Skin condition prediction method

Abstract

The invention belongs to the technical field of computer application, and provides a skin condition prediction method, which comprises the following steps: collecting the real-time state of skin through a wearable device to obtain physiological data; acquiring an image of skin through skin imaging equipment to obtain image data; preprocessing the physiological data and the image data, and extracting a physiological feature vector and an image feature vector; performing data fusion on the physiological features and the image features to obtain a multi-source data feature vector; inputting the multi-source data features into a skin condition prediction model to generate a skin condition prediction result; based on a skin condition standard, evaluating the skin condition prediction result to obtain an evaluation result; according to the method, comprehensive and quantitative skin condition evaluation is provided, individual differences are combined, personalized nursing suggestions are provided, early detection of skin problems is facilitated, and targeted intervention and nursing are promoted.

Description

Technical Field

[0001] The invention belongs to the technical field of computer application, and in particular to a skin condition prediction method. Background Art

[0002] As people pay more and more attention to skin health, the monitoring and evaluation of skin conditions has gradually become a research hotspot; nowadays, women have greater interest in this aspect, and most women use cosmetics to keep their skin in the best condition.

[0003] However, some traditional technologies rely solely on skin image acquisition and analysis. These methods can only obtain visual information on the skin surface and cannot reflect physiological changes deeper in the skin. For example, image analysis has difficulty detecting real-time changes in temperature, humidity, and oil secretion inside the skin, and these physiological indicators play a key role in judging skin health. When the skin is in the early stages of allergy or inflammation, the skin's physiological indicators may have changed, but there may not be obvious symptoms in appearance. Relying solely on image data cannot detect these potential problems in a timely manner;

[0004] Many existing skin condition prediction technologies focus solely on one or a few aspects of the skin, failing to establish a comprehensive, multi-dimensional assessment system. For example, some methods determine skin dryness solely based on moisture content, ignoring oil secretion, changes in skin texture, and the impact of the external environment on the skin. In reality, skin health is the result of the interaction of multiple factors, and a single-dimensional assessment cannot accurately reflect the skin's true condition.

[0005] In some traditional skin assessment methods, doctors or professionals rely on visual observation and personal experience to determine skin condition. This approach is subject to significant subjective factors, and different evaluators may produce different results, resulting in a lack of objectivity and consistency in the assessment results. Furthermore, manual assessments struggle to quantify skin conditions, making it difficult to provide accurate data support for subsequent treatment and care.

[0006] To this end, those skilled in the art have proposed a skin condition prediction method, which aims to provide a comprehensive and quantitative skin condition assessment, combine individual differences, and provide personalized care recommendations, which helps to detect skin problems early and promote targeted intervention and care. Summary of the Invention

[0007] In order to solve the above technical problems, the present invention provides a skin condition prediction method to solve the problems raised in the background technology.

[0008] A method for predicting skin condition, comprising the following steps:

[0009] S1. Collect the real-time state of the skin through the wearable device to obtain physiological data; collect the image of the skin through the skin imaging device to obtain image data;

[0010] S2. Preprocessing the physiological data to extract physiological feature vectors;

[0011] S3, preprocessing the image data and extracting image feature vectors;

[0012] S4, fusing the physiological features and image features to obtain a multi-source data feature vector;

[0013] S5. Inputting the multi-source data features into a skin condition prediction model to generate a skin condition prediction result;

[0014] S6. Evaluate the skin condition prediction result based on the skin condition standard to obtain an evaluation result.

[0015] Preferably, the preprocessing of the physiological data and feature extraction of physiological feature vectors includes:

[0016] The physiological characteristic vector P includes skin temperature p1, skin humidity p2 and oil secretion p3;

[0017] The rate of change of skin temperature at adjacent time points is calculated using the following formula: T :

[0018] Among them, R T is the rate of change of the skin temperature, which is used to reflect the dynamic change of the skin temperature, Δt is the time interval, T n and T n-1 The temperature data at adjacent time points are used to obtain the skin temperature characteristics;

[0019] The following formula is used to calculate the change rate R of the skin moisture at adjacent time points: H :

[0020] Among them, R H is the rate of change of skin humidity, which is used to reflect the dynamic change of skin humidity. Δt is the time interval, H n and H n-1 The skin moisture data at adjacent time points are used to obtain the skin moisture characteristics;

[0021] The following formula is used to calculate the rate of change R of the oil secretion at adjacent time points: O :

[0022] Among them, R Ois the rate of change of oil secretion, which is used to reflect the dynamic change of skin oil secretion. Δt is the time interval. n and O n-1 The oil secretion data at adjacent time points are used to obtain the oil secretion characteristics.

[0023] Preferably, the preprocessing of the image data and feature extraction of image feature vectors includes:

[0024] Calculating a gray level co-occurrence matrix of the image data, wherein the gray level co-occurrence matrix is ​​used to describe the spatial relationship between pixels of different gray values ​​in the image data, wherein the gray level co-occurrence matrix is ​​expressed as: P(i, j, d, θ), where i and j are gray values, d is the distance between pixel pairs, and θ is the direction, to obtain skin texture features;

[0025] The number of pixels with different grayscale values ​​in the image data is counted to obtain a color histogram, and the probability of the grayscale value appearing in k pixels is calculated, which is expressed as: where n k is the number of pixels with gray value k, N is the total number of pixels in the image, H(k) represents the probability of occurrence of a pixel with gray value k, k = 0, 1, ..., L-1, and the skin color feature is obtained;

[0026] Based on the preset spot threshold Q, the image data is binarized to obtain a binary image B(x, y), which is expressed as: Where I(x, y) is the pixel value, I(x, y)>Q means that the pixel is marked as a spot pixel, and I(x, y)≤Q means that the pixel is marked as a background pixel; by performing connected area analysis on the binary image, the number of spots is counted as N s , the area of ​​each spot A i ,i=1,2,...,N s and the total area of ​​the spots Get skin spot characteristics;

[0027] The image feature vector E includes skin texture features e1, skin color features e2 and skin spot features e3.

[0028] Preferably, the data fusion of the physiological feature vector and the image feature vector to obtain a multi-source data feature vector includes:

[0029] Assign weight ω to the physiological feature vector p ;

[0030] Assign weight ω to the image feature vector e ;

[0031] Perform weighted summation on the physiological feature vector P and image feature vector E of the same dimension to obtain the multi-source data feature vector:

[0032] F=ω p P+ω e E

[0033] Among them, ω p +ω e =1, F is the multi-source data feature vector.

[0034] Preferably, inputting the multi-source data features into a skin condition prediction model to generate a skin condition prediction result includes:

[0035] The skin condition prediction model was constructed using a multi-layer perceptron, which consists of an input layer, multiple hidden layers, and an output layer.

[0036] The input layer has m neurons, corresponding to the m dimensions of the multi-source data feature vector F, where the number of neurons in the lth layer is n l , the activation function is σ l (·);

[0037] In the hidden layer and output layer, the output h of the lth layer l and the output h of the l-1 layer l-1 The relationship between them is:

[0038] h l =σ l (W l h l-1 +b l )

[0039] Among them, W l is the weight matrix of the lth layer, b l is the bias vector;

[0040] The input layer h0=F, after multiple layers of calculation, the output layer h L is the prediction result of skin condition, where L is the total number of layers of the network, and the skin condition prediction model is obtained;

[0041] The new multi-source data feature vector F * As input, it is input into the skin condition prediction model, starting from the input layer, and then through the calculation of each hidden layer, and finally the result of the output layer is obtained. but The skin condition prediction results are generated.

[0042] Preferably, the skin condition prediction result is evaluated based on the skin condition standard to obtain the evaluation result, including:

[0043] According to the skin condition standard, the evaluation indexes are expressed as C1, C2, ..., C n , the weight corresponding to each evaluation indicator is α1,α2,...,α n ,and

[0044] According to the criteria of each evaluation index, the skin condition prediction result Converted into the score S(C i );

[0045] The evaluation result Z is obtained by weighted summation, which is expressed as:

[0046]

[0047] Among them, Z is the evaluation result reflecting the overall condition of the skin. The higher the Z value, the better the skin condition; conversely, the lower the Z value, the worse the skin condition.

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

[0049] 1. This invention integrates physiological data and image data to evaluate skin conditions from multiple perspectives, avoiding the limitations of a single data source and reflecting the true condition of the skin more comprehensively and accurately. Furthermore, through data processing and model prediction, it achieves quantitative assessment of skin conditions, provides objective and accurate assessment results, and facilitates early detection and intervention of skin problems.

[0050] 2. This invention establishes an evaluation system based on multi-dimensional skin condition standards, focusing not only on the skin's appearance but also on the skin's physiological state, making the evaluation results more scientific and practical. Through model training and evaluation, it can provide each user with personalized skin condition predictions and care recommendations, which is different from traditional general evaluation methods and is more in line with individual differences. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of the skin condition prediction method of the present invention. DETAILED DESCRIPTION

[0052] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0053] As attached Figure 1 As shown:

[0054] Embodiment 1: The present invention provides a method for predicting skin condition, comprising the following steps:

[0055] S1. The real-time state of the skin is collected through wearable devices to obtain physiological data; the skin image is collected through skin imaging equipment to obtain image data; the physiological data can reflect the internal physiological state of the skin, and the image data can intuitively present the external performance of the skin.

[0056] S2. Preprocess the physiological data and extract physiological feature vectors; the physiological feature vector P includes skin temperature p1, skin humidity p2 and oil secretion p3;

[0057] Calculate the rate of change of skin temperature at adjacent time points to reflect the dynamic change of skin temperature. The rate of change of skin temperature R T Expressed as: Where Δt is the time interval, T n and T n-1 The temperature data at adjacent time points are used to obtain the skin temperature characteristics;

[0058] Calculate the rate of change of skin humidity at adjacent time points to reflect the dynamic changes of skin humidity. The rate of change of skin humidity R H Expressed as: Where Δt is the time interval, H n and H n-1 The skin moisture data at adjacent time points are used to obtain the skin moisture characteristics;

[0059] Calculate the rate of change of oil secretion at adjacent time points to reflect the dynamic changes of skin oil secretion. The rate of change of oil secretion R O Expressed as: Where Δt is the time interval, O n and O n-1 The oil secretion data at adjacent time points are used to obtain the oil secretion characteristics.

[0060] S3, preprocessing the image data and extracting the image feature vector; the image feature vector E includes skin texture feature e1, skin color feature e2 and skin spot feature e3;

[0061] Calculate the gray-level co-occurrence matrix of the image data to describe the spatial relationship between pixels of different gray values ​​in the image data. The gray-level co-occurrence matrix is ​​expressed as: P(i, j, d, θ), where i and j are gray values, d is the distance between pixel pairs, and θ is the direction, to obtain skin texture features;

[0062] Count the number of pixels with different grayscale values ​​in the image data to obtain a color histogram, and calculate the probability of the grayscale value appearing in k pixels, which is expressed as: where n kis the number of pixels with gray value k, N is the total number of pixels in the image, H(k) represents the probability of occurrence of a pixel with gray value k, k = 0, 1, ..., L-1, and the skin color feature is obtained;

[0063] Based on the preset spot threshold Q, the image data is binarized to obtain a binary image B(x,y), which is expressed as: Where I(x,y) is the pixel value, I(x,y)>Q means that the pixel is marked as a spot pixel, and I(x,y)≤Q means that the pixel is marked as a background pixel. By performing connected region analysis on the binary image, the number of spots is counted as N s , the area of ​​each spot A i ,i=1,2,...,N s and the total area of ​​the spots Get skin spot characteristics.

[0064] S4, fusing physiological features and image features to obtain multi-source data feature vectors;

[0065] Assign weight ω to the physiological feature vector p ;

[0066] Assign weight ω to the image feature vector e ;

[0067] The weighted sum of the physiological feature vector P and the image feature vector E with the same dimension is expressed as:

[0068] F=ω p P+ω e E

[0069] Among them, ω p +ω e =1, F is the multi-source data feature vector.

[0070] Features extracted from physiological data and image data are fused to form multi-source data features, thereby comprehensively utilizing the advantages of both data and making up for the shortcomings of a single data source; multi-source data features can more comprehensively describe skin conditions, integrate physiological and image information, and provide richer and more comprehensive feature vectors for accurate prediction of skin conditions, which helps to improve the accuracy and reliability of predictions.

[0071] S5. Inputting the multi-source data features into the skin condition prediction model to generate a skin condition prediction result; a skin condition prediction model is constructed using a multi-layer perceptron, which consists of an input layer, multiple hidden layers, and an output layer;

[0072] The input layer has m neurons, corresponding to the m dimensions of the multi-source data feature vector F, where the number of neurons in the lth layer is n l , the activation function is σl (·);

[0073] In the hidden layer and output layer, the output h of the lth layer is l and the output h of the l-1 layer l-1 The relationship between them is:

[0074] h l =σ l (W l h l-1 +b l )

[0075] Among them, W l is the weight matrix of the lth layer, b l is the bias vector;

[0076] Input layer h0=F, after multiple layers of calculation, output layer h L is the prediction result of skin condition, where L is the total number of layers of the network, and the skin condition prediction model is obtained;

[0077] The new multi-source data feature vector F * As input, it is input into the skin condition prediction model. Starting from the input layer, it passes through the calculation of each hidden layer in turn, and finally obtains the result of the output layer. but The skin condition prediction results are generated.

[0078] The model is trained using the fused multi-source data features. By learning patterns and regularities within the data, it establishes a mapping relationship between input features and skin conditions. The trained model can predict skin conditions based on the input multi-source data features and output corresponding predictions, providing a quantitative basis for skin condition assessment.

[0079] S6. Based on the skin condition standard, the skin condition prediction result is evaluated to obtain an evaluation result.

[0080] According to the skin condition standard, the evaluation index is expressed as C1, C2, ..., C n , the weight corresponding to each evaluation indicator is α1,α2,...,α n ,and

[0081] According to the criteria of each evaluation index, the skin condition prediction result h L * Converted into the score S(C i );

[0082] For skin evaluation index C1, if The corresponding eigenvalue is x, and the ideal range of its index is [a, b]. When x∈[a, b], S(C1)=100; when x<a, Where v1 is a constant, min is the minimum possible value of the eigenvalue; when x>a, Where v1′ is a constant and max is the maximum possible value of the eigenvalue.

[0083] For other evaluation indicators C i , predict the results based on their relationship with skin condition in the same way The relationship between the corresponding features in the definition score function S(C i );

[0084] The evaluation result Z is obtained by weighted summation, which is expressed as:

[0085]

[0086] Among them, Z is the evaluation result reflecting the overall condition of the skin. The higher the Z value, the better the skin condition; conversely, the lower the Z value, the worse the skin condition.

[0087] Based on pre-set skin condition standards, the prediction results are evaluated from multiple dimensions to obtain comprehensive evaluation results; the prediction results are converted into practical skin condition descriptions for easy understanding and application; the evaluation can intuitively understand the overall condition of the skin and its performance in various dimensions, providing users with detailed skin condition analysis, which helps to formulate targeted skin care plans.

[0088] Experimental example:

[0089] Skin temperature, humidity, and oil secretion data of 100 subjects were collected through wearable devices, recorded once every hour for one week.

[0090] Professional skin imaging equipment was used to collect facial skin images of the subjects, including frontal, left and right side images, and a total of 300 images were obtained.

[0091] Clean, standardize and extract features of physiological data, extract statistical features (mean, standard deviation, etc.) and change rate;

[0092] Grayscale, denoise, and normalize the image data, and then extract skin texture, color, and spot features, including gray-level co-occurrence matrix features and color histogram features;

[0093] The physiological features and image features are fused in a weighted manner to obtain multi-source data features;

[0094] A multi-layer perceptron (MLP) was used as a skin condition prediction model, with 70% of the data used as a training set and 30% of the data used as a test set. The model was trained on the training set and then predicted on the test set to obtain the skin condition prediction results.

[0095] In the data preprocessing results, some statistical characteristics of physiological data are shown in the following table:

[0096]

[0097]

[0098] In the data preprocessing results, the feature extraction results of some image data are shown in the following table:

[0099] Subject number Texture Energy Texture contrast Color mean Number of spots 1 0.12 0.35 150.2 5 2 0.15 0.32 148.6 3

[0100] Based on the set skin condition standards, the skin is evaluated from three dimensions: skin health, smoothness, and color. The weights of each dimension are 0.4, 0.3, and 0.3 respectively. The comprehensive evaluation results are calculated as shown in the following table:

[0101]

[0102] Experimental results demonstrate that this skin condition prediction method effectively integrates physiological and image data features, uses a multi-layer perceptron model to predict skin conditions, and obtains reasonable evaluation results based on the established evaluation criteria. Combined with a practical questionnaire, this method demonstrates its feasibility and effectiveness in predicting skin conditions.

[0103] It is important to note that the construction and arrangement of the present application shown in a plurality of different exemplary embodiments are merely illustrative. Although only a few embodiments are described in detail in this disclosure, it will be readily understood by those who consult this disclosure that many modifications are possible without departing substantially from the novel teachings and advantages of the subject matter described in this application. Other replacements, modifications, changes, and omissions may be made in the design, operating conditions, and arrangement of the exemplary embodiments without departing from the scope of the present invention. Therefore, the present invention is not limited to specific embodiments, but extends to a variety of modifications still falling within the scope of the appended claims.

[0104] Additionally, in order to provide a concise description of exemplary embodiments, all features of an actual embodiment (i.e., those features that are not relevant to the best mode presently contemplated for carrying out the invention or those that are not relevant to implementing the invention) may not be described.

[0105] It will be understood that in the development of any actual embodiment, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but for those of ordinary skill having the benefit of this disclosure, the development effort will be a routine task of design, fabrication, and production without undue experimentation.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A skin condition prediction method, characterized in that: The following steps are involved: S1. Collect the real-time state of the skin through the wearable device to obtain physiological data; collect the image of the skin through the skin imaging device to obtain image data; S2. Preprocessing the physiological data to extract physiological feature vectors; S3, preprocessing the image data and extracting image feature vectors; S4, fusing the physiological features and image features to obtain a multi-source data feature vector; S5. Inputting the multi-source data features into a skin condition prediction model to generate a skin condition prediction result; S6. Evaluate the skin condition prediction result based on the skin condition standard to obtain an evaluation result.

2. A skin condition prediction method according to claim 1, characterized in that: The preprocessing of the physiological data and feature extraction of physiological feature vectors include: The physiological characteristic vector P includes skin temperature p1, skin humidity p2 and oil secretion p3; The rate of change of skin temperature at adjacent time points is calculated using the following formula: T : Among them, R T is the rate of change of the skin temperature, which is used to reflect the dynamic change of the skin temperature, Δt is the time interval, T n and T n-1 The temperature data at adjacent time points are used to obtain the skin temperature characteristics; The following formula is used to calculate the change rate R of the skin moisture at adjacent time points: H : Among them, R H is the rate of change of skin humidity, which is used to reflect the dynamic change of skin humidity. Δt is the time interval, H n and H n-1 The skin moisture data at adjacent time points are used to obtain the skin moisture characteristics; The following formula is used to calculate the rate of change R of the oil secretion at adjacent time points: O : Among them, R O is the rate of change of oil secretion, which is used to reflect the dynamic change of skin oil secretion. Δt is the time interval. n and O n-1 The oil secretion data at adjacent time points are used to obtain the oil secretion characteristics.

3. A skin condition prediction method according to claim 1, characterized in that: The preprocessing of the image data and feature extraction of image feature vectors includes: Calculating a gray level co-occurrence matrix of the image data, wherein the gray level co-occurrence matrix is ​​used to describe the spatial relationship between pixels of different gray values ​​in the image data, wherein the gray level co-occurrence matrix is ​​expressed as: P(i, j, d, θ), where i and j are gray values, d is the distance between pixel pairs, and θ is the direction, to obtain skin texture features; The number of pixels with different grayscale values ​​in the image data is counted to obtain a color histogram, and the probability of the grayscale value appearing in k pixels is calculated, which is expressed as: where n k is the number of pixels with gray value k, N is the total number of pixels in the image, H(k) represents the probability of occurrence of a pixel with gray value k, k = 0, 1, ..., L-1, and the skin color feature is obtained; Based on the preset spot threshold Q, the image data is binarized to obtain a binary image B(x, y), which is expressed as: Where I(x, y) is the pixel value, I(x, y)>Q means that the pixel is marked as a spot pixel, and I(x, y)≤Q means that the pixel is marked as a background pixel; by performing connected area analysis on the binary image, the number of spots is counted as N s , the area of ​​each spot A i ,i=1,2,...,N s and the total area of ​​the spots Get skin spot characteristics; The image feature vector E includes skin texture features e1, skin color features e2 and skin spot features e3.

4. A skin condition prediction method according to claim 1, characterized in that: The step of fusing the physiological feature vector and the image feature vector to obtain a multi-source data feature vector includes: Assign weight ω to the physiological feature vector p ; Assign weight ω to the image feature vector e ; Perform weighted summation on the physiological feature vector P and image feature vector E of the same dimension to obtain the multi-source data feature vector: F=ω p P+ω e E Among them, ω p +ω e =1, F is the multi-source data feature vector.

5. The skin condition prediction method according to claim 1, wherein: Inputting the multi-source data features into a skin condition prediction model to generate a skin condition prediction result includes: The skin condition prediction model was constructed using a multi-layer perceptron, which consists of an input layer, multiple hidden layers, and an output layer. The input layer has m neurons, corresponding to the m dimensions of the multi-source data feature vector F, where the number of neurons in the lth layer is n l , the activation function is σ l (·); In the hidden layer and output layer, the output h of the lth layer l and the output h of the l-1 layer l-1 The relationship between them is: h l =σ l (W l h l-1 +b l ) Among them, W l is the weight matrix of the lth layer, b l is the bias vector; The input layer h0=F, after multiple layers of calculation, the output layer h L is the prediction result of skin condition, where L is the total number of layers of the network, and the skin condition prediction model is obtained; The new multi-source data feature vector F * As input, it is input into the skin condition prediction model, starting from the input layer, and then through the calculation of each hidden layer, and finally the result of the output layer is obtained. but The skin condition prediction results are generated.

6. A skin condition prediction method according to claim 1, characterized in that: The step of evaluating the skin condition prediction result based on the skin condition standard to obtain an evaluation result includes: According to the skin condition standard, the evaluation indexes are expressed as C1, C2, ..., C n , the weight corresponding to each evaluation indicator is α1,α2,...,α n ,and According to the criteria of each evaluation index, the skin condition prediction result Converted into the score S(C i ); The evaluation result Z is obtained by weighted summation, which is expressed as: Among them, Z is the evaluation result reflecting the overall condition of the skin. The higher the Z value, the better the skin condition; conversely, the lower the Z value, the worse the skin condition.

Images