresponding to adjacent spacing - matching analysis layers, and output to obtain the unit - spacing difference; Calculate the ratio of the unit - matching difference to the unit - spacing difference, and output to obtain the unit - correlation ratio; Perform averaging calculation on the unit - correlation coefficients corresponding to all adjacent - analysis groups, and output to obtain the suspected - correlation coefficient; Based on the suspected - correlation coefficient, combine with the dot spacings corresponding to the spacing - matching analysis layers in the deep validation set, and perform product calculation to output the suspected - matching evaluation value; Combine all suspected - matching evaluation values with the matching evaluation values in the corresponding spacing - matching analysis layers in the deep validation set to obtain multiple model - validation groups; Input the multiple model - validation groups into the Euclidean calculation formula, and output to obtain the validation - difference value; Specifically, the Euclidean calculation formula is: , calculate to obtain the validation - difference value , Denoted as the suspected matching evaluation value within the th model verification group, Denoted as the matching evaluation value within the th model verification group; If the verification difference value is greater than the verification difference threshold, it indicates that the suspected correlation coefficient has limitations and is denoted as a non-model correlation coefficient; If the verification difference value is less than or equal to the verification difference threshold, it indicates that the suspected correlation coefficient has no limitations and is denoted as a model correlation coefficient; Based on the model correlation coefficient, with the X-axis as the dot pitch and the Y-axis as the matching evaluation value corresponding to the anastomotic sub-region, a dot pitch-light guiding intensity model is constructed, and the corresponding model equation is: , where Denoted as the correlation coefficient; Step 4: According to the constructed dot pitch-light guiding intensity model, obtain the non-anastomotic pitch optimization value, and formulate an optimization strategy based on the non-anastomotic pitch optimization value; In some embodiments, extract the model equation corresponding to the dot pitch-light guiding intensity model ; Subtract all matching evaluation values from the matching evaluation threshold, take the absolute value, and output to obtain the matching evaluation difference; Input it into the model equation corresponding to the dot pitch-light guiding intensity model, and output to obtain the non-anastomotic pitch optimization value; Compare the non-anastomotic pitch optimization values corresponding to all non-anastomotic sub-regions, and optimize the non-anastomotic sub-regions in descending order until all non-anastomotic sub-regions on the light guide plate are optimized; The specific solution of the embodiment of the present invention is: respectively obtain the dot pitch between the main light guiding dots and the auxiliary light guiding dots within the anastomotic sub-region, combine the light guiding intensity within the anastomotic sub-region, and perform deep learning through a convolutional neural network to construct a dot pitch-light guiding intensity model, thereby not only improving the accuracy of the constructed model, but also improving the accuracy when optimizing the light guiding performance of non-anastomotic sub-regions according to the model. According to the constructed dot pitch-light guiding intensity model, obtain the non-anastomotic pitch optimization value, and formulate an optimization strategy based on the non-anastomotic pitch optimization value, effectively solving the problem of insufficient light guiding performance of individual sub-regions on the light guide plate in a specific environment, improving the light guiding performance of the light guide plate, reducing the light guiding intensity and angle difference, and improving the light guiding uniformity and overall quality.

[0022] Embodiment 3 Based on the same inventive concept as the method for optimizing the dot distribution of a light guide plate based on deep learning in the foregoing embodiment, as Figure 2 shown, the present application provides a system for optimizing the dot distribution of a light guide plate based on deep learning, wherein the system specifically includes: Dot type recognition module: Analyze the types of dots on the light guide plate through a convolutional neural network to identify the main light guide dots and auxiliary light guide dots; Coincidence analysis and evaluation module: According to the coordinate positions of the main light guide dots and auxiliary light guide dots on the light guide plate, use a convolutional neural network to divide the light guide plate into regions to obtain sub-regions, and combine the actual light guide sub-regions of the light guide plate to perform a matching analysis of light guide intensity, and screen out non-coincident sub-regions and coincident sub-regions; Deep learning construction module: Respectively obtain the dot pitch between the main light guide dots and the auxiliary light guide dots in the coincident sub-region, combine the light guide intensity in the coincident sub-region, and perform deep learning through a convolutional neural network to construct a pitch-light guide intensity model; Model optimization design module: According to the constructed pitch-light guide intensity model, obtain the optimized value of the non-coincident pitch, and formulate an optimization strategy based on the optimized value of the non-coincident pitch.

[0023] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An optimization design method for the dot distribution of a light guide plate based on deep learning, characterized in that: Including: Performing type analysis on the dot patterns on the light guide plate through a convolutional neural network to identify the main light guide dot patterns and the auxiliary light guide dot patterns; According to the coordinate positions of the main light guide dot patterns and the auxiliary light guide dot patterns on the light guide plate, using a convolutional neural network to divide the light guide plate into sub-regions, and combining the required light guide intensity of the light guide sub-regions on the actual light guide plate to perform light guide intensity matching analysis, and screening out non-matching sub-regions and matching sub-regions; Respectively obtaining the dot pattern spacing between the main light guide dot patterns and the auxiliary light guide dot patterns within the matching sub-regions, combining the light guide intensity within the matching sub-regions, and performing deep learning through a convolutional neural network to construct a spacing-light guide intensity model; According to the constructed spacing-light guide intensity model, obtaining the optimized value of the non-matching spacing, and formulating an optimization strategy according to the optimized value of the non-matching spacing.

2. The optimized design method for the dot distribution of a light guide plate based on deep learning according to claim 1, wherein: Performing type analysis on the dot patterns on the light guide plate through a convolutional neural network, and the process is as follows: Arbitrarily select a dot pattern on the light guide plate as the target analysis dot pattern, obtain the scattering intensity of the target analysis dot pattern under light rays with different incident angles, and sort them from large to small according to the scattering intensity to obtain a scattering intensity convolution set; In the scattering intensity convolution set, combine the scattering intensities within adjacent convolution layers to obtain a convolution intensity combination; Input the convolution intensity combination into the Euclidean distance formula, and output to obtain the target intensity value.

3. The optimized design method for the dot distribution of a light guide plate based on deep learning according to claim 2, characterized in that: The identification process of the main light guide dot patterns and the auxiliary light guide dot patterns is as follows: If the target intensity value is greater than the target intensity threshold, it is recorded as an auxiliary light guide dot pattern; If the target intensity is less than or equal to the target intensity threshold, it is recorded as the main light guide dot pattern.

4. A method for optimizing the dot distribution design of a light guide plate based on deep learning according to claim 1, characterized in that: According to the coordinate positions of the main light guide dot patterns and the auxiliary light guide dot patterns on the light guide plate, using a convolutional neural network to divide the light guide plate into regions, and the process is as follows: Using a convolutional neural network to perform spatial transformation on the light guide region of the light guide plate to obtain a two-dimensional space model; Within the two-dimensional space model, respectively obtain the coordinate points corresponding to the main light guide dot patterns and the auxiliary light guide dot patterns, and divide the two-dimensional space model to obtain divided sub-regions.

5. A method for optimizing the dot distribution design of a light guide plate based on deep learning according to claim 1, characterized in that: Combining the actual light guide sub-regions of the light guide plate to perform matching analysis on the light guide intensity to obtain a light guide angle matching ratio, and the process is as follows: Perform averaging processing on the scattering intensities of the main light guide dot patterns within the divided sub-regions under different incident angles, and output to obtain the main light guide intensity; Respectively extract the maximum scattering intensity and the minimum scattering intensity of the auxiliary light guide dot patterns within the divided sub-regions under different incident angles, and perform averaging processing to output the auxiliary light guide intensity; Perform summation calculation on the main light guide intensity and the auxiliary light guide intensity, output to obtain the light guide intensity value, subtract it from the light guide intensity corresponding to the actual light guide sub-region, take the absolute value, obtain the light guide intensity matching value, and obtain the proportion of the actual light guide sub-region corresponding light guide intensity to obtain the light intensity matching ratio.

6. The optimized design method for the dot distribution of a light guide plate based on deep learning according to claim 5, characterized in that: The screening process of the non-matching sub-regions and the matching sub-regions is as follows: Extract the scattering angles of the main light guide dot patterns within the divided sub-regions, subtract them from the light guide angles corresponding to the actual light guide sub-regions, take the absolute value, obtain the light guide angle matching value, and obtain the proportion of the actual light guide sub-region corresponding light guide angle to obtain the light guide angle matching ratio; Perform geometric product calculation on the light guide intensity matching ratio and the light guide angle matching ratio, and output the matching evaluation value; If the matching evaluation value is greater than the matching evaluation threshold, it is recorded as a non-matching sub-region; If the matching evaluation value is less than or equal to the matching evaluation threshold, it is recorded as a matching sub-region.

7. A method for optimizing the dot distribution design of a light guide plate based on deep learning according to claim 1, characterized in that: Perform deep learning through a convolutional neural network to obtain a suspected correlation coefficient. The process is as follows: In a two-dimensional space model, extract the coordinates of the main light guide points and the coordinates of the auxiliary light guide points within each matching sub-region, and input them into the coordinate point distance formula to obtain the distance between the points; Extract the matching evaluation value corresponding to each matching sub-region, combine it with the corresponding distance between the points, and incorporate it into the convolutional layer in the convolutional neural network to obtain multiple distance matching analysis layers. Then, sort the multiple distance matching analysis layers according to the size of the matching evaluation value to obtain a distance matching sequence; Divide the distance matching sequence into a deep training set and a deep validation set, and combine the adjacent distance matching analysis layers in the deep training set to obtain multiple adjacent analysis groups; Within each adjacent analysis group, perform difference processing on the matching evaluation value and the distance between the points corresponding to the adjacent distance matching analysis layers respectively to obtain the unit matching difference and the unit distance difference; After performing ratio calculation on the unit matching difference and the unit distance difference corresponding to each adjacent analysis group, perform averaging processing to obtain the suspected correlation coefficient.

8. A method for optimizing the dot distribution design of a light guide plate based on deep learning according to claim 7, characterized in that: The construction process of the distance-light guide intensity model is as follows: Based on the suspected correlation coefficient, combine the distance between the points corresponding to the distance matching analysis layer in the deep validation set to perform product calculation, and output the suspected matching evaluation value; Combine all the suspected matching evaluation values with the matching evaluation values in the distance matching analysis layer in the corresponding deep validation set to obtain multiple model validation groups, and input them into the Euclidean calculation formula to output the validation difference value; If the verification difference value is less than or equal to the verification difference threshold, it is recorded as the model correlation coefficient. Based on the model correlation coefficient, with the X-axis being the grid spacing and the Y-axis being the matching evaluation value corresponding to the anastomosis sub-region, a spacing-light guiding intensity model is constructed, and the corresponding model equation is: , where is expressed as the correlation coefficient.

9. A method for optimizing the dot distribution design of a light guide plate based on deep learning according to claim 5, characterized in that: Obtain the non-matching distance optimization value, and formulate an optimization strategy according to the non-matching distance optimization value. The process is as follows: Extraction of the model equation corresponding to the extraction pitch - light guiding intensity model ; Subtract all the matching evaluation values from the matching evaluation threshold, take the absolute value, and output the matching evaluation difference; Input it into the model equation corresponding to the distance-light guide intensity model, and output the non-matching distance optimization value; Compare the sizes of all the non-matching distance optimization values corresponding to the non-matching sub-regions, and optimize the non-matching sub-regions in descending order until all the non-matching sub-regions on the light guide plate are optimized.

10. A system for optimizing the dot distribution design of a light guide plate based on deep learning, characterized in that: Including: Dot type recognition module: Analyze the types of dots on the light guide plate through a convolutional neural network to identify the main light guide dots and the auxiliary light guide dots; Matching analysis and evaluation module: According to the coordinate positions of the main light guide dots and the auxiliary light guide dots on the light guide plate, use a convolutional neural network to divide the sub-regions of the light guide plate, and combine the required light guide intensity of the light guide sub-regions on the actual light guide plate to perform light guide intensity matching analysis, and screen out the non-matching sub-regions and the matching sub-regions; Deep learning construction module: Respectively obtain the distance between the main light guide dots and the auxiliary light guide dots within the matching sub-regions, combine the light guide intensity within the matching sub-regions, and perform deep learning through a convolutional neural network to construct a distance-light guide intensity model; Model optimization design module: Obtain the optimized value of the non-matching spacing according to the constructed spacing-light guiding intensity model, and formulate an optimization strategy based on the optimized value of the non-matching spacing.

Images