Skin moisture analysis system based on computer vision technology

Abstract

The invention relates to the technical field of image analysis, in particular to a skin moisture analysis system based on a computer vision technology, which comprises a skin region segmentation module, a moisture distribution modeling module, a texture enhancement optimization module and a moisture distribution classification module. According to the method, the boundary, the area, the pixel value and the brightness value of the skin image region are accurately analyzed, the moisture distribution state is delimited, the weighted structure diagram is generated, the moisture difference between the regions is reflected, accurate data support is provided for subsequent analysis, and the node relation and the moisture value difference are deeply analyzed; a moisture aggregation mode is disclosed, a high-accuracy distribution rule model is formed, moisture characteristics are accurately displayed, by extracting a gradient value of a moisture dense region and utilizing a texture contrast enhancement technology, moisture difference is more remarkable, image details are optimized, a finer moisture distribution diagram is presented, and by means of grading classification, the moisture state is accurately reflected, so that the moisture distribution accuracy is improved. Differentiated analysis is supported, and the analysis precision and the application value are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a skin moisture analysis system based on computer vision technology. Background Art

[0002] The field of image analysis technology is a comprehensive discipline that combines computer vision, image processing and artificial intelligence algorithms. It is mainly used to extract meaningful information from image or video data. Its core includes image acquisition, preprocessing, feature extraction, pattern recognition and presentation of analysis results. The main methods involved in this field include edge detection, segmentation algorithms, texture analysis, deep learning models (such as convolutional neural networks) and classifier design. It is widely used in medical imaging diagnosis, industrial detection, behavior monitoring, biometric recognition and environmental monitoring. Its key goal is to identify, classify, quantify or reconstruct target features or areas in the image through efficient algorithms to support decision-making in scientific research or practical applications.

[0003] Among them, the skin moisture analysis system is an application tool based on image analysis technology, which is specifically used to evaluate the moisture content and distribution status of the skin surface and its structure. The system collects skin area images and combines image processing and feature analysis technology to calculate the moisture-related parameters of the skin, thereby realizing quantitative evaluation and visualization of the skin's moisture status. Its main uses include verification of the effectiveness of skin care products, formulation of personalized skin care plans, and auxiliary diagnosis of skin diseases in clinical medicine, providing technical support for skin science research and daily care.

[0004] When dealing with skin moisture distribution, existing technologies rely on traditional image processing methods, such as simple color value and brightness value comparison, which cannot accurately capture the slight moisture differences in various parts of the skin. Most existing systems lack sufficient accuracy in boundary recognition and moisture distribution analysis, resulting in the overall distribution of skin moisture being too vague and difficult to reflect detailed changes. Due to the lack of sufficient weighting and optimization processing in region division and moisture modeling, the results do not have high-precision moisture prediction capabilities. The existing systems are relatively simple in texture optimization, and it is difficult to effectively improve the details of skin images. Therefore, they cannot truly restore the moisture state and texture of the skin. In actual applications, some skin areas have excessive concentration of moisture, and traditional methods cannot distinguish subtle differences in regions, resulting in a decrease in the accuracy of the analysis results. This not only affects the accuracy of skin care product effect verification, but also restricts the formulation of personalized care plans and the accuracy of medical diagnosis, ultimately limiting the practical application value of image analysis technology in skin moisture assessment. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a skin moisture analysis system based on computer vision technology.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: A skin moisture analysis system based on computer vision technology comprises:

[0007] The skin region segmentation module extracts the color and brightness values ​​of pixels based on the input skin image, analyzes the region boundary position and area, summarizes the region adjacency relationship, establishes the region division result, labels the weights of the moisture content range of the regional nodes, performs weighted analysis on the node connection relationship, and generates the skin region weighted structure;

[0008] The moisture distribution modeling module extracts the edge weight distribution value between nodes based on the skin area weighted structure, performs moisture clustering operation according to the distribution pattern by superimposing the characteristic difference value and the global characteristic value, obtains the moisture distribution law parameter, compares the difference of moisture values ​​between regions, analyzes the corresponding distribution relationship between regions through the difference numerical value, and obtains the moisture distribution pattern model;

[0009] The texture enhancement optimization module extracts the grayscale gradient value of the moisture-dense area based on the moisture distribution pattern model, adjusts the grayscale value according to the texture gradient characteristics, optimizes the pixel difference value in the area, performs texture contrast enhancement, hierarchically refines the enhanced texture features, generates a skin texture enhancement structure, and globally matches the regional characteristic values ​​to obtain a texture optimization feature map;

[0010] The moisture distribution classification module extracts the moisture content value of the local area and the global distribution characteristics based on the texture optimization feature map, divides the levels according to the regional moisture content range, matches the distribution pattern characteristics, compares the regions according to the distribution level, and generates the skin moisture distribution classification results.

[0011] As a further solution of the present invention, the step of obtaining the region division result is specifically as follows:

[0012] Based on the input skin image, the color value and brightness value of each pixel in the skin image are extracted, a color distribution matrix is ​​constructed for the color value of the pixel, and a brightness distribution matrix is ​​constructed for the brightness value. The overall distribution range of the color and brightness values ​​is analyzed according to the color distribution matrix and the brightness distribution matrix to obtain the preliminary distribution range parameters of the color and brightness;

[0013] According to the color and brightness preliminary distribution range parameters, the color values ​​and brightness values ​​of adjacent pixels are subjected to difference analysis, and the color and brightness difference thresholds are set based on the calculation results. The boundary analysis is performed on the adjacent pixel pairs of the pixel points, using the formula:

[0014]

[0015] Calculate the boundary value of the pixel points, mark the pixel points greater than the threshold as boundary points, and generate the region boundary matrix;

[0016] Among them, B i,j Indicates the boundary value of the pixel, C i,j and C i+1,j Respectively represent the color values ​​of adjacent pixels, L i,j and L i,j+1 Represents the brightness value of adjacent pixels, W c and W l are the normalized weight coefficients for color and brightness respectively;

[0017] Based on the region boundary matrix, the regions are identified and segmented by performing connectivity analysis on the marked boundary points, the areas of the segmented regions are counted, the shape parameters of each region are calculated according to the segmentation results, the adjacency relationship between regions is analyzed, and the region division results are established.

[0018] As a further solution of the present invention, the step of acquiring the weighted structure of the skin region is specifically:

[0019] Based on the regional division results, regional node range extraction is performed, and moisture content data of nodes in the region are extracted, upper and lower limits are analyzed, abnormal data is filtered, and moisture data of edge nodes is corrected to obtain a regional node moisture content range table;

[0020] Based on the regional node moisture content range table, normalize the upper and lower limits of the node moisture content, adjust the normalized weight, optimize the weight balance, recalibrate the weight value, and obtain the regional node weighted table;

[0021] Based on the regional node weighted table, the node connection relationship is analyzed, the geometric distance between nodes and the normalized weight difference of the moisture content are calculated, the weight matrix is ​​corrected, the node information and the weight matrix are integrated, and the skin regional weighted structure is generated.

[0022] As a further solution of the present invention, the step of obtaining the water distribution law parameters is specifically as follows:

[0023] Based on the skin area weighted structure, the edge weight values ​​between each node are extracted, and the edge weight values ​​are preliminarily summarized by calculating the local characteristic difference values ​​between the nodes and superimposing the global characteristic values. The weighted node distribution matrix is ​​obtained by combining the distribution characteristics of the edge weight values ​​between the nodes and the summary results;

[0024] The weight parameters recorded in the weighted node distribution matrix are used to perform quantitative analysis of the characteristic difference value and the global characteristic value, using the formula:

[0025]

[0026] Get the feature distribution difference matrix;

[0027] Among them, T krepresents the characteristic distribution difference value, W a and W b Represent the local and global weight values, D k is the local characteristic value, G a is the global feature value, n and m are the total number of local and global features, respectively;

[0028] Based on the characteristic distribution difference matrix, the nodes are clustered by analyzing the weighted weights between the nodes, the moisture regional characteristics are divided according to the node clustering mode, the relationship between the node distribution law and the regional characteristics is extracted, and the moisture distribution law parameters are obtained.

[0029] As a further solution of the present invention, the steps of obtaining the moisture distribution pattern model are specifically as follows:

[0030] Based on the moisture distribution law parameters, extract the moisture content values ​​of nodes in the region, group the node data by region, fit the moisture value distribution law, and generate a regional moisture distribution statistical table;

[0031] Based on the regional moisture distribution statistical table, compare the regional moisture distribution statistical parameters, analyze the moisture distribution morphological deviations between regions, integrate the inter-regional correlation patterns, and obtain regional moisture distribution difference relationship data;

[0032] Based on the regional moisture distribution difference relationship data, the moisture distribution difference values ​​and correlation parameters between regions are extracted, the parameters are aggregated into a unified model, the regional difference values ​​are mapped to parameter curves, the moisture distribution parameters are integrated, and the moisture distribution pattern model is obtained.

[0033] As a further solution of the present invention, the step of obtaining the skin texture enhancement structure is specifically:

[0034] Based on the moisture distribution pattern model, the pixel grayscale values ​​in the moisture-dense area are extracted, the grayscale gradient value is calculated, the gradient difference threshold is set according to the overall grayscale distribution of the area, and the grayscale gradient distribution matrix is ​​generated;

[0035] The regional grayscale gradient values ​​in the grayscale gradient distribution matrix are used to adjust the grayscale values ​​through texture characteristics, and the grayscale difference values ​​are optimized by combining the regional grayscale distribution and the texture gradient characteristic weights. The formula is:

[0036]

[0037] Generate optimized grayscale values;

[0038] Among them, G t is the optimized gray value, H x and H y is the grayscale difference between horizontal and vertical directions, W g and S gare the feature weight parameters in the horizontal and vertical directions, P t is the gradient distribution characteristic value, k is the texture enhancement adjustment coefficient, E g Distribute weights for texture features;

[0039] The optimized grayscale value is called to analyze the texture features, and the texture layering characteristics are refined by layering the grayscale difference value after optimizing the texture gradient according to the characteristics, the layering characteristics are superimposed, and the layering gradient is adjusted according to the texture enhancement rule to generate a skin texture enhancement structure.

[0040] As a further solution of the present invention, the step of acquiring the texture optimization feature map is specifically as follows:

[0041] Based on the skin texture enhancement structure, the region is partitioned according to the texture distribution characteristics, segmented by detecting the direction change of the regional texture, and the characteristics are extracted according to the density between the textures, and the texture characteristics are digitally reorganized to generate a regional characteristic value matrix;

[0042] Based on the regional characteristic value matrix, the characteristics of adjacent regions are compared item by item, the local values ​​of regional characteristic differences are extracted, the distribution of adjacent regional characteristics is adjusted, the overall smoothing of regional characteristic data is performed, and a global matching characteristic value matrix is ​​generated;

[0043] Based on the global matching characteristic value matrix, the weights of regional texture characteristics are redistributed, the adjacent unit values ​​of the characteristic value distribution pattern are adjusted, the global texture characteristics are optimized and reconstructed, and a texture optimization feature map is obtained.

[0044] As a further solution of the present invention, the step of obtaining the skin moisture distribution classification result is specifically:

[0045] Extract the moisture content value and global distribution characteristics of the local area in the texture optimization feature map, match the moisture content value with the local area coordinate parameters, and generate the local area moisture content mean and global moisture distribution characteristics;

[0046] The local area moisture content mean and the global moisture distribution characteristics are classified and matched, and the classification is performed according to the difference between the local area moisture content mean and the global distribution characteristic parameters, using the formula:

[0047]

[0048] Calculate the difference weight between local area and global moisture distribution to generate moisture distribution level;

[0049] Among them, W d represents the difference weight between local and global water distribution, M r Represents the average moisture content in the local area, C grepresents the global moisture distribution characteristic parameter, Q t is the matching weight coefficient, L u is the variance parameter of the local area position;

[0050] The moisture distribution level is called, the moisture distribution patterns of the local areas are classified and compared according to the level, the consistency weights of the moisture content and distribution characteristics between the areas are matched, and the skin moisture distribution classification result is generated.

[0051] Compared with the prior art, the advantages and positive effects of the present invention are:

[0052] In the present invention, by accurately analyzing the boundaries, areas, pixel values ​​and brightness values ​​of the regions in the skin image, the moisture distribution state of each region can be effectively delineated, and a weighted structure diagram is generated for each region. This structured moisture information not only reflects the moisture difference in different regions, but also can provide accurate data support for further moisture analysis. In the process of moisture distribution modeling, through the in-depth analysis of the relationship between nodes and the difference in moisture values, the potential aggregation mode of moisture in various parts of the skin can be revealed, thereby forming a moisture distribution law model with high accuracy. The model accurately shows the specific distribution characteristics of moisture on the skin surface, providing a reliable basis for the subsequent optimization of texture processing. The texture optimization process extracts the gradient value of the moisture-intensive area and uses texture contrast enhancement technology to make the moisture difference in the image more significant, thereby visually presenting a more delicate skin moisture distribution map. Through the final classification, the processing flow not only accurately reflects the moisture content of the skin, but also clearly shows the moisture state of each region, and performs differential analysis according to the moisture level of different regions. The overall solution is optimized layer by layer, greatly improving the accuracy of moisture analysis, and making the formulation of skin care programs more scientific, with higher visualization effect and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a system flow chart of the present invention;

[0054] Figure 2 It is a flow chart of the regional division result in the present invention;

[0055] Figure 3 A flowchart of the skin region weighting structure in the present invention;

[0056] Figure 4 It is a flow chart of the parameters of the water distribution law in the present invention;

[0057] Figure 5 It is a flow chart of the water distribution pattern model in the present invention;

[0058] Figure 6 is a flow chart of the skin texture enhancement structure in the present invention;

[0059] Figure 7 A flowchart of a texture optimization feature map in the present invention;

[0060] Figure 8 The flowchart of the skin moisture distribution classification result in the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0062] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0063] See also Figure 1 , the skin moisture analysis system based on computer vision technology includes:

[0064] The skin region segmentation module extracts the color and brightness values ​​of pixels based on the input skin image, analyzes the region boundary position and area, summarizes the region adjacency relationship, establishes the region division result, labels the weights of the moisture content range of the regional nodes, performs weighted analysis on the node connection relationship, and generates the skin region weighted structure;

[0065] The moisture distribution modeling module is based on the skin area weighted structure, extracts the edge weight distribution value between nodes, performs moisture clustering operation according to the distribution pattern by superimposing the characteristic difference value and the global characteristic value, obtains the moisture distribution law parameters, compares the difference of moisture values ​​between regions, analyzes the corresponding distribution relationship between regions through the difference numerical value, and obtains the moisture distribution pattern model;

[0066] The texture enhancement and optimization module extracts the grayscale gradient value of the moisture-dense area based on the moisture distribution pattern model, adjusts the grayscale value according to the texture gradient characteristics, optimizes the pixel difference value in the area, performs texture contrast enhancement, hierarchically refines the enhanced texture features, generates a skin texture enhancement structure, and globally matches the regional characteristic values ​​to obtain a texture optimization feature map;

[0067] The moisture distribution classification module is based on the texture optimization feature map, extracts the local area moisture content value and the global distribution characteristics, divides the levels according to the regional moisture content range, matches the distribution pattern characteristics, compares the regions according to the distribution level, and generates the skin moisture distribution classification results.

[0068] The regional division results include boundary position, area, and adjacency relationship. The skin area weighted structure includes moisture content range and connection relationship weight. The moisture distribution parameters include distribution value difference and global characteristic analysis. The moisture distribution pattern model includes clustering operation results and characteristic difference analysis. The skin texture enhancement structure includes texture contrast enhancement refinement and feature stratification. The texture optimization feature map includes grayscale value adjustment and global characteristic matching. The skin moisture distribution classification results include moisture content level and matching distribution pattern characteristics.

[0069] See also Figure 2 , the specific steps for obtaining the regional division results are:

[0070] Based on the input skin image, the color value and brightness value of each pixel in the skin image are extracted, a color distribution matrix is ​​constructed for the color value of the pixel, and a brightness distribution matrix is ​​constructed for the brightness value. The overall distribution range of the color and brightness values ​​is analyzed according to the color distribution matrix and the brightness distribution matrix to obtain the preliminary distribution range parameters of the color and brightness;

[0071] By scanning pixel by pixel, the RGB color values ​​and brightness values ​​of all pixels are recorded in order of coordinates. A color distribution matrix is ​​generated for the extracted color values ​​according to the distribution of pixels, wherein the color distribution matrix contains the correspondence between the pixel positions and the color values. The main color distribution intervals are extracted by calculating the distribution density of the color histogram. A brightness distribution matrix is ​​generated for the brightness values. The brightness distribution matrix records the brightness value and its position of each pixel. The overall distribution characteristics of the brightness matrix are calculated by comparing the range of brightness distribution. All pixels in the matrix are preliminarily screened based on the distribution law of color values ​​and the distribution characteristics of brightness values. Combined with the statistical analysis of the distribution change amplitude in the color histogram and the proportion of pixels in each brightness interval in the brightness range, the pixel range that meets the distribution characteristics is screened through the joint calculation results of the color and brightness matrices, and finally the preliminary distribution range parameters of color and brightness are obtained.

[0072] According to the color and brightness preliminary distribution range parameters, the color and brightness values ​​of adjacent pixels are analyzed for difference, and the color and brightness difference thresholds are set based on the calculation results. The boundary analysis is performed on the adjacent pixel pairs of the pixel points using the formula:

[0073]

[0074] Calculate the boundary value of the pixel points, mark the pixel points greater than the threshold as boundary points, and generate the region boundary matrix;

[0075] Among them, B i,j Indicates the boundary value of the pixel, C i,j and C i+1,j Respectively represent the color values ​​of adjacent pixels, L i,j and L i,j+1 Represents the brightness value of adjacent pixels, W c and W l are the normalized weight coefficients for color and brightness respectively;

[0076] The benefit of the formula is that it calculates the color and brightness differences between adjacent pixels and combines the normalized weight parameters to comprehensively reflect the joint influence of color and brightness on the boundary, thereby enhancing the accuracy of boundary calculation and reducing the deviation caused by a single parameter.

[0077] Collect the color values ​​C of adjacent pixels in the color value matrix i,j and C i+1,j , and the brightness values ​​L of adjacent pixels in the brightness value matrix i,j and L i,j+1 , calculate the difference;

[0078] Divide the color value difference and brightness value difference by the weight coefficient W respectively c and W l , normalize the color and brightness values;

[0079] The normalized color value difference and brightness value difference are squared and added, and the square root of the sum of the squares is calculated to obtain the boundary value of the corresponding pixel point in the boundary value matrix;

[0080] Let C i,j =100, C i+1,j =120,L i,j =80,L i,j+1 =60,W c =10,W l =20, substitute it into the formula and calculate as follows:

[0081]

[0082] The result shows that the boundary value of the pixel in the boundary value matrix is ​​2.236. If the boundary value is greater than the threshold, it is marked as a boundary point and further used to generate the boundary matrix.

[0083] Based on the regional boundary matrix, the connectivity analysis of the marked boundary points is performed to identify and segment the regions, the area of ​​the segmented regions is counted, the shape parameters of each region are calculated according to the segmentation results, the adjacency relationship between regions is analyzed, and the regional division results are established;

[0084] Through pixel-by-pixel comparison, the pixels in the boundary matrix that are greater than the set threshold are marked, and the pixels marked as boundary points are recorded as a boundary marker set. Connectivity analysis is performed through the boundary marker set. All pixels are divided into different regions according to the connectivity between adjacent pixels. The area of ​​the pixel set in each region is calculated separately. The area calculation is completed by counting the number of all pixels in the region. The area statistics result is called to perform shape parameter extraction on each region in turn. The shape parameter extraction calculates the perimeter of each region based on the area statistics value, and extracts the regional geometric features in combination with the ratio of the width and height of the region. The extracted geometric features are used to compare each region with the adjacent pixel set of the region, establish the regional division result, and record the adjacency relationship of all regions.

[0085] See also Figure 3 , the steps for obtaining the weighted structure of the skin region are as follows:

[0086] Based on the regional division results, the regional node range is extracted, and the moisture content data of the nodes in the region are extracted, the upper and lower limits are analyzed, the abnormal data is filtered, the moisture data of the edge nodes is corrected, and the moisture content range table of the regional nodes is obtained;

[0087] The moisture content data of each node is analyzed in turn, the initial data range of the node is grouped according to fixed intervals, the moisture distribution frequency of each interval is calculated one by one, and the credibility of the node data is determined according to the frequency distribution characteristics. The upper and lower limits are analyzed, and the moisture content of the regional nodes is calculated according to the upper and lower limits of the interval to obtain the moisture distribution density value and the mean of the current region. By multiple comparisons of the density value and the mean, it is determined whether the data is abnormal, the abnormal data is filtered, and the abnormal data points are removed from the calculation range. For the moisture content data of the edge nodes, the weighted smoothing method is used to correct the edge data based on the data distribution law of the neighboring nodes, and the proportional influence is exerted on the moisture data of the neighboring nodes through the weight distribution method. Finally, the corrected value of the edge node is calculated, and the moisture data of all nodes in the region are reintegrated to form a moisture content range table for the regional nodes, in which the content in the table clearly specifies the upper and lower limits of the moisture content range of each node and the corresponding credibility assessment.

[0088] Based on the regional node moisture content range table, the upper and lower limits of the node moisture content are normalized, the normalized weight is adjusted, the weight balance is optimized, the weight value is recalibrated, and the regional node weighted table is obtained;

[0089] By mapping the upper and lower limits to a unified normalized interval, the normalized moisture value of each node is calculated, and the normalized weight is adjusted. By analyzing the data of the uniformity of node distribution in the entire area, the degree of influence of neighboring nodes on the node, and the fluctuation characteristics of node moisture content, the normalized weight value is reallocated to optimize the weight balance. By verifying whether the node weight conforms to the overall equilibrium principle in the region, the moisture distribution density of nodes in the region is used as a constraint condition, the weight distribution of individual nodes is adjusted to eliminate the interference of extreme weight values ​​on the overall analysis accuracy, and the weight value is recalibrated. By comprehensively considering the spatial distribution characteristics and moisture content differences of regional nodes, the corrected weight value is numerically fitted to ensure that the actual moisture characteristics of the node can still be effectively reflected after the weight adjustment. Finally, a regional node weighted table is formed to lay a data foundation for subsequent analysis.

[0090] Based on the regional node weighted table, the node connection relationship is analyzed, the geometric distance between nodes and the normalized weight difference of moisture content are calculated, the weight matrix is ​​corrected, the node information and the weight matrix are integrated, and the skin regional weighted structure is generated;

[0091] By calculating the geometric distance between nodes and using triangular geometric relationships to determine the spatial connection structure between nodes one by one, the geometric distance between nodes and the normalized weight difference of moisture content are calculated, and the moisture weight value of each pair of nodes is differentially analyzed. The moisture content weight difference is used as an important parameter of the connection relationship. Combining the geometric distance and weight difference, a preliminary association matrix between nodes is established, and the weight matrix is ​​corrected. According to the balance index of the node weight distribution in the association matrix and the influence coefficient of the geometric distance, the unreasonable weight difference in the matrix is ​​corrected. The weights are reallocated through weighted adjustment to ensure that the matrix can accurately reflect the moisture distribution characteristics in the region. The node information and the weight matrix are integrated, and the moisture content data of the node is organically combined with the corrected weight matrix to further generate the weighted structure of the skin area for subsequent research and analysis of regional moisture distribution characteristics.

[0092] See also Figure 4 , the specific steps for obtaining the parameters of water distribution law are as follows:

[0093] Based on the weighted structure of the skin area, the edge weight values ​​between each node are extracted. The edge weight values ​​are preliminarily summarized by calculating the local characteristic difference values ​​between nodes and superimposing the global characteristic values. The weighted node distribution matrix is ​​obtained by combining the distribution characteristics of the edge weight values ​​between nodes with the induction results.

[0094] By reading the initial edge weight value between each node, the edge weight value is standardized using the characteristic parameters of the skin area, the edge weight values ​​between all nodes are extracted, and they are preliminarily screened to remove abnormal values ​​and invalid values. Then, all weight values ​​are normalized according to the standardization rules. The normalization process includes statistics on the maximum and minimum values, and the normalization interval range is calculated by difference. At the same time, the normalized edge weight is corrected in combination with the regional characteristic value corresponding to the node. The correction process includes comparing the difference between the node characteristic value and the edge weight distribution value, adjusting the edge weight value according to the difference, generating the final standardized edge weight, and finally obtaining the weighted node distribution matrix by analyzing the associated weight distribution of each node.

[0095] Using the weight parameters recorded in the weighted node distribution matrix, quantitative analysis of the characteristic difference value and the global characteristic value is performed using the formula:

[0096]

[0097] Get the feature distribution difference matrix;

[0098] Among them, T k represents the characteristic distribution difference value, W a and W b Represent the local and global weight values, D k is the local characteristic value, G a is the global feature value, n and m are the total number of local and global features, respectively;

[0099] The benefit of the formula is that it amplifies the impact of local characteristic differences by weighted analysis of the differences between local and global characteristic values, and at the same time, through the weight normalization processing method, the calculation results can accurately reflect the relationship between local and global characteristics, and at the same time improve the applicability of the results in subsequent analysis;

[0100] Collect local characteristic values ​​D required for characteristic analysis k and the global characteristic value set G a , by monitoring the local characteristic distribution value of the nodes in the skin area, D k , G is obtained by statistically analyzing the characteristic data of the entire skin area. a ;

[0101] Calculate the absolute value of the difference between the local feature value and the global feature value, and use it as the core input of the feature difference;

[0102] Introducing the weight parameter W a and W b , the weight value is obtained by quantifying the importance of the feature, where the local weight value W aIt is squared to amplify the impact of local differences, and the difference value is multiplied by the corresponding weight square value;

[0103] All weighted difference values ​​are accumulated and the accumulated result is divided by the sum of the global weight values ​​∑W b , realize the normalization of the global range;

[0104] Set the local characteristic value D k =40, global characteristic value G a =50, local weight value W a =0.7, global weight value W b =1.2, then the calculation is as follows:

[0105] |D k -G a |=|40-50|=10;

[0106]

[0107] ∑W b =1.2;

[0108]

[0109] The result shows that the characteristic distribution difference value is 4.083, which means that there is a certain difference between the local characteristic value and the global characteristic value under the current calculation conditions. This value will be used as the core parameter for subsequent characteristic distribution difference analysis.

[0110] Based on the characteristic distribution difference matrix, the nodes are clustered by analyzing the weighted weights between nodes, the moisture regional characteristics are divided according to the node clustering pattern, the relationship between the node distribution law and the regional characteristics is extracted, and the moisture distribution law parameters are obtained;

[0111] The weight distribution values ​​between nodes recorded in the matrix are called, and the weight values ​​of each node are classified and clustered for analysis. The analysis content includes the range distribution of node weights, the differences in associated weights between nodes, and the distribution of overall regional characteristic parameters. First, the weight parameters in the characteristic distribution difference matrix are extracted, and all weight values ​​are sorted in ascending order. The distribution frequency and number of nodes of the weight values ​​in each interval are counted. Then, the weight values ​​of each node are grouped according to the clustering rules through the classification rules. The classification rules are set according to the frequency changes of weight distribution and the node difference characteristics within the weight range. By analyzing the node characteristics within each weight distribution range in the grouping results, the characteristic superposition parameters of each group are calculated. The characteristic superposition parameters include the sum of weights, the weighted average value between nodes, and the maximum weight value. The distribution pattern between groups is recorded. The moisture distribution law parameters are obtained through the grouping characteristics analysis between nodes, including the weight distribution characteristics of the moisture area and the grouping clustering relationship between nodes.

[0112] See also Figure 5 , the specific steps for obtaining the moisture distribution pattern model are:

[0113] Based on the parameters of moisture distribution law, the moisture content values ​​of nodes in the region are extracted, the node data are grouped by region, the moisture value distribution law is fitted, and the regional moisture distribution statistical table is generated;

[0114] The corresponding moisture content value data is obtained from each node one by one. By combining the geographic spatial location and moisture characteristics of each node, the distribution trend of the node moisture data in the region is clarified. The node data is grouped by region, and the region is divided into multiple small sub-regions to ensure that the node data within each sub-region can reflect the local moisture characteristics. The node data is integrated into independent groups by region through data aggregation and rearrangement, and the moisture value distribution law is fitted. The mean, maximum, minimum and standard deviation of the node moisture content are calculated in each group. The moisture distribution curve in the region is fitted based on the parameters. According to the distribution law of each region, the local moisture distribution trend is calibrated, and a regional moisture distribution statistical table is generated. The distribution law parameters of all sub-regions are summarized into a table. The content of the statistical table includes the node distribution range, moisture mean and its distribution characteristic parameters, so as to more comprehensively display the overall picture of moisture distribution in the skin area and provide reliable data support.

[0115] Based on the regional moisture distribution statistical table, the statistical parameters of regional moisture distribution are compared, the deviation of moisture distribution morphology between regions is analyzed, the correlation patterns between regions are integrated, and the relationship data of regional moisture distribution differences are obtained;

[0116] By comparing the moisture distribution parameters in the statistical tables of each region, the mean moisture difference and standard deviation change between regions are calculated, and the difference range between the maximum and minimum values ​​is gradually analyzed. The characteristic parameter difference values ​​of each region are extracted, and the deviation of moisture distribution morphology between regions is analyzed. Combined with the spatial distribution characteristics of moisture data between regions, the deviation of moisture distribution trend in each region is calculated. According to the fitting curve of the distribution law, the moisture change trends between regions are compared, the main difference characteristics of moisture distribution morphology between regions are determined, and the regional correlation model is integrated. By analyzing the overlap of node data between adjacent regions, the correlation degree of moisture characteristics between different regions is judged. The difference data and correlation information between regions are aggregated into a unified model, and the regional moisture distribution difference relationship data is obtained, which provides an analytical basis for further exploring the moisture distribution connection between regions.

[0117] Based on the regional moisture distribution difference relationship data, the moisture distribution difference values ​​and correlation parameters between regions are extracted, the parameters are aggregated into a unified model, the regional difference values ​​are mapped to the parameter curve, the moisture distribution parameters are integrated, and the moisture distribution pattern model is obtained;

[0118] By comparing the difference values ​​of regional moisture data, analyzing the relevant parameters between regions one by one, combining the node moisture characteristics, extracting the key parameters related to moisture distribution, and grouping the difference values, the parameters are aggregated into a unified model, and the difference parameter data between regions are integrated with the associated parameters. The parameter data is standardized through a unified calculation model to ensure that the data from different regions have consistent comparative significance under the same model, and the regional difference values ​​are mapped to the parameter curve. For each set of parameter data, a parameter curve is generated according to the change trend of the difference values ​​between regions. The correlation and difference characteristics of moisture distribution between regions are reflected in the form of the curve, the moisture distribution parameters are integrated, all parameters and difference values ​​are unified into the regional difference analysis model, and the moisture distribution pattern model is generated in combination with the distribution law and the difference relationship, which provides a visual model basis and accurate data support for the study of moisture distribution characteristics in skin areas.

[0119] See also Figure 6 ,The steps for obtaining the skin texture enhancement structure are as follows:

[0120] Based on the moisture distribution pattern model, the pixel grayscale values ​​in the moisture-dense area are extracted, the grayscale gradient value is calculated, the gradient difference threshold is set according to the overall grayscale distribution of the area, and the grayscale gradient distribution matrix is ​​generated;

[0121] The pixel grayscale matrix of the moisture-dense area is analyzed by scanning pixel by pixel, the grayscale value of each pixel is called and the difference between it and the grayscale values ​​of the surrounding adjacent pixels is calculated, and the grayscale difference value is obtained by absolute value operation as the grayscale gradient value of the pixel. Then, according to the overall grayscale characteristics of the moisture-dense area, the distribution range of the grayscale gradient value is analyzed, and the threshold interval of the grayscale difference is set. The grayscale gradient points exceeding the set threshold range are classified as high-difference gradient intervals, and the remaining points are divided into low-difference gradient intervals. The grayscale gradient distribution pattern diagram is generated by statistically analyzing the distribution ratio of high-difference gradient points and low-difference gradient points. At the same time, combined with the overall change characteristics of the regional grayscale distribution, the high-difference intervals and low-difference intervals are summarized respectively, and the grayscale value distribution of the pixels in each interval is recorded. Finally, a grayscale gradient distribution matrix is ​​generated, which records the grayscale gradient value of each pixel point and the grayscale characteristic distribution in the area for subsequent processing.

[0122] Using the regional grayscale gradient value in the grayscale gradient distribution matrix, the grayscale value is adjusted by the texture characteristics, and the grayscale difference value is optimized by combining the regional grayscale distribution and the texture gradient characteristic weights. The formula is:

[0123]

[0124] Generate optimized grayscale values;

[0125] Among them, G t is the optimized gray value, H x and Hy is the grayscale difference between horizontal and vertical directions, W g and S g are the feature weight parameters in the horizontal and vertical directions, P t is the gradient distribution characteristic value, k is the texture enhancement adjustment coefficient, E g Distribute weights for texture features;

[0126] The benefit of the formula is that by introducing the gradient characteristic weight and texture enhancement adjustment coefficient, the grayscale gradients in the horizontal and vertical directions are uniformly normalized, and at the same time, the adaptability of the result to the texture distribution is improved by superimposing the characteristic weight value, making the grayscale optimization result smoother and adaptable to complex texture characteristics;

[0127] Get the horizontal grayscale difference value H from the grayscale gradient distribution matrix x and vertical grayscale difference value H y , the horizontal difference value is calculated by comparing the gray values ​​of adjacent horizontal pixels pixel by pixel, and the vertical difference value is calculated by comparing the gray values ​​of adjacent vertical pixels pixel by pixel;

[0128] Determine the characteristic weight parameters W in the horizontal and vertical directions according to the regional grayscale distribution characteristics g and S g By analyzing the changes in the regional grayscale gradient distribution range and the texture distribution characteristic value, the weight value is calculated, and the texture enhancement adjustment coefficient k and the texture characteristic distribution weight E are obtained at the same time. g ;

[0129] Substitute the above parameters into the formula, normalize the grayscale difference values ​​in the horizontal and vertical directions through normalization, and obtain the grayscale optimization result through square sum calculation;

[0130] Set the horizontal grayscale difference value H x =20, vertical grayscale difference value H y =15, horizontal feature weight W g =0.4, gradient distribution characteristic value P t =2.5, vertical characteristic weight S g =0.6, texture enhancement adjustment coefficient k=2.0, texture characteristic distribution weight E g =10, the calculation is as follows:

[0131]

[0132] The result shows that the optimized grayscale value is 45.2, which can be effectively used for further optimization of texture enhancement characteristics.

[0133] The optimized grayscale value is called to analyze the texture features, and the texture layering characteristics are refined by layering the grayscale difference value after optimizing the texture gradient according to the characteristics, and the layering characteristics are superimposed, and the layering gradient is adjusted according to the texture enhancement rule to generate a skin texture enhancement structure;

[0134] The texture characteristic distribution of the grayscale value is optimized pixel by pixel, the optimized grayscale gradient value is extracted, and it is divided into layers according to the gradient value range. The texture characteristic values ​​within the grayscale gradient range of each layer are called to generate the corresponding texture characteristic layered matrices. Then, combined with the characteristic analysis of each layered matrix, a layer-by-layer cumulative analysis is performed, and the texture enhancement characteristics of the cumulative results are optimized. The optimization process includes the removal of abnormal gradient points in the texture characteristic matrix, the normalization adjustment of the texture characteristic range, and the distribution smoothing of the accumulated gradient. The overall texture characteristic distribution is optimized by the superposition result of the layered characteristic matrix, and the optimized texture distribution result is called. By performing comprehensive analysis and induction operations on the final texture characteristic matrix, it is ensured that the result can accurately record the enhanced characteristics of the regional texture distribution, and finally the skin texture enhancement structure is obtained.

[0135] See also Figure 7 , the steps for obtaining the texture optimization feature map are as follows:

[0136] Based on the skin texture enhancement structure, the region is partitioned according to the texture distribution characteristics, segmented by detecting the direction change of the regional texture, and the characteristics are extracted according to the density between the textures, and the texture characteristics are digitally reorganized to generate a regional characteristic value matrix;

[0137] By calculating the direction vector of the texture distribution in each pixel area, analyzing the change of pixel intensity gradient point by point, clarifying the texture direction trend of each small area, and segmenting by detecting the direction change of regional texture, a fixed-step sliding analysis window is used to count the main vector of the texture direction in each window, and compare the direction difference between adjacent windows. The area with significant direction change is used as the segmentation boundary to complete the directional partitioning of the entire area. The characteristics are extracted according to the density of textures. The distribution density of texture intersection points in the unit area is calculated to extract the density characteristic value of the texture in each area. Combined with the repetitive characteristics of the local texture, the main frequency value and amplitude characteristic value of the texture are further extracted to reorganize the texture characteristics into data. The texture directionality, density, main frequency value and other characteristics of each partition are expressed in numerical form, and the extracted characteristics are arranged in regional order to finally form a regional characteristic value matrix as the core data input for subsequent analysis of skin texture.

[0138] Based on the regional characteristic value matrix, the characteristics of adjacent regions are compared item by item, the local values ​​of regional characteristic differences are extracted, the distribution of adjacent regional characteristics is adjusted, the overall regional characteristic data is smoothed, and a global matching characteristic value matrix is ​​generated;

[0139] The texture direction, density, main frequency value and other characteristics of adjacent areas in the matrix are compared item by item, the absolute difference of each characteristic item is calculated, and the cumulative sum of the differences is used as a quantitative indicator of regional characteristic differences, the main characteristic difference points between regions are clarified, the local values ​​of regional characteristic differences are extracted, and the difference values ​​in regional characteristic differences that are higher than the global mean are extracted as key area data for local adjustment. At the same time, the regional boundaries where the maximum characteristic difference is located are recorded for subsequent adjustments, and the characteristic distribution of adjacent areas is adjusted. New characteristic values ​​are calculated by interpolation at the boundaries of adjacent areas, and the mutation part of the characteristic distribution curve is smoothly transitioned to make the characteristic distribution between regions more continuous, reduce the analysis error caused by texture characteristic jumps, and perform overall smoothing of regional characteristic data. The characteristic values ​​of all regions are weighted averaged according to the density, directionality and frequency characteristics, and the weights are assigned to the inverse of the texture similarity of adjacent regions. The global matching characteristic value matrix is ​​constructed with the smoothed characteristic values ​​to provide unified global data support for subsequent texture optimization operations.

[0140] Based on the global matching characteristic value matrix, the weights of regional texture characteristics are redistributed, the adjacent unit values ​​of the characteristic value distribution pattern are adjusted, the global texture characteristics are optimized and reconstructed, and the texture optimization feature map is obtained;

[0141] By analyzing the proportion of texture characteristics in each area in the global matching characteristic value matrix, calculating the influence of regional characteristics on the entire texture distribution, reallocating weight values ​​according to the influence, adjusting the adjacent unit values ​​of the characteristic value distribution pattern, and interpolating the characteristic values ​​of adjacent units according to the values ​​of each unit in the global matching characteristic value matrix, adjusting its distribution pattern, making the characteristic values ​​of adjacent units in the matrix gradually transition, avoiding local numerical mutations, ensuring the overall consistency of texture characteristics, and optimizing and reconstructing global texture characteristics. The adjusted characteristic values ​​are used as the basic data for optimization, and the texture characteristic parameters of each area are recalculated in combination with the weights. The directional distribution, density distribution and frequency characteristic distribution curves of the regional texture are reconstructed, and the optimized characteristics are stored in matrix form to obtain the texture optimization feature map. The final characteristic matrix is ​​superimposed with the original skin texture image to show the spatial distribution and optimization results of the skin surface texture characteristics, providing an accurate reference for subsequent moisture analysis and skin quality assessment.

[0142] See also Figure 8 , the steps for obtaining the skin moisture distribution classification results are as follows:

[0143] Extract the moisture content value and global distribution characteristics of the local area in the texture optimization feature map, match the moisture content value with the local area coordinate parameters, and generate the local area moisture content mean and global moisture distribution characteristics;

[0144] By scanning the extracted texture optimization feature map pixel by pixel, grayscale conversion is performed for each pixel value, the moisture value of each pixel point is uniformly quantized into a grayscale value range of 0 to 255, and the local area is divided into multiple grid units, each grid unit contains a certain number of pixels, and then the pixels in each grid are statistically analyzed to calculate the mean moisture value of the local area. The mean is calculated using the arithmetic mean of the grayscale values ​​of all pixels. The local moisture distribution characteristics are completed through local difference analysis of the moisture mean. Taking the center point coordinates of each area as a reference, the mean deviation between adjacent areas is calculated, and the standard deviation of the moisture mean distribution in the global range is statistically calculated. The standard deviation is obtained by calculating the discrete degree of the local area mean. Then, the global moisture distribution characteristics are normalized by combining the moisture mean and standard deviation of the local area, and the normalized results are organized into a table for further analysis, and finally the local area moisture content mean and global moisture distribution characteristics are generated.

[0145] The local area moisture content mean and the global moisture distribution characteristics are classified and matched, and the classification is carried out according to the difference between the local area moisture content mean and the global distribution characteristic parameters, using the formula:

[0146]

[0147] Calculate the difference weight between local area and global moisture distribution to generate moisture distribution level;

[0148] Among them, W d represents the difference weight between local and global water distribution, M r Represents the average moisture content in the local area, C g represents the global moisture distribution characteristic parameter, Q t is the matching weight coefficient, L u is the variance parameter of the local area position;

[0149] The benefit of the formula is that by combining the comprehensive parameters of regional moisture mean, global characteristics and location differences, the calculation of moisture distribution weights is refined, which can reflect the degree of correlation between local areas and global characteristics, thereby improving the accuracy of classification results;

[0150] M r Indicates the mean moisture content of a local area, which is calculated based on the mean pixel value of the area. For example, if the pixel value of a local area is [50, 60, 70, 80], calculate M r =(50+60+70+80) / 4=65;

[0151] C g Represents the global moisture distribution characteristics, which is calculated by the statistical standard deviation of all regional means. For example, if the regional mean is [65, 70, 75], calculate Cg =(|65-70|+|70-75|+|75-65|) / 3=5;

[0152] Q t Represents the matching weight coefficient, which is calculated by normalizing the ratio of the number of pixels covered in the current local area to the maximum number of pixels covered. For example, if the number of pixels covered is 150, the maximum number of pixels covered is 200, and the minimum number of pixels covered is 100, calculate Q t =(150-100) / (200-100)=0.5;

[0153] L u Represents the variance of the pixel position in the local area. Combined with the pixel coordinates, the discrete degree in the area is calculated. For example, if the coordinates are [(1, 1), (1, 2), (2, 1), (2, 2)], calculate L u =0.5;

[0154] Substituting the above parameters into the formula

[0155] The results show that the difference weight between the local and global moisture distribution characteristics is 32.25, which is used to further classify and match the moisture distribution levels and provide accurate characteristic parameters for subsequent distribution analysis.

[0156] Call the moisture distribution level, classify and compare the moisture distribution patterns of local areas according to the level, match the consistency weights of moisture content and distribution characteristics between regions, and generate skin moisture distribution classification results;

[0157] According to the grading results of the moisture content values ​​of local areas and the corresponding distribution pattern characteristics, a parameter form of the matching mode is constructed. By classifying the moisture content values ​​of local areas and the distribution levels, the local areas within the same distribution level are compared, and the deviation between the moisture content values ​​of each area and the standard distribution level is calculated. The deviation is obtained by taking the absolute value of the current local area moisture value minus the mean of the standard level. For the moisture distribution characteristics between matching areas, the consistency weight is calculated according to the distribution level of adjacent areas. The consistency weight is obtained by the ratio of the number of pixels in the overlapping areas of the regional moisture distribution level. Further combined with the distribution statistics of the consistency weights between regions, according to the comparison results of the distribution levels, the overall distribution model is constructed, and finally the consistency of the moisture distribution pattern of the local area and the global characteristics is matched to generate the skin moisture distribution classification result.

[0158] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. The skin moisture analysis system based on computer vision technology is characterized by: The system comprises: The skin region segmentation module extracts the color and brightness values ​​of pixels based on the input skin image, analyzes the region boundary position and area, summarizes the region adjacency relationship, establishes the region division result, labels the weights of the moisture content range of the regional nodes, performs weighted analysis on the node connection relationship, and generates the skin region weighted structure; The moisture distribution modeling module extracts the edge weight distribution value between nodes based on the skin area weighted structure, performs moisture clustering operation according to the distribution pattern by superimposing the characteristic difference value and the global characteristic value, obtains the moisture distribution law parameter, compares the difference of moisture values ​​between regions, analyzes the corresponding distribution relationship between regions through the difference numerical value, and obtains the moisture distribution pattern model; The texture enhancement optimization module extracts the grayscale gradient value of the moisture-dense area based on the moisture distribution pattern model, adjusts the grayscale value according to the texture gradient characteristics, optimizes the pixel difference value in the area, performs texture contrast enhancement, hierarchically refines the enhanced texture features, generates a skin texture enhancement structure, and globally matches the regional characteristic values ​​to obtain a texture optimization feature map; The moisture distribution classification module extracts the moisture content value of the local area and the global distribution characteristics based on the texture optimization feature map, divides the levels according to the regional moisture content range, matches the distribution pattern characteristics, compares the regions according to the distribution level, and generates the skin moisture distribution classification results.

2. The skin moisture analysis system based on computer vision technology according to claim 1, characterized in that: The steps for obtaining the region division result are specifically as follows: Based on the input skin image, the color value and brightness value of each pixel in the skin image are extracted, a color distribution matrix is ​​constructed for the color value of the pixel, and a brightness distribution matrix is ​​constructed for the brightness value. The overall distribution range of the color and brightness values ​​is analyzed according to the color distribution matrix and the brightness distribution matrix to obtain the preliminary distribution range parameters of the color and brightness; According to the color and brightness preliminary distribution range parameters, the color values ​​and brightness values ​​of adjacent pixels are subjected to difference analysis, and the color and brightness difference thresholds are set based on the calculation results. The boundary analysis is performed on the adjacent pixel pairs of the pixel points, using the formula: Calculate the boundary value of the pixel points, mark the pixel points greater than the threshold as boundary points, and generate the region boundary matrix; Among them, B i,j Indicates the boundary value of the pixel, C i,j and C i+1,j Respectively represent the color values ​​of adjacent pixels, L i,j and L i,j+1 Respectively represent the brightness values ​​of adjacent pixels, W c and W l are the normalized weight coefficients for color and brightness respectively; Based on the region boundary matrix, the regions are identified and segmented by performing connectivity analysis on the marked boundary points, the areas of the segmented regions are counted, the shape parameters of each region are calculated according to the segmentation results, the adjacency relationship between regions is analyzed, and the region division results are established.

3. The skin moisture analysis system based on computer vision technology according to claim 2, characterized in that: The steps of acquiring the skin region weighted structure are specifically as follows: Based on the regional division results, regional node range extraction is performed, and moisture content data of nodes in the region are extracted, upper and lower limits are analyzed, abnormal data is filtered, and moisture data of edge nodes is corrected to obtain a regional node moisture content range table; Based on the regional node moisture content range table, normalize the upper and lower limits of the node moisture content, adjust the normalized weight, optimize the weight balance, recalibrate the weight value, and obtain the regional node weighted table; Based on the regional node weighted table, the node connection relationship is analyzed, the geometric distance between nodes and the normalized weight difference of the moisture content are calculated, the weight matrix is ​​corrected, the node information and the weight matrix are integrated, and the skin regional weighted structure is generated.

4. The skin moisture analysis system based on computer vision technology according to claim 3, characterized in that: The steps for obtaining the water distribution law parameters are specifically as follows: Based on the skin area weighted structure, the edge weight values ​​between each node are extracted, and the edge weight values ​​are preliminarily summarized by calculating the local characteristic difference values ​​between the nodes and superimposing the global characteristic values. The weighted node distribution matrix is ​​obtained by combining the distribution characteristics of the edge weight values ​​between the nodes and the summary results; The weight parameters recorded in the weighted node distribution matrix are used to perform quantitative analysis of the characteristic difference value and the global characteristic value, using the formula: Get the feature distribution difference matrix; Among them, T k represents the characteristic distribution difference value, W a and W b Represent the local and global weight values ​​respectively, D k is the local characteristic value, G a is the global feature value, n and m are the total number of local and global features, respectively; Based on the characteristic distribution difference matrix, the nodes are clustered by analyzing the weighted weights between the nodes, the moisture regional characteristics are divided according to the node clustering mode, the relationship between the node distribution law and the regional characteristics is extracted, and the moisture distribution law parameters are obtained.

5. The skin moisture analysis system based on computer vision technology according to claim 4, characterized in that: The steps for obtaining the moisture distribution pattern model are specifically as follows: Based on the moisture distribution law parameters, extract the moisture content values ​​of nodes in the region, group the node data by region, fit the moisture value distribution law, and generate a regional moisture distribution statistical table; Based on the regional moisture distribution statistical table, compare the regional moisture distribution statistical parameters, analyze the moisture distribution morphological deviations between regions, integrate the inter-regional correlation patterns, and obtain regional moisture distribution difference relationship data; Based on the regional moisture distribution difference relationship data, the moisture distribution difference values ​​and correlation parameters between regions are extracted, the parameters are aggregated into a unified model, the regional difference values ​​are mapped to parameter curves, the moisture distribution parameters are integrated, and the moisture distribution pattern model is obtained.

6. The skin moisture analysis system based on computer vision technology according to claim 5, characterized in that: The steps for obtaining the skin texture enhancement structure are specifically as follows: Based on the moisture distribution pattern model, the pixel grayscale values ​​in the moisture-dense area are extracted, the grayscale gradient value is calculated, the gradient difference threshold is set according to the overall grayscale distribution of the area, and the grayscale gradient distribution matrix is ​​generated; The regional grayscale gradient values ​​in the grayscale gradient distribution matrix are used to adjust the grayscale values ​​through texture characteristics, and the grayscale difference values ​​are optimized by combining the regional grayscale distribution and the texture gradient characteristic weights. The formula is: Generate optimized grayscale values; Among them, G t is the optimized gray value, H x and H y is the grayscale difference between horizontal and vertical directions, W g and S g are the feature weight parameters in the horizontal and vertical directions, P t is the gradient distribution characteristic value, k is the texture enhancement adjustment coefficient, E g Distribute weights for texture features; The optimized grayscale value is called to analyze the texture features, and the texture layering characteristics are refined by layering the grayscale difference value after optimizing the texture gradient according to the characteristics, the layering characteristics are superimposed, and the layering gradient is adjusted according to the texture enhancement rule to generate a skin texture enhancement structure.

7. The skin moisture analysis system based on computer vision technology according to claim 6, characterized in that: The steps for obtaining the texture optimization feature map are specifically as follows: Based on the skin texture enhancement structure, the region is partitioned according to the texture distribution characteristics, segmented by detecting the direction change of the regional texture, and the characteristics are extracted according to the density between the textures, and the texture characteristics are digitally reorganized to generate a regional characteristic value matrix; Based on the regional characteristic value matrix, the characteristics of adjacent regions are compared item by item, the local values ​​of regional characteristic differences are extracted, the distribution of adjacent regional characteristics is adjusted, the overall smoothing of regional characteristic data is performed, and a global matching characteristic value matrix is ​​generated; Based on the global matching characteristic value matrix, the weights of regional texture characteristics are redistributed, the adjacent unit values ​​of the characteristic value distribution pattern are adjusted, the global texture characteristics are optimized and reconstructed, and a texture optimization feature map is obtained.

8. The skin moisture analysis system based on computer vision technology according to claim 7, characterized in that: The steps for obtaining the skin moisture distribution classification result are specifically as follows: Extract the moisture content value and global distribution characteristics of the local area in the texture optimization feature map, match the moisture content value with the local area coordinate parameters, and generate the local area moisture content mean and global moisture distribution characteristics; The local area moisture content mean and the global moisture distribution characteristics are classified and matched, and the classification is performed according to the difference between the local area moisture content mean and the global distribution characteristic parameters, using the formula: Calculate the difference weight between local area and global moisture distribution to generate moisture distribution level; Among them, W d represents the difference weight between local and global water distribution, M r Represents the average moisture content in the local area, C g represents the global moisture distribution characteristic parameter, Q t is the matching weight coefficient, L u is the variance parameter of the local area position; The moisture distribution level is called, the moisture distribution patterns of the local areas are classified and compared according to the level, the consistency weights of the moisture content and distribution characteristics between the areas are matched, and the skin moisture distribution classification result is generated.

Images