Deep learning-based scalp detection method and system

Abstract

The invention relates to the technical field of scalp detection, in particular to a scalp detection method and system based on deep learning, and the method comprises the following steps: extracting the light intensity value distribution and spatial position value of image pixel points based on scalp image data, analyzing the regional feature boundaries of an epidermal layer, a corium layer and a hair follicle layer, and carrying out the partitioning processing of the whole region, and generating a scalp layering segmentation result. According to the method, through combined analysis of a light intensity value and a spatial position in scalp image data, layered region division is refined, and through quantification of sebaceous gland distribution area, a relation between hair follicle spacing and density and corium layer thickness characteristics, interaction characteristics of sebaceous gland secretion value, hair follicle spacing and thickness value are dynamically analyzed; the fluctuation trend of the health state in the time sequence is disclosed, the accuracy of health state analysis is improved, the efficiency and comprehensiveness of anomaly positioning are enhanced, and more scientific support is provided for medical diagnosis and beauty care.

Description

Technical Field

[0001] The present invention relates to the technical field of scalp detection, and particularly to a scalp detection method and system based on deep learning. Background Art

[0002] The technical field of scalp detection includes methods for collecting, analyzing, and processing the surface conditions of the scalp, and involves applications in multiple fields such as medical diagnosis and beauty care. The core content of this technical field is to analyze specific indicators such as the health status of the scalp, hair follicle density, and distribution of skin secretions through the acquisition and processing of scalp image data. The overall scalp detection technology includes links such as data collection, image processing and analysis, and result output. By imaging and feature extraction of the scalp surface of the detection object, and combining data analysis technology, a systematic description and evaluation of the scalp state are achieved.

[0003] Among them, the scalp detection method based on deep learning refers to a method that uses a deep learning model to automatically process scalp image data. This patent theme aims at feature recognition and state analysis in scalp image data, and uses a convolutional neural network model to segment and classify features such as the texture, color, and hair follicle morphology of scalp images in specific dimensions, and ensures the accuracy and robustness of classification through specific model training and optimization techniques. The processing process includes multi-scale feature extraction of scalp images, automatic marking and segmentation of specific regions, and result mapping operations after analysis to achieve refined processing and accurate analysis of scalp features.

[0004] When the prior art processes scalp image data, it mainly relies on simple analysis of overall features and is difficult to accurately capture the regional boundaries and stratification characteristics at different levels. This method usually ignores the individual characteristics of the epidermis, dermis, and hair follicle layer, making the assessment of the scalp health status remain at a shallow level. The scope of feature extraction is relatively limited, and it fails to refine the distribution law between pixel points, the characteristics of stratified regions, and the data of specific regions, resulting in the lack of sufficient depth in the health assessment results. In the monitoring of dynamic change trends, a single-variable static statistical model is mostly used, which is difficult to reflect the complex interaction effects between parameters and limits the accurate identification of health status fluctuations in time series. In addition, for the zonal assessment of local abnormalities, the existing methods lack a detailed description of the deviation of regional characteristics and often have difficulty in timely discovering potential problems. This deficiency may lead to an incomplete assessment range, thereby affecting the diagnosis and prevention effects of health problems. Summary of the Invention

[0005] The purpose of the present invention is to solve the disadvantages existing in the prior art, and to propose a scalp detection method and system based on deep learning.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: A scalp detection method based on deep learning, comprising the following steps:

[0007] S1: Based on the scalp image data, extract the light intensity value distribution and spatial position value of the image pixel points, analyze the regional feature boundaries of the epidermis, dermis and hair follicle layers, calculate the distribution range between pixels within the region, classify the pixel points to determine their hierarchical attribution, and perform zoning processing on the overall region to generate the scalp layered segmentation result;

[0008] S2: Based on the scalp layered segmentation result, extract the sebaceous gland distribution area value of the epidermis, fit the curve characteristics of the dermis thickness, screen the distribution relationship data of the hair follicle spacing and hair follicle density, and analyze the local and overall characteristics of the layered data to generate the layered key feature matrix;

[0009] S3: Based on the layered key feature matrix, analyze the interactive dynamic characteristics of the sebaceous gland secretion value, hair follicle spacing value and epidermis thickness value, calculate the active change rate of the sebaceous gland, compare the weight of the hair follicle distribution fluctuation range, summarize the parameter influence relationship of the interactive result, and perform verification analysis processing on the parameter collaboration to generate the multi-parameter collaborative optimization result;

[0010] S4: Based on the multi-parameter collaborative optimization result, analyze the sebaceous gland active trajectory change value, hair follicle spacing volatility and dermis thickness change value in the time series, calculate the dynamic change rate and the trajectory distribution trend, screen the characteristic fluctuation result of the time series, and generate the scalp health dynamic trend result;

[0011] S5: Based on the scalp health dynamic trend result, analyze the abnormal fluctuation value of the sebaceous gland region, the abnormal characteristic distribution value of the hair follicle spacing and the dermis thickness fluctuation value, calculate the difference value of the zoned characteristic parameters, mark the offset characteristic distribution region, and generate the scalp health zoned evaluation value.

[0012] As a further solution of the present invention, the scalp hierarchical segmentation result includes the epidermal layer region feature boundary, the dermal layer region feature boundary, the hair follicle layer region feature boundary, the pixel point distribution range, the pixel point hierarchical attribution, and the overall region partition. The hierarchical key feature matrix includes the epidermal layer sebaceous gland distribution area value, the dermal layer thickness curve characteristic, the hair follicle spacing distribution relationship, the hair follicle density distribution relationship, the local characteristics of the hierarchical data, the overall characteristics of the hierarchical data, and the health parameters. The multi-parameter collaborative optimization result includes the interactive dynamic characteristics of the sebaceous gland secretion value, the interactive dynamic characteristics of the hair follicle spacing value, the interactive dynamic characteristics of the epidermal layer thickness value, the sebaceous gland activity change rate, the weight of the hair follicle distribution fluctuation range, the parameter influence relationship, and the collaborative optimization verification result. The scalp health dynamic trend result includes the sebaceous gland activity trajectory change value, the hair follicle spacing volatility, the dermal layer thickness change value, the dynamic change rate, the trajectory distribution trend, and the time series characteristic fluctuation result. The scalp health partition evaluation value includes the abnormal fluctuation value of the sebaceous gland region, the abnormal characteristic distribution value of the hair follicle spacing, the dermal layer thickness fluctuation value, the partition characteristic parameter difference value, the offset characteristic distribution region, and the region characteristic matrix.

[0013] As a further solution of the present invention, based on the scalp image data, extract the light intensity value distribution and spatial position value of the image pixel points, analyze the regional feature boundaries of the epidermal layer, dermal layer, and hair follicle layer, calculate the distribution range between pixels within the region, classify the pixel points to determine their hierarchical attribution, and perform partition processing on the overall region. The specific steps for generating the scalp hierarchical segmentation result are as follows:

[0014] S101: Based on the scalp image data, extract the light intensity value of the image pixel points and their spatial positions in the image, establish a pixel point feature mapping of the light intensity value and spatial position, calculate the distribution density of the pixel points in the two-dimensional space, use the image coordinates to locate their positions in the image data, and store the light intensity value and position data in correspondence as a data table to generate pixel feature data;

[0015] S102: Based on the pixel feature data, evaluate the aggregation of pixel points in combination with the spatial distribution density, use the spatial adjacency characteristics to identify the pixel distribution regions of the epidermal layer, dermal layer, and hair follicle layer, calculate the regional boundary distribution range through the pixel point position change trend, and screen the hierarchical boundary lines based on the light intensity value fluctuation characteristics in combination with the regional coordinate points to generate a regional boundary mapping;

[0016] S103: Based on the regional boundary mapping, group the pixel points within the region layer by layer, calculate the change range of the light intensity value within the region, and determine the belonging layer. According to the hierarchical attribution, re-divide the spatial distribution of the pixel points into three independent regions: the epidermal layer, the dermal layer, and the hair follicle layer, to generate the scalp hierarchical segmentation result.

[0017] As a further solution of the present invention, based on the scalp layer segmentation result, the distribution area value of sebaceous glands in the epidermis layer is extracted, the thickness curve characteristics of the dermis layer are fitted, the distribution relationship data of hair follicle spacing and hair follicle density are screened, and the local and overall characteristics of the layer data are analyzed. The specific steps for generating the layer key feature matrix are as follows:

[0018] S201: Based on the scalp layer segmentation result, locate the pixel points in the epidermis layer, and screen the sebaceous gland-related pixels. Using the light intensity value and spatial position parameters, analyze the distribution of sebaceous gland pixels, calculate the total area value within the sebaceous gland region, associate the sebaceous gland area with the corresponding spatial coordinates, and store them in matrix format to generate the sebaceous gland distribution matrix;

[0019] S202: Based on the sebaceous gland distribution matrix, extract the spatial position data of the pixel points in the dermis layer, combine with the regional boundary coordinates of the epidermis layer, calculate the thickness distance, fit the thickness change curve of the dermis layer point by point, correct the spatial position deviation, and generate the curve continuity result to generate the dermis layer thickness curve matrix;

[0020] S203: Based on the dermis layer thickness curve matrix, extract the spatial positions of the pixel points in the hair follicle region, calculate the distance between adjacent hair follicles, combine with the distribution of hair follicles in the region, calculate the hair follicle density, screen the eigenvalue of the relationship between the spacing change and the density distribution, and integrate the overall characteristics and local characteristics of the region to generate the layer key feature matrix.

[0021] As a further solution of the present invention, the specific formula for the thickness distance is as follows:

[0022]

[0023] where T′ represents the thickness distance, x i1 represents the horizontal spatial coordinate of the pixel point in the dermis layer, y i1 represents the vertical spatial coordinate of the pixel point in the dermis layer, x b1 and y b1 respectively represent the horizontal and vertical spatial coordinates of the boundary pixel points in the epidermis layer, I i1 represents the light intensity value of the pixel point in the dermis layer, I b1 represents the light intensity value of the boundary pixel points in the epidermis layer, g represents the number of adjacent pixel points, x j1 and y j1 respectively represent the horizontal and vertical spatial coordinates of the adjacent pixel points.

[0024] As a further solution of the present invention, based on the hierarchical key feature matrix, analyze the interactive dynamic characteristics of sebum gland secretion value, hair follicle spacing value, and epidermal layer thickness value, calculate the active change rate of the sebum gland, compare the weight of the hair follicle distribution fluctuation range, summarize the parameter influence relationship of the interactive result, and perform verification analysis and processing on the parameter collaboration to generate the specific steps of the multi-parameter collaborative optimization result as follows:

[0025] S301: Based on the hierarchical key feature matrix, extract the time series data of sebum gland secretion value, hair follicle spacing value, and epidermal layer thickness value, divide the data area according to the time series, calculate their change rates, analyze the dynamic interaction characteristics between the sebum gland secretion value and the hair follicle spacing value, and generate an interactive dynamic characteristic matrix;

[0026] S302: Based on the interactive dynamic characteristic matrix, calculate the active change rate of the sebum gland, count the fluctuation range of the hair follicle spacing value, analyze its distribution weight, summarize the influence characteristics between the sebum gland secretion value and the hair follicle spacing value, and combine the change data of the epidermal layer thickness value to analyze the collaborative dynamic characteristics of the three, and generate a parameter collaborative characteristic matrix;

[0027] S303: Based on the parameter collaborative characteristic matrix, analyze the collaboration of sebum gland, hair follicle, and epidermal layer thickness in the interactive result, extract the key parameter group, calculate the collaborative optimization value, perform verification analysis and dynamic adjustment on the extracted parameter group, summarize the parameter collaboration relationship, and generate a multi-parameter collaborative optimization result.

[0028] As a further solution of the present invention, the specific formula for the distribution weight value is as follows:

[0029]

[0030] Among them, W′ represents the distribution weight value, D i2 represents the i-th value in the time series data of the sebum gland secretion value, represents the mean value of the time series data of the sebum gland secretion value, M i2 represents the i-th value in the time series data of the hair follicle spacing value, n 1 represents the number of time series data points of the sebum gland secretion value and the hair follicle spacing value, T j2 represents the j-th value in the change data of the epidermal layer thickness value, represents the mean value of the change data of the epidermal layer thickness value, m 1 represents the number of data points in the change data of the epidermal layer thickness value.

[0031] As a further solution of the present invention, based on the multi-parameter collaborative optimization result, parsing the sebaceous gland activity trajectory change value, hair follicle spacing volatility, and dermis layer thickness change value in the time series, calculating the dynamic change rate and the trajectory distribution trend, and screening the characteristic fluctuation result of the time series, the specific steps for generating the scalp health dynamic trend result are as follows:

[0032] S401: Based on the multi-parameter collaborative optimization result, extract the sebaceous gland activity trajectory change value, hair follicle spacing volatility, and dermis layer thickness change value in the time series, segment the data according to the time axis, calculate the dynamic change rate, analyze the characteristics of the parameter trajectory distribution in the time series, and generate a dynamic change rate matrix;

[0033] S402: Based on the dynamic change rate matrix, extract the change rate fluctuation value in the time series for zoning processing, calculate the fluctuation characteristics, make a regional judgment on the abnormality and stability of the fluctuation value, screen the time interval with significant change rate fluctuation, and associate the distribution trend to generate a time series characteristic fluctuation matrix;

[0034] S403: Based on the time series characteristic fluctuation matrix, analyze the dynamic collaborative change characteristics of the sebaceous gland, hair follicle spacing, and dermis layer thickness within the time period, extract the relevant time series parameters for combination, calculate the trajectory distribution trend, and summarize the dynamic interaction characteristics between the parameters to generate the scalp health dynamic trend result.

[0035] As a further solution of the present invention, based on the scalp health dynamic trend result, parsing the abnormal fluctuation value of the sebaceous gland area, the abnormal characteristic distribution value of the hair follicle spacing, and the dermis layer thickness fluctuation value, calculating the difference value of the zonal characteristic parameters, marking the offset characteristic distribution area, and the specific steps for generating the scalp health zonal evaluation value are as follows:

[0036] S501: Based on the scalp health dynamic trend result, extract the time series data of the sebaceous gland area, screen the abnormal fluctuation value, analyze the spatial position and time interval of the fluctuation value, combine the hair follicle spacing distribution value and the dermis layer thickness change value, calculate the fluctuation range, and generate a characteristic fluctuation distribution matrix;

[0037] S502: Based on the characteristic fluctuation distribution matrix, statistically analyze the sebaceous gland fluctuation value, hair follicle spacing value, and dermis layer thickness value within the zone, calculate the characteristic parameter difference, compare the parameter offset trend between zones, mark the distribution area with offset characteristics and its corresponding position, and generate a zonal characteristic offset matrix;

[0038] S503: Based on the zonal characteristic offset matrix, analyze the health characteristic differences within the differential zones, combine the zonal characteristic offset trend, calculate the zonal health evaluation value, integrate the zonal data to generate an evaluation matrix, and re-divide the spatial position to generate the scalp health zonal evaluation value.

[0039] A scalp detection system based on deep learning, comprising:

[0040] The image layering module extracts the light intensity value distribution and spatial position value of image pixel points based on scalp image data, analyzes the regional feature boundaries of the epidermis layer, dermis layer and hair follicle layer, calculates the distribution range between pixels within the region, and performs zoning processing on the overall region to generate a scalp layering segmentation result;

[0041] The feature extraction module extracts the distribution area value of sebaceous glands in the epidermis layer based on the scalp layering segmentation result, screens the distribution relationship data of hair follicle spacing and hair follicle density, and analyzes the local and overall characteristics of the layering data to generate a layering key feature matrix;

[0042] The interactive verification module analyzes the interactive dynamic characteristics of sebaceous gland secretion values, hair follicle spacing values and epidermis layer thickness values based on the layering key feature matrix, calculates the active change rate of sebaceous glands, and compares the weight of the hair follicle distribution fluctuation range to perform verification analysis processing on parameter coordination to generate a multi-parameter coordination optimization result;

[0043] The dynamic trend module analyzes the change values of the active trajectory of sebaceous glands, the volatility of hair follicle spacing and the change value of the dermis layer thickness in the time series based on the multi-parameter coordination optimization result, screens the characteristic fluctuation results of the time series, and generates a scalp health dynamic trend result;

[0044] The health assessment module analyzes the abnormal fluctuation value of the sebaceous gland area, the abnormal characteristic distribution value of hair follicle spacing and the fluctuation value of the dermis layer thickness based on the scalp health dynamic trend result, calculates the difference value of the zonal characteristic parameters, and generates a scalp health zonal assessment value.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0046] In the present invention, through the combined analysis of the light intensity value and spatial position in scalp image data, the zonal division of the layering is refined. By quantifying the distribution area of sebaceous glands, the relationship between hair follicle spacing and density, and the characteristics of the dermis layer thickness, the interactive characteristics of sebaceous gland secretion values, hair follicle spacing and thickness values are dynamically analyzed, revealing the fluctuation trend of the health state in the time series, improving the accuracy of health state analysis, enhancing the efficiency and comprehensiveness of abnormal positioning, and providing more scientific support for medical diagnosis and beauty care. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 It is a schematic diagram of the step flow of the present invention;

[0049] Figure 2 It is a step flow chart of S1 of the present invention;

[0050] Figure 3 It is a step flow chart of S2 of the present invention;

[0051] Figure 4 It is a step flow chart of S3 of the present invention;

[0052] Figure 5 It is a step flow chart of S4 of the present invention;

[0053] Figure 6 It is a step flow chart of S5 of the present invention;

[0054] Figure 7 It is a system module diagram of the present invention. Detailed implementation manners

[0055] The following will describe the technical solutions in the present invention with reference to the accompanying drawings.

[0056] In the embodiments of the present invention, words such as "exemplarily", "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of the word "example" aims to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two can be selected.

[0057] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0058] In the embodiments of the present invention, sometimes subscripts such as W 1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0059] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0060] Please refer to Figure 1, a scalp detection method based on deep learning, comprising the following steps:

[0061] S1: Based on scalp image data, extract the light intensity value distribution and spatial position value of image pixel points, analyze the regional feature boundaries of the epidermis, dermis, and hair follicle layers, calculate the distribution range between pixels within the region, classify pixel points to determine their hierarchical attribution, and perform zoning processing on the overall region to generate a scalp layered segmentation result;

[0062] S2: Based on the scalp layered segmentation result, extract the distribution area value of sebaceous glands in the epidermis, fit the curve characteristics of the dermis thickness, screen the distribution relationship data of hair follicle spacing and hair follicle density, and analyze the local and overall characteristics of the layered data to generate a layered key feature matrix;

[0063] S3: Based on the layered key feature matrix, analyze the interactive dynamic characteristics of sebaceous gland secretion values, hair follicle spacing values, and epidermis thickness values, calculate the active change rate of sebaceous glands, compare the weight of the hair follicle distribution fluctuation range, summarize the parameter influence relationship of the interactive results, and perform verification analysis processing on parameter collaboration to generate a multi-parameter collaborative optimization result;

[0064] S4: Based on the multi-parameter collaborative optimization result, analyze the sebaceous gland active trajectory change value, hair follicle spacing volatility, and dermis thickness change value in the time series, calculate the dynamic change rate and trajectory distribution trend, screen the characteristic fluctuation results of the time series, and generate a scalp health dynamic trend result;

[0065] S5: Based on the scalp health dynamic trend result, analyze the abnormal fluctuation value of the sebaceous gland region, the abnormal characteristic distribution value of hair follicle spacing, and the dermis thickness fluctuation value, calculate the difference value of the zoning characteristic parameters, mark the offset characteristic distribution region, and generate a scalp health zoning evaluation value.

[0066] The scalp layered segmentation results include the boundary of the epidermal layer region features, the boundary of the dermal layer region features, the boundary of the hair follicle layer region features, the pixel point distribution range, the pixel point hierarchical attribution, and the overall area partition. The layered key feature matrix includes the sebaceous gland distribution area value in the epidermal layer, the characteristics of the dermal layer thickness curve, the distribution relationship of hair follicle spacing, the distribution relationship of hair follicle density, the local characteristics of the layered data, the overall characteristics of the layered data, and the health parameters. The multi-parameter collaborative optimization results include the interactive dynamic characteristics of the sebaceous gland secretion value, the interactive dynamic characteristics of the hair follicle spacing value, the interactive dynamic characteristics of the epidermal layer thickness value, the active change rate of the sebaceous gland, the weight of the hair follicle distribution fluctuation range, the parameter influence relationship, and the collaborative optimization verification result. The scalp health dynamic trend results include the active trajectory change value of the sebaceous gland, the hair follicle spacing volatility, the dermal layer thickness change value, the dynamic change rate, the trajectory distribution trend, and the time series characteristic fluctuation result. The scalp health partition evaluation value includes the abnormal fluctuation value of the sebaceous gland area, the abnormal characteristic distribution value of the hair follicle spacing, the dermal layer thickness fluctuation value, the partition characteristic parameter difference value, the offset characteristic distribution area, and the regional characteristic matrix.

[0067] Please refer to Figure 2 , the specific steps of S1 are as follows:

[0068] S101: Based on the scalp image data, extract the light intensity value of the image pixel points and their spatial positions in the image, establish the pixel feature mapping of the light intensity value and the spatial position, calculate the distribution density of the pixel points in the two-dimensional space, use the image coordinates to locate their positions in the image data, store the light intensity value and the position data in correspondence as a data table, and generate pixel feature data;

[0069] Regarding the light intensity value of the image pixel points and their spatial positions in the image, first extract the light intensity value of each image pixel point one by one through image segmentation technology, record the two-dimensional spatial coordinates of each pixel point, calculate the weighted density of the spatial distribution of the light intensity value, construct a spatial distribution mapping using the image coordinate data, further use region search and clustering methods to divide the spatial regions of different pixel points, combine the spatial distribution characteristics of these regions, and screen out representative spatial points according to the frequency distribution pattern of the light intensity value, establish the pixel feature mapping of the light intensity value and the spatial position, and at the same time match and store the light intensity value data with the spatial coordinate points into a table data structure, and finally generate pixel feature data.

[0070] S102: Based on the pixel feature data, combine the spatial distribution density to evaluate the aggregation of pixel points, use the spatial adjacency characteristics to identify the pixel distribution regions of the epidermal layer, dermal layer and hair follicle layer, calculate the regional boundary distribution range through the pixel point position change trend, and screen the layered boundary lines according to the light intensity value fluctuation characteristics combined with the regional coordinate points to generate a regional boundary mapping;

[0071] First, based on the fluctuation characteristics of the light intensity value and the spatial distribution density of the pixel points, where the spatial distribution density is estimated by calculating the number of other pixel points around each pixel point, a high-density area may indicate the aggregation of the cortical structure area. Then, according to the spatial adjacency characteristics of the pixel points, the clustering analysis method is used to identify the pixel point distribution areas belonging to the epidermis, dermis, and hair follicle layers. This identification process includes classifying the pixel points and dividing them into different layers. Subsequently, by analyzing the changing trend of the pixel point positions, the distribution range of the regional boundaries of these layers is calculated. The fluctuation analysis of the light intensity value helps to determine the stratification boundary lines, which are defined by the mutation points of the light intensity value. The mutation points mark the boundaries between different layers. Based on these data, a regional boundary map is generated, which details the spatial positioning of different cortical structures.

[0072] S103: Based on the regional boundary map, group the pixel points in the region layer by layer, calculate the change range of the light intensity value in the region, and determine the belonging layer. According to the layer belonging, re-divide the spatial distribution of the pixel points into three independent regions: the epidermis layer, the dermis layer, and the hair follicle layer, and generate the scalp stratification segmentation result;

[0073] First, extract the light intensity value distribution data of the pixel points corresponding to the boundary lines and label them according to the stratification belonging. Combining the changing trend of the light intensity value of each layer, divide the light intensity value in the region by layer. The change of the light intensity value extracts its peak distribution and frequency characteristics through the spectral analysis method. Use these characteristics to determine the layer and label the belonging regions of the pixel points as the epidermis layer, the dermis layer, and the hair follicle layer. Finally, re-divide the image spatial distribution region according to the labeling result and update the boundary segmentation data to generate the final scalp stratification segmentation result.

[0074] Please refer to Figure 3 , the specific steps of S2 are as follows:

[0075] S201: Based on the scalp stratification segmentation result, locate the pixel points of the epidermis layer, screen the pixel points associated with the sebaceous glands, analyze the distribution of the sebaceous gland pixels using the light intensity value and the spatial position parameters, calculate the total area value within the sebaceous gland region, associate the sebaceous gland area with the corresponding spatial coordinates, and store them in matrix format to generate the sebaceous gland distribution matrix;

[0076] First, accurately locate the pixel points of the epidermis layer, and screen out the pixel points associated with sebaceous glands by setting the range of light intensity values. For the screened pixel points, extract their light intensity values and corresponding spatial position parameters, analyze the distribution law of the sebaceous gland area using the statistical characteristics of the light intensity value distribution, calculate the total area value by fitting the spatial coordinate data of the sebaceous gland pixel points into a polygon area, and at the same time correct the calculation error of the area boundary using the boundary coordinate information of the pixel points during the fitting process. Associate the finally generated sebaceous gland area with its corresponding spatial coordinates, and archive the sebaceous gland distribution information in the form of a storage matrix to finally generate a sebaceous gland distribution matrix.

[0077] S202: Based on the sebaceous gland distribution matrix, extract the spatial position data of the dermal layer pixel points, combine with the regional boundary coordinates of the epidermis layer, calculate the thickness distance, fit the thickness change curve of the dermal layer point by point, correct the spatial position deviation, and generate the curve continuity result to generate a dermal layer thickness curve matrix;

[0078] The specific formula for the thickness distance is:

[0079]

[0080] Among them, T′ represents the thickness distance, x i1 represents the horizontal spatial coordinate of the dermal layer pixel point, y i1 represents the vertical spatial coordinate of the dermal layer pixel point, x b1 and y b1 respectively represent the horizontal and vertical spatial coordinates of the epidermis layer boundary pixel points, I i1 represents the light intensity value of the dermal layer pixel point, I b1 represents the light intensity value of the epidermis layer boundary pixel point, g represents the number of adjacent pixel points, x j1 and y j1 respectively represent the horizontal and vertical spatial coordinates of the adjacent pixel points.

[0081] Set the actual monitoring parameters as follows:

[0082] The horizontal spatial coordinate x of the dermal layer pixel point i1 = 150, the vertical spatial coordinate y i1 = 100, the horizontal spatial coordinate x of the epidermis layer boundary pixel point b1 = 140, the vertical spatial coordinate y b1 = 90, the light intensity value I of the dermal layer pixel point i1 = 85, the light intensity value I of the epidermis layer boundary pixel point b1 = 100, the number of adjacent pixel points g = 4, and the spatial coordinates of the adjacent pixel points are (x j1 , y j1) = (145, 95), (142, 93), (138, 87), (150, 92).

[0083] (x i1 - x b1 ) 2 +(y i1 - y b1 ) 2 : This calculates the square of the Euclidean distance between points i1 and b1 in two-dimensional space. x i1 , y i1 and x b1 , y b1 are the coordinates of two points obtained through a geographic information system or a coordinate measuring device.

[0084] This represents the absolute value of the relative length change, where l i1 and l b1 can be the length of a line segment or other measurement criteria, obtained through actual measurement.

[0085] This is an adjustment factor, where g is the total number of points considered (such as points in a population). This part takes into account the influence of the sum of the squared distances from point b1 to all other points, whose coordinates are obtained through similar measurement techniques.

[0086] Numerical example:

[0087] Assume the specific coordinate data is: x i1 = 10, y i1 = 20, x b1 = 5, y b1 = 15, the length l i1 = 4, l b1 = 2, and there are 2 additional points g = 2, with coordinates x j1 = [3, 4], y j1 = [6, 7].

[0088] Calculation process: Square of the Euclidean distance: (10 - 5) 2 +(20 - 15) 2 = 25 + 25 = 50 Absolute value of the length change: Calculation of the adjustment factor: For j = 1: (3 - 5) 2 +(6 - 15) 2 = 4 + 81 = 85 For j = 2: (4 - 5) 2 +(7 - 15) 2 = 1 + 64 = 65 Factor value:

[0089]

[0090] Combine the above results to calculate the final T′:

[0091]

[0092] Result interpretation:

[0093] The calculated weighted distance T′ = 0.608 indicates that after considering the length change and the distance weighting of surrounding points, the actual perceived distance between two points decreases, which may represent the actual interaction distance or the distance of mutual influence between points. This result can be used for further analysis of the relative position or connection strength of the two points in the spatial relationship.

[0094] S203: Based on the dermal layer thickness curve matrix, extract the spatial positions of the pixel points in the hair follicle area, calculate the distances between adjacent hair follicles, combine the distribution of hair follicles in the area, calculate the hair follicle density, screen the eigenvalues of the relationship between the spacing change and the density distribution, and integrate the overall characteristics and local characteristics of the area to generate a hierarchical key feature matrix;

[0095] For the pixel points in the hair follicle area, first extract their spatial coordinate data, calculate the distances between adjacent hair follicles using the spatial position relationship between the coordinate points, calculate the overall density of the hair follicle area by statistically analyzing the distance data between each pair of hair follicles and using the density estimation formula, combine the distribution characteristics of hair follicles in the area, further calculate the fluctuation range of the spacing between adjacent hair follicles and its impact on the hair follicle density, extract the spacing change characteristics as eigenvalues, integrate the local and overall spatial distribution laws of hair follicles in the area, and finally generate result data for subsequent analysis by constructing a hierarchical key feature matrix containing local and overall characteristics.

[0096] Please refer to Figure 4 , and the specific steps of S3 are as follows:

[0097] S301: Based on the hierarchical key feature matrix, extract the time series data of the sebaceous gland secretion value, hair follicle spacing value, and epidermis layer thickness value, divide the data area according to the time series, and calculate their change rates, analyze the dynamic interaction characteristics between the sebaceous gland secretion value and the hair follicle spacing value, and generate an interaction dynamic characteristic matrix;

[0098] Use the sliding window method to divide the time series into multiple continuous data segments according to the time range, calculate the change rate within each data segment, specifically use the time difference method to obtain the change rate of each time period, construct the time series dynamic change models for the sebaceous gland secretion value and the hair follicle spacing value respectively, calculate the time series coupling degree of the two during the dynamic change interaction, further use dynamic interaction analysis to fit the change characteristics of the two, store the characteristic values of the dynamic interaction of the two in a matrix and form a time dynamic interaction characteristic matrix, and finally output the interaction dynamic characteristic matrix of the sebaceous gland and the hair follicle spacing.

[0099] S302: Calculate the active change rate of the sebaceous glands based on the interaction dynamic characteristic matrix, statistically analyze the fluctuation range of the hair follicle spacing values, and analyze their distribution weights. Summarize the influence characteristics between the sebaceous gland secretion values and the hair follicle spacing values, and combine the change data of the epidermal layer thickness values to analyze the collaborative dynamic characteristics of the three, and generate a parameter collaborative characteristic matrix;

[0100] The specific calculation formula for the distribution weight value is:

[0101]

[0102] Among them, W′ represents the distribution weight value, D i2 represents the i-th value in the time series data of the sebaceous gland secretion value, represents the mean value of the time series data of the sebaceous gland secretion value, M i2 represents the i-th value in the time series data of the hair follicle spacing value, n 1 represents the number of time series data points of the sebaceous gland secretion value and the hair follicle spacing value, T j2 represents the j-th value in the change data of the epidermal layer thickness value, represents the mean value of the change data of the epidermal layer thickness value, m 1 represents the number of data points of the change data of the epidermal layer thickness value.

[0103] Set specific monitoring parameters

[0104] The time series data of the sebaceous gland secretion value is D i2 = [85, 90, 80, 95, 88] units, and the number of data points n 1 = 5, and its mean value is

[0105]

[0106] The time series data of the hair follicle spacing value is M i2 = [2.1, 2.3, 2.0, 2.4, 2.2] mm, corresponding to each time point.

[0107] The change data of the epidermal layer thickness value is T j2 = [1.8, 1.9, 2.0, 2.2, 1.7] mm, and the number of data points m 1 = 5, and its mean value is

[0108]

[0109] Calculate the numerator part of the weighted sum of the deviation term of the sebaceous gland secretion value and the hair follicle spacing value according to the formula for calculating the numerator term in the formula:

[0110]

[0111] For each item, calculate:

[0112] When i = 1:

[0113]

[0114] When i = 2:

[0115]

[0116] When i = 3:

[0117]

[0118] When i = 4:

[0119]

[0120] When i = 5:

[0121]

[0122] The sum obtained is:

[0123]

[0124] Calculate the square root of the variance of the epidermal layer thickness value, which is the sum of squares in the denominator term calculated according to the formula:

[0125]

[0126] For each item, calculate:

[0127] When j = 1:

[0128]

[0129] When j = 2:

[0130]

[0131] When j = 3:

[0132]

[0133] When j = 4:

[0134]

[0135] When j = 5:

[0136]

[0137] The sum obtained is:

[0138]

[0139] Square root calculation:

[0140]

[0141] Comprehensively calculate the distribution weight value W′

[0142] Substitute into the formula for calculation:

[0143]

[0144] Result description:

[0145] The calculated distribution weight value W′ is 23.29. This value indicates the comprehensive influence intensity of the time - series deviation between the sebaceous gland secretion value and the hair follicle spacing value combined with the change in the epidermal layer thickness. This result is used to summarize the dynamic relationship between the sebaceous gland secretion value and the hair follicle spacing value and analyze the cooperative dynamic characteristics of the three, and finally generate a parameter cooperative characteristic matrix.

[0146] S303: Based on the parameter cooperative characteristic matrix, analyze the cooperation of the sebaceous gland, hair follicle, and epidermal layer thickness in the interaction result, extract the key parameter group, calculate the cooperative optimization value, conduct verification analysis and dynamic adjustment on the extracted parameter group, summarize the parameter cooperative relationship, and generate a multi - parameter cooperative optimization result;

[0147] Extract the key parameter group and perform dynamic adjustment on it. First, conduct statistical analysis on the dynamic change curves of the sebaceous gland secretion value and the hair follicle spacing value, extract key parameter points based on the inflection point characteristics of the change curves, generate the cooperative optimization value of the key parameters through a fitting model, verify the calculation results of the optimization value using the multi - parameter comparison analysis method, combine the dynamic analysis of the epidermal layer thickness change range, calculate the influence characteristics of the optimization result on the cooperation of the three, and summarize the final result as a multi - parameter cooperative optimization result and generate corresponding data.

[0148] Please refer to Figure 5 , the specific steps of S4 are as follows:

[0149] S401: Based on the multi - parameter cooperative optimization result, extract the sebaceous gland active trajectory change value, hair follicle spacing volatility, and dermal layer thickness change value in the time series, segment the data according to the time axis, calculate the dynamic change rate, analyze the characteristics of the parameter trajectory distribution in the time series, and generate a dynamic change rate matrix;

[0150] The data is segmented according to the time axis. During the segmentation process, the data is normalized according to the continuity of the time series. The time difference method is used to calculate the dynamic change rate of each segment of data. Specifically, the change rate is obtained by calculating the ratio of the change value between adjacent time points to the reference value. The change rates of the parameters in the matrix are analyzed by normalization, and the time series distribution characteristics of the parameters are analyzed through the normalized matrix. Finally, a dynamic change rate matrix is generated for subsequent characteristic analysis.

[0151] S402: Based on the dynamic change rate matrix, extract the fluctuation values of the change rate in the time series for zoning processing, and calculate the fluctuation characteristics. Make a regional judgment on the abnormality and stability of the fluctuation values, screen the time intervals with significant fluctuations in the change rate, and associate the distribution trend to generate a time series characteristic fluctuation matrix;

[0152] Based on the dynamic change rate matrix, according to the formula

[0153]

[0154] Calculate the change rate fluctuation characteristics, the standard deviation of the fluctuation value, and the dynamic synergy.

[0155] In the formula, F w represents the change rate of the time series, ΔR is the data change amount in the adjacent time period, R 0 is the reference value, and T is the time interval; Z a is the standard deviation of the fluctuation value, x i is the fluctuation value sample, μ is the mean value of the fluctuation value, and n is the number of samples; P b is the synergy integral, t 1 , t 2 is the time range.

[0156] The change rate of the time series is calculated by the ratio of the change value to the time interval. Set the reference value R 0 = 50 units, the time interval T = 10 seconds, and the change amount ΔR = 5 units. It is calculated that

[0157]

[0158] The standard deviation of the fluctuation value is calculated by the variance of the sample. Let the number of samples n = 5, and the sample values x i are 1, 2, 3, 4, 5, and the mean value μ = 3. It is calculated that

[0159]

[0160] The dynamic synergy is calculated by the product integral of the change rate and the standard deviation. Let the time range be from t 1 = 0 seconds to t 2 = 10 seconds, and calculate the integral

[0161]

[0162] This result indicates that there is a certain synergy relationship between the time series change rate and the fluctuation characteristics, which can be further used to screen the time intervals with significant fluctuations in the change rate and correlate the distribution trends.

[0163] S403: Based on the time series characteristic fluctuation matrix, analyze the dynamic synergy change characteristics of the sebaceous gland, hair follicle spacing, and dermal layer thickness within the time period, extract relevant time series parameters for combination, calculate the trajectory distribution trend, summarize the dynamic interaction characteristics between the parameters, and generate the scalp health dynamic trend result;

[0164] Regarding the dynamic synergy change characteristics of the sebaceous gland active trajectory, hair follicle spacing, and dermal layer thickness within the time period, first extract the relevant parameter combinations in the time series, conduct dynamic trajectory analysis on these parameter combinations, calculate the trajectory distribution trend of the parameters within each time period, obtain the dynamic trend characteristics by calculating the slope and trajectory shape characteristics of the parameter changes within each time period, summarize the temporal distribution of the parameter changes, integrate the summarized time series dynamic interaction characteristics into the scalp health dynamic trend result, and perform the final analysis and output based on the change characteristics of the parameter synergy.

[0165] Please refer to Figure 6 , the specific steps of S5 are as follows:

[0166] S501: Based on the scalp health dynamic trend result, extract the time series data of the sebaceous gland area, screen the abnormal fluctuation values, analyze the spatial position and time interval of the fluctuation values, combine the hair follicle spacing distribution value and the dermal layer thickness change value, calculate the fluctuation range, and generate the characteristic fluctuation distribution matrix;

[0167] The screening process uses the time series segmentation comparison method, calculates the deviation amount between the sebaceous gland secretion value and the standard distribution within each time period, and marks the part where the deviation amount exceeds the normal range as the abnormal fluctuation value. Analyze the distribution range of the abnormal fluctuation value through the corresponding relationship between the spatial position coordinates and the time interval, combine the relevant characteristics of the hair follicle spacing distribution value and the dermal layer thickness change value, calculate the sebaceous gland fluctuation range, construct the characteristic fluctuation distribution matrix by statistically analyzing the extreme values, mean values, and standard deviations of the fluctuation range, and store the matrix data for subsequent analysis.

[0168] S502: Based on the characteristic fluctuation distribution matrix, statistically analyze the sebaceous gland fluctuation value, hair follicle spacing value, and dermal layer thickness value within the partition, calculate the characteristic parameter differences, compare the parameter offset trends between partitions, mark the distribution areas with offset characteristics and their corresponding positions, and generate the partition characteristic offset matrix;

[0169] Based on the characteristic fluctuation distribution matrix, according to the formula

[0170]

[0171] Calculate the parameter differences, fluctuation characteristics, and dynamic offset trends within the partitions.

[0172] In the formula, D v represents the parameter difference between partitions, P i and P j are the parameter values within two partitions respectively, Q c represents the standard deviation of the partition fluctuation characteristics, d k is the sample of the fluctuation value, μ is the sample mean, n is the number of samples, R o is the integral of the dynamic offset trend, t a and t b are the integral time ranges.

[0173] The partition parameter difference is calculated through the relative difference of the parameter values between partitions. Suppose the parameter value P i in partition 1 is 12 units, and the parameter value P j in partition 2 is 10 units. The calculation gives

[0174]

[0175] The standard deviation of the partition fluctuation characteristics is calculated through the variance of the fluctuation value samples. Suppose the number of samples n = 5, and the fluctuation sample values are 2, 3, 4, 5, 6, and the sample mean μ = 4. The calculation gives

[0176]

[0177] The dynamic offset trend is calculated through the integral of the product of the difference value and the fluctuation characteristics. Suppose the time range is from t a = 0 seconds to t b = 10 seconds. Calculate the integral

[0178]

[0179] This result shows that there are certain fluctuation characteristics in the dynamic offset trend of the partition parameters, which can be used to mark the distribution areas and their corresponding positions of the offset characteristics.

[0180] S503: Based on the partition characteristic offset matrix, analyze the health characteristic differences within the differentiated partitions, combine the partition characteristic offset trends, calculate the partition health assessment values, integrate the partition data to generate an assessment matrix, and re-divide the spatial positions to generate the scalp health partition assessment values;

[0181] Perform characteristic statistics on the parameters of the differentiated areas, extract the dynamic change characteristics of the sebaceous gland activity value, hair follicle spacing, and dermal layer thickness within the partition, calculate the health assessment value of each partition by combining the partition characteristic offset trend, dynamically summarize the data within the partition through the comparison and analysis of the assessment value and the partition location, re-divide the spatial location of the partition, generate the scalp health partition assessment value based on the new partition assessment data, and output the result.

[0182] Please refer to Figure 7 , a scalp detection system based on deep learning, comprising:

[0183] The image layering module extracts the light intensity value distribution and spatial position value of the image pixel points based on the scalp image data, analyzes the regional feature boundaries of the epidermis, dermis, and hair follicle layers, calculates the distribution range between pixels within the region, and performs partition processing on the overall region to generate the scalp layering segmentation result;

[0184] The feature extraction module extracts the sebaceous gland distribution area value of the epidermis based on the scalp layering segmentation result, screens the distribution relationship data of the hair follicle spacing and hair follicle density, and analyzes the characteristics of the local and overall layering data to generate the layering key feature matrix;

[0185] The interactive verification module analyzes the interactive dynamic characteristics of the sebaceous gland secretion value, hair follicle spacing value, and epidermis layer thickness value based on the layering key feature matrix, calculates the sebaceous gland activity change rate, and compares the weight of the hair follicle distribution fluctuation range to perform verification analysis and processing on the parameter collaboration to generate the multi-parameter collaboration optimization result;

[0186] The dynamic trend module analyzes the sebaceous gland activity trajectory change value, hair follicle spacing volatility, and dermal layer thickness change value in the time series based on the multi-parameter collaboration optimization result, screens the characteristic fluctuation results of the time series, and generates the scalp health dynamic trend result;

[0187] The health assessment module analyzes the abnormal fluctuation value of the sebaceous gland area, the abnormal characteristic distribution value of the hair follicle spacing, and the dermal layer thickness fluctuation value based on the scalp health dynamic trend result, calculates the difference value of the partition characteristic parameters, and generates the scalp health partition assessment value.

[0188] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claimed rights.

Claims

1. A scalp detection method based on deep learning, characterized in that: The following steps are involved: S1: Based on the scalp image data, the light intensity distribution and spatial position values ​​of the image pixels are extracted, the regional characteristic boundaries of the epidermis, dermis and hair follicle layer are analyzed, the distribution range of pixels in the region is calculated, the pixels are classified to determine their layer affiliation, and the overall region is partitioned to generate the scalp layer segmentation result; S2: Based on the scalp stratification segmentation results, extract the distribution area value of the epidermis sebaceous glands, fit the dermis thickness curve characteristics, filter the distribution relationship data of the hair follicle spacing and the hair follicle density, and analyze the local and overall characteristics of the stratified data to generate a stratified key feature matrix; S3: Based on the hierarchical key feature matrix, the interactive dynamic characteristics of the sebaceous gland secretion value, the hair follicle spacing value and the epidermal thickness value are analyzed, the sebaceous gland activity change rate is calculated, and the weight of the hair follicle distribution fluctuation range is compared, the parameter influence relationship of the interactive result is summarized, and the parameter collaboration is verified and analyzed to generate a multi-parameter collaborative optimization result; S4: Based on the multi-parameter collaborative optimization results, the sebaceous gland active trajectory change value, the hair follicle spacing fluctuation rate and the dermis thickness change value in the time series are analyzed, the dynamic change rate and trajectory distribution trend are calculated, the characteristic fluctuation results of the time series are screened, and the scalp health dynamic trend results are generated; S5: Based on the scalp health dynamic trend results, analyze the abnormal fluctuation value of the sebaceous gland area, the abnormal characteristic distribution value of the hair follicle spacing and the dermis thickness fluctuation value, calculate the difference value of the partition characteristic parameters, mark the offset characteristic distribution area, and generate the scalp health partition evaluation value.

2. The scalp detection method based on deep learning according to claim 1, characterized in that: The scalp stratified segmentation result includes the characteristic boundary of the epidermis region, the characteristic boundary of the dermis region, the characteristic boundary of the hair follicle region, the pixel distribution range, the pixel hierarchical attribution, and the overall regional partition. The stratified key feature matrix includes the epidermis sebaceous gland distribution area value, the dermis thickness curve characteristics, the follicle spacing distribution relationship, the follicle density distribution relationship, the stratified data local characteristics, the stratified data overall characteristics, and the health parameters. The multi-parameter collaborative optimization result includes the interactive dynamic characteristics of the sebaceous gland secretion value, the interactive dynamic characteristics of the follicle spacing value, the interactive dynamic characteristics of the epidermal thickness value, the sebaceous gland activity change rate, the hair follicle distribution fluctuation range weight, the parameter influence relationship, and the collaborative optimization verification result. The scalp health dynamic trend result includes the sebaceous gland activity trajectory change value, the hair follicle spacing fluctuation rate, the dermis thickness change value, the dynamic change rate, the trajectory distribution trend, and the time series characteristic fluctuation result. The scalp health partition evaluation value includes the sebaceous gland area abnormal fluctuation value, the hair follicle spacing abnormal characteristic distribution value, the dermis thickness fluctuation value, the partition characteristic parameter difference value, the offset characteristic distribution area, and the regional characteristic matrix.

3. The scalp detection method based on deep learning according to claim 1, characterized in that: Based on the scalp image data, the intensity distribution and spatial position values ​​of the image pixels are extracted, the regional characteristic boundaries of the epidermis, dermis and hair follicle layer are analyzed, the distribution range of pixels in the region is calculated, the pixels are classified to determine their layer affiliation, and the overall region is partitioned. The specific steps to generate the scalp layer segmentation results are as follows: S101: based on the scalp image data, extract the light intensity value of the image pixel point and its spatial position in the image, establish a pixel feature mapping of the light intensity value and the spatial position, calculate the distribution density of the pixel point in the two-dimensional space, locate its position in the image data using the image coordinates, store the light intensity value and the position data in a data table, and generate pixel feature data; S102: Based on the pixel feature data, the aggregation of the pixels is evaluated in combination with the spatial distribution density, and the pixel distribution areas of the epidermis, dermis and hair follicle layer are identified by using the spatial adjacency characteristics. The distribution range of the regional boundary is calculated according to the change trend of the pixel position, and the layered boundary lines are screened according to the light intensity fluctuation characteristics combined with the regional coordinate points to generate the regional boundary map; S103: Based on the region boundary mapping, the pixels in the region are grouped layer by layer, the range of light intensity values ​​in the region is calculated, and the layer to which they belong is determined. The spatial distribution of the pixels is re-divided into three independent regions, namely the epidermis, dermis and hair follicle layer, according to the layer affiliation, to generate a scalp layer segmentation result.

4. The scalp detection method based on deep learning according to claim 1, characterized in that: Based on the scalp stratification segmentation results, the distribution area value of the sebaceous glands in the epidermis is extracted, the thickness curve characteristics of the dermis are fitted, the distribution relationship data of the hair follicle spacing and the hair follicle density are screened, and the local and overall characteristics of the stratified data are analyzed. The specific steps of generating the stratified key feature matrix are as follows: S201: Based on the scalp layer segmentation result, locate the epidermal pixel points, and select the sebaceous gland-related pixels, use the light intensity value and the spatial position parameter to analyze the distribution of the sebaceous gland pixels, calculate the total area value in the sebaceous gland area, associate the sebaceous gland area with the corresponding spatial coordinates, and store them in a matrix format to generate a sebaceous gland distribution matrix; S202: Based on the sebaceous gland distribution matrix, extract the spatial position data of the dermis pixel points, combine the regional boundary coordinates of the epidermis, calculate the thickness distance, fit the dermis thickness change curve point by point, correct the spatial position deviation, and generate the curve continuity result to generate the dermis thickness curve matrix; S203: Based on the dermis thickness curve matrix, the spatial position of the pixel points in the hair follicle area is extracted, the distance between adjacent hair follicles is calculated, the hair follicle density is calculated based on the distribution of hair follicles in the area, the characteristic values ​​of the relationship between the spacing change and the density distribution are screened, the overall characteristics of the area and the local characteristics are integrated and processed to generate a hierarchical key feature matrix.

5. The scalp detection method based on deep learning according to claim 4, characterized in that: The thickness distance calculation formula is specifically: Where T′ represents the thickness distance, x i1 Represents the horizontal spatial coordinate of the pixel point in the dermis, y i1 Represents the longitudinal spatial coordinate of the pixel point in the dermis, x b1 and b1 Represent the horizontal and vertical spatial coordinates of the epidermal boundary pixel points, I i1 Represents the light intensity value of the dermis pixel, I b1 represents the light intensity value of the pixel at the boundary of the epidermis, g represents the number of adjacent pixels, and x j1 and j1 Represents the horizontal and vertical spatial coordinates of adjacent pixels respectively.

6. The scalp detection method based on deep learning according to claim 1, characterized in that: Based on the hierarchical key feature matrix, the interactive dynamic characteristics of sebaceous gland secretion value, hair follicle spacing value and epidermal thickness value are analyzed, the sebaceous gland activity change rate is calculated, and the weight of the hair follicle distribution fluctuation range is compared. The parameter influence relationship of the interactive result is summarized, and the parameter collaboration is verified and analyzed. The specific steps for generating the multi-parameter collaborative optimization result are as follows: S301: based on the hierarchical key feature matrix, extracting the time series data of sebaceous gland secretion value, hair follicle distance value and epidermal thickness value, dividing the data area according to the time series, and calculating the change rate thereof, analyzing the dynamic interaction characteristics of the sebaceous gland secretion value and the hair follicle distance value, and generating an interaction dynamic characteristic matrix; S302: Based on the interactive dynamic characteristic matrix, the sebaceous gland activity change rate is calculated, the fluctuation range of the hair follicle spacing value is counted, and its distribution weight is analyzed, the influence characteristics between the sebaceous gland secretion value and the hair follicle spacing value are summarized, and the synergistic dynamic characteristics of the three are analyzed in combination with the epidermal thickness value change data to generate a parameter synergistic characteristic matrix; S303: Based on the parameter synergy characteristic matrix, the synergy of sebaceous glands, hair follicles and epidermal thickness in the interaction results is analyzed, key parameter groups are extracted, and synergy optimization values ​​are calculated. The extracted parameter groups are verified, analyzed and dynamically adjusted, the parameter synergy relationship is summarized and processed, and a multi-parameter synergy optimization result is generated.

7. The scalp detection method based on deep learning according to claim 6, characterized in that: The distribution weight value calculation formula is specifically: Where W′ represents the distribution weight value, D i2 represents the i-th value in the time series data of sebaceous gland secretion value, D2 represents the mean of the time series data of sebaceous gland secretion value, M i2 represents the i-th value in the time series data of the hair follicle spacing value, n1 represents the number of time series data points of sebaceous gland secretion value and hair follicle spacing value, T j2 represents the jth value in the epidermal thickness change data, T2 represents the mean of the epidermal thickness change data, and m1 represents the number of epidermal thickness change data points.

8. The scalp detection method based on deep learning according to claim 1, characterized in that: Based on the multi-parameter collaborative optimization results, the specific steps of analyzing the sebaceous gland active trajectory change value, hair follicle spacing fluctuation rate and dermis thickness change value in the time series, calculating the dynamic change rate and trajectory distribution trend, screening the characteristic fluctuation results of the time series, and generating the scalp health dynamic trend results are as follows: S401: Based on the multi-parameter collaborative optimization result, extract the sebaceous gland active trajectory change value, hair follicle spacing fluctuation rate and dermis thickness change value in the time series, segment the data according to the time axis, calculate the dynamic change rate, analyze the characteristics of the parameter trajectory distribution in the time series, and generate a dynamic change rate matrix; S402: Based on the dynamic change rate matrix, extract the change rate fluctuation value in the time series for partition processing, calculate the fluctuation characteristics, make regional judgments on the abnormality and stability of the fluctuation value, screen the time interval with significant change rate fluctuation, and associate the distribution trend to generate a time series characteristic fluctuation matrix; S403: Based on the time series characteristic fluctuation matrix, the dynamic coordinated change characteristics of the sebaceous glands, the distance between the hair follicles and the thickness of the dermis within the time period are analyzed, the time series parameters with correlation are extracted and combined, and the trajectory distribution trend is calculated, the dynamic interaction characteristics between the parameters are summarized, and the scalp health dynamic trend results are generated.

9. The scalp detection method based on deep learning according to claim 1, characterized in that: Based on the scalp health dynamic trend results, the specific steps of analyzing the abnormal fluctuation value of the sebaceous gland area, the abnormal characteristic distribution value of the hair follicle spacing and the fluctuation value of the dermis thickness, calculating the difference value of the partition characteristic parameters, marking the offset characteristic distribution area, and generating the scalp health partition evaluation value are as follows: S501: Based on the scalp health dynamic trend result, extract the time series data of the sebaceous gland area, screen the abnormal fluctuation value, analyze the spatial position and time interval of the fluctuation value, combine the hair follicle spacing distribution value and the dermis thickness change value, calculate the fluctuation range, and generate a characteristic fluctuation distribution matrix; S502: Based on the characteristic fluctuation distribution matrix, statistics are collected on the sebaceous gland fluctuation value, the hair follicle spacing value and the dermis thickness value in the partition, characteristic parameter differences are calculated, parameter deviation trends between partitions are compared, distribution areas with deviation characteristics and their corresponding positions are marked, and a partition characteristic deviation matrix is ​​generated; S503: Based on the partition characteristic offset matrix, analyze the health characteristic differences within the differentiated partitions, combine the partition characteristic offset trend, calculate the partition health assessment value, integrate the partition data to generate an assessment matrix, and re-divide the spatial position to generate a scalp health partition assessment value.

10. A scalp detection system based on deep learning, characterized in that: According to a deep learning-based scalp detection method according to any one of claims 1 to 9, the system comprises: The image stratification module extracts the light intensity distribution and spatial position value of image pixels based on scalp image data, analyzes the regional characteristic boundaries of the epidermis, dermis and hair follicle layer, calculates the distribution range between pixels in the region, and partitions the entire region to generate scalp stratification segmentation results; The feature extraction module extracts the distribution area value of the sebaceous glands in the epidermis based on the scalp stratification segmentation result, selects the distribution relationship data of the hair follicle spacing and the hair follicle density, and analyzes the local and overall characteristics of the stratified data to generate a stratified key feature matrix; The interactive verification module analyzes the interactive dynamic characteristics of the sebaceous gland secretion value, the hair follicle spacing value and the epidermal thickness value based on the hierarchical key feature matrix, calculates the sebaceous gland activity change rate, and compares the weight of the hair follicle distribution fluctuation range, verifies and analyzes the parameter collaboration, and generates a multi-parameter collaborative optimization result; The dynamic trend module analyzes the sebaceous gland active trajectory change value, hair follicle spacing fluctuation rate and dermis thickness change value in the time series based on the multi-parameter collaborative optimization result, screens the characteristic fluctuation results of the time series, and generates a scalp health dynamic trend result; The health assessment module analyzes the abnormal fluctuation value of the sebaceous gland area, the abnormal characteristic distribution value of the hair follicle spacing and the fluctuation value of the dermis thickness based on the scalp health dynamic trend results, calculates the difference value of the partition characteristic parameters, and generates a scalp health partition assessment value.

Images