A skin oil evaluation method based on test paper

By using a test strip-based method for evaluating skin oil, combined with machine learning models and image features, the accuracy and cost issues of skin oil detection have been resolved, achieving low-cost and stable skin oil assessment.

CN115343279BActive Publication Date: 2024-11-15SHANGHAI CHINA-NORM MANAGEMENT CONSULTING CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210839123.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-18
Publication Date
2024-11-15
Estimated Expiration
2042-07-18

AI Technical Summary

Technical Problem

Existing methods for detecting skin oil are not accurate enough for remote evaluation, manual evaluation is costly and subjective, laboratory instruments cannot meet the needs of home-based self-testing, and image recognition methods are greatly affected by ambient light and shooting angle.

Method used

A skin oil assessment method based on test strips is adopted. Skin oil is collected by test strips, image feature information is obtained, and machine learning models are used to evaluate it by combining area and light features. GBDT, XGBoost, LinearModels or SVM models are constructed to accurately judge the skin oil status.

Benefits of technology

It achieves accurate, stable, and low-cost evaluation of skin oil conditions, overcomes human subjective error, is suitable for large-scale promotion, and has an average relative error of less than 15%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115343279B_ABST
    Figure CN115343279B_ABST
Patent Text Reader

Abstract

The present invention relates to a skin oil evaluation method based on a test paper, comprising: step (1): collecting skin oil through a test paper and obtaining a test paper image with skin oil; step (2): collecting feature information of irregular oil regions in the test paper image; step (3): inputting the feature information of the irregular oil regions into a skin oil evaluation model and outputting the oil condition of the skin. The present invention can effectively evaluate the skin oil condition and improve the efficiency of skin oil condition evaluation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of computer image processing, and in particular to a skin oil evaluation method based on test paper. Background Art

[0002] Clinically, skin can be roughly divided into several types according to the amount of lipids on the skin surface: oily skin, dry skin, neutral skin and mixed skin. People with oily skin have more active secretion of sebaceous glands, and their faces are greasy, difficult to clean, and have oily sheen that affects their appearance. They are also prone to corresponding facial skin diseases, such as acne and seborrheic dermatitis. Controlling excessive secretion of sebaceous glands and reducing oily symptoms are the key to maintaining healthy skin. Therefore, the oil condition of the skin is a good indicator of skin health.

[0003] When evaluating the oiliness of the subjects remotely, since it is impossible to touch the skin, judging by images alone will result in extremely inaccurate evaluation results due to factors such as light, angle, ambient light, camera, and recognition technology. Manual remote evaluation is highly subjective and ambiguous, and the diagnosis results are closely related to the doctor's experience. At the same time, manual evaluation takes a long time and has high labor costs. Therefore, photo recognition or remote evaluation by doctors can no longer meet the needs of accurate online diagnosis.

[0004] Here are two typical skin oil detection methods:

[0005] Laboratory instrument Sebumeter detection method: The measuring instrument needs to be used with a sebum measurement box, which contains a roll of matte tape with a thickness of 0.1 mm. There is a mirror under the measuring part of the tape. During measurement, the measuring part of the tape is pressed against the skin, and when the tape absorbs skin oil, it becomes translucent. Then insert the measuring box into the detection hole of the device, and the light emitted by the light source in the detection hole is reflected to the photodiode through the mirror behind the measuring tape, and the light intensity is recorded. The light transmittance indicates the oil content in the measuring area of ​​the tape. Advantages of the method: fast speed, high accuracy, stability and reliability, widely used in laboratory clinical testing; Disadvantages of the method: The instrument is expensive and bulky, and cannot meet the needs of subjects for self-sampling at home.

[0006] Recognition method based on image deep learning: This method builds an end-to-end (facial image, oil value) deep learning model. It can collect a large number of skin photos taken by mobile phones or other instruments to build an image data set, and use big data to continuously increase the accuracy and reliability of the model. Advantages of the method: It can also be well recognized in different scenes and image quality conditions. At the same time, the judgment rules are unified and stable, and the oil evaluation efficiency is also high; Disadvantages of the method: When acquiring and recognizing images, ambient light, shooting angle, shooting hardware, photo clarity, recognition technology and other factors will affect the accuracy of the oil results. In addition, this method does not contact the subject's skin and only relies on the light information reflected in the picture. The reliability is still under discussion.

[0007] Both methods have their own advantages and disadvantages. The Sebumeter is widely recognized and is definitely the first choice for laboratory measurements. Although the deep learning modeling method innovatively uses a large number of images for learning, the reliability of modeling and prediction based only on optical information remains to be discussed. In addition, due to the impact of image quality issues, the application recognition of this method may be low. Summary of the invention

[0008] The technical problem to be solved by the present invention is to provide a skin oil evaluation method based on test paper, which can effectively evaluate the skin oil condition and improve the efficiency of skin oil condition evaluation.

[0009] The technical solution adopted by the present invention to solve the technical problem is: to provide a skin oil evaluation method based on a test paper, comprising:

[0010] Step (1): collecting skin oil through a test paper and obtaining an image of the test paper with skin oil;

[0011] Step (2): collecting characteristic information of irregular oil regions in the test paper image;

[0012] Step (3): input the characteristic information of the irregular oily area into a skin oil evaluation model, and output the oily condition of the skin.

[0013] The characteristic information of the irregular grease area in step (2) includes area characteristics and light characteristics.

[0014] The area feature includes a first area feature, and the first area feature is obtained by the following steps:

[0015] Step (21): detecting the contour of the irregular grease region in the center of the test paper image, and calculating the centroid of the irregular grease region based on the contour of the irregular grease region;

[0016] Step (22): intercepting the irregular grease area based on the coordinates of the centroid;

[0017] Step (23): Calculate the area of ​​the irregular oil region and use it as the first area feature input into the skin oil evaluation model.

[0018] The area feature includes a second area feature, and the second area feature is obtained by the following steps:

[0019] Step (24): performing image opening operation on the irregular grease area to obtain a plurality of small contour areas;

[0020] Step (25): sorting the areas of the small contours from large to small, and summing them up in bins as the second area feature input into the skin oil evaluation model.

[0021] The optical characteristics are obtained by the following steps:

[0022] Step (26): converting the image corresponding to the irregular grease area into the HSI color space, separating the I channel image, and calculating the optical density integral value of the irregular grease area in the I channel image;

[0023] Step (27): bin and sum the optical density integral values ​​of the irregular oily areas in the I channel image to serve as the optical features input into the skin oil evaluation model.

[0024] The step (21) uses an edge detection operator to detect the contour of the irregular grease area in the center of the test paper image.

[0025] The step (1) specifically comprises: pressing the test paper on the skin for a preset time to obtain the test paper with skin oil, photographing the test paper with skin oil, and obtaining an image of the test paper with skin oil.

[0026] The step (3) also includes: training the skin oil evaluation model by inputting a plurality of skin oil images, wherein each skin oil image corresponds to a Sebumeter value, and continuously optimizing the skin oil evaluation model by comparing the output value of each skin oil image with the Sebumeter value corresponding to each skin oil image.

[0027] The skin oil evaluation model in step (3) is a GBDT model, an XGBoost model, a LinearModels model or a SVM model.

[0028] Beneficial Effects

[0029] Due to the adoption of the above-mentioned technical scheme, the present invention has the following advantages and positive effects compared with the prior art: the present invention utilizes a machine learning method to construct a skin oil evaluation model, and realizes accurate evaluation of skin oil by combining an artificial intelligence model algorithm with a test strip. According to experimental analysis, the average relative percentage error of the present invention is within 15%. A stable light source is set during image data acquisition, and multiple models of cameras are used for image acquisition. The model has strong adaptability. The present invention can provide an objective and stable evaluation standard for the evaluation of skin oil conditions by collecting area features and light features. The collection method based on the test strip is stable and reliable, easy to operate, and low in cost. At the same time, the evaluation based on the artificial intelligence model overcomes the errors caused by human subjective reasons. It is simple to use, highly reliable, and suitable for large-scale promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a schematic diagram of the skin oil collection and evaluation process according to an embodiment of the present invention;

[0031] Figure 2 It is a schematic diagram of the process of extracting features from an image by a skin oil evaluation model according to an embodiment of the present invention;

[0032] Figure 3 It is a schematic diagram of the collection position of the skin oil evaluation model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0034] The embodiment of the present invention relates to a skin oil evaluation method based on a test paper, please refer to Figure 1 and Figure 2 ,include:

[0035] Step (1): collecting skin oil through a test paper and obtaining an image of the test paper with skin oil;

[0036] Step (2): collecting characteristic information of irregular oil regions in the test paper image;

[0037] Step (3): input the characteristic information of the irregular oily area into a skin oil evaluation model, and output the oily condition of the skin.

[0038] Furthermore, the step (1) is specifically as follows: first, use the test paper to press on the skin to collect oil, be careful not to let fingers or other objects contaminate the test area of ​​the test paper, and use a single finger of one hand to press on the skin for about 30 seconds. Then use a camera or a mobile phone to take pictures of the test paper with oil, and it is recommended to magnify it by about 2.5 times, keep the lens about 20 cm away from the test paper, use a side light source to shoot (avoid shadows), and pay attention to the four sides of the camera being parallel to the test paper, and pay attention to not blur the photos.

[0039] The test paper is pressed on the skin for a preset time to obtain the test paper with skin oil, and the test paper with skin oil is photographed to obtain an image of the test paper with skin oil.

[0040] Furthermore, the characteristic information of the irregular grease area in step (2) includes area characteristics and light characteristics, which are described in detail below:

[0041] 1. The area feature includes a first area feature, and the first area feature is obtained by the following steps:

[0042] Step (21): using an edge detection operator (such as the Canny operator and findContours in OpenCV) to detect the contour of the irregular grease area in the center of the test paper image, and calculating the centroid of the irregular grease area based on the contour of the irregular grease area;

[0043] Step (22): intercepting the irregular grease area based on the coordinates of the centroid;

[0044] Step (23): Calculate the area of ​​the irregular oily region and use it as the first area feature input into the skin oil evaluation model; calculate the area using a grayscale image + threshold segmentation method, and take two areas obtained by two strict and loose thresholds respectively (both areas are used as features to input into the subsequent skin oil evaluation model), for the two extreme cases of lighter and heavier conditions.

[0045] 2. The area feature includes a second area feature, and the second area feature is obtained by the following steps:

[0046] Step (24): performing image opening operation on the irregular grease area to obtain a number of small contour areas; the opening operation is a mathematical morphology image processing method, that is, first performing an erosion operation and then performing an expansion operation, which can achieve the effect of segmenting finely connected graphics while keeping the overall pixel area unchanged;

[0047] Step (25): sorting the small contour areas by size, and summing them up in bins to serve as the second area feature input into the skin oil evaluation model.

[0048] In this embodiment, after sorting the small contour areas from large to small, for example, the area sizes of the 1st to 3rd, 4th to 8th, 9th to 25th, 26th to 45th, and 46th to 100th intervals are merged as the characteristics of the five box intervals and input into the skin oil evaluation model.

[0049] 3. The optical characteristics are obtained by the following steps:

[0050] Step (26): converting the image corresponding to the irregular grease area into the HSI color space, separating the I channel image, and calculating the optical density integral value of the irregular grease area in the I channel image, the feature including the area and the image reflection intensity information;

[0051] Step (27): bin and sum the optical density integral values ​​of the irregular oily areas in the I channel image to serve as the optical features input into the skin oil evaluation model.

[0052] Since the value range of each pixel in the image I channel is 0 to 255, for example, this embodiment divides 0 to 255 into 32 equal-width box intervals, the first interval is 0 to 7, the second interval is 8 to 15, and so on. The pixels in the corresponding intervals of the I channel image are summed up (such as counting the number of pixels with intensity values ​​of 0 to 7, and then summing the pixels with intensity values ​​of 0 to 7 to obtain the area), which is an integral result value of the comprehensive area and intensity value, which is the optical density integral in a logical sense, and a total of 32 groups of box interval features can be obtained. In actual use, these 32 equal-width intervals are appropriately selected (such as selecting the 8th to 18th groups of intervals of the 32 equal-width box intervals) and then input into the skin oil evaluation model to express stronger characteristic attributes.

[0053] Furthermore, the skin oil evaluation model of this embodiment is established based on a machine learning method, and is used to evaluate the skin oiliness in combination with oil test strips. The model first needs to prepare a large number of photos of oil test strips as a database, and each photo has a corresponding Sebumeter value. Due to accuracy considerations, one test strip is photographed at a time.

[0054] In this embodiment, the skin oil image based on the test paper is defined as a database, and image processing is performed on the image data of the database to obtain different feature values. Then, a machine learning model for predicting specific oil values ​​is prepared and built, that is, the processed feature data is input into the skin oil evaluation model for training, and the training results are obtained. The trained model is performance tested. If the model cannot meet the performance requirements, the model training parameters are adjusted and the model is retrained until a skin oil evaluation model that meets the performance requirements is trained.

[0055] Collection of skin oil as shown in the figure Figure 3As shown, A is the test paper collection point, and B is the Sebumeter instrument collection point. This implementation method is as close to the Sebumeter test position as possible, and selects a visually consistent area, and the sampling time is kept consistent with the instrument test time.

[0056] The machine learning model constructed in this embodiment can be a model for processing regression problems such as a GBDT model, an XGBoost model, a LinearModels model or an SVM model. At the beginning of the skin oil evaluation model training, the parameters are set according to the detection requirements of the model. For example, in the XGBoost model, the following settings can be made: the tree depth is set to 6, the evaluation index is customized such as the percentage mean relative error, the learning rate is set to 0.1, the column sampling coefficient is set to 0.8, the KFold cross-validation function is written, and so on.

[0057] When performing a performance test on the constructed skin oil evaluation model, first prepare several test sample data (test set) that are independent of the training data, and then process them according to the image feature extraction process during training, use the trained model to evaluate the obtained feature data, and finally perform consistency analysis statistics and evaluation index calculation on the model evaluation results and Sebumeter. If the consistency between the model test results and the test data annotation results is greater than 95%, and the percentage average relative error is less than 15%, the detection model is considered to meet the requirements. This embodiment uses three hundred test data when testing the detection model. Any value can be used in actual applications, generally not less than three hundred. The more test data samples there are, the higher the model accuracy.

[0058] The present invention utilizes a machine learning method to construct a skin oil evaluation model, and realizes accurate evaluation of skin oil by combining an artificial intelligence model algorithm with a test strip. Experimental analysis shows that the average relative percentage error of the present invention is within 15%. A stable light source is set during image data acquisition, and multiple models of cameras are used for image acquisition. The model has strong adaptability. The present invention can provide an objective and stable evaluation standard for the evaluation of skin oil conditions. The acquisition method based on the test strip is stable and reliable, easy to operate, and low in cost. At the same time, the evaluation is based on the artificial intelligence model, which overcomes the errors caused by human subjective reasons. The method is simple to use, highly reliable, and suitable for large-scale promotion.

[0059] The above schematically describes the invention and its implementation methods, which is not restrictive. Without departing from the spirit or basic features of the invention, the invention can be implemented in other specific forms. What is shown in the accompanying drawings is only one of the implementation methods of the invention. The actual structure is not limited to this, and any figure mark in the claims should not limit the claims involved. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the invention, a structural method and an embodiment similar to the technical solution are designed without creativity, which should all fall within the scope of protection of this patent. In addition, the word "including" does not exclude other elements or steps, and the word "one" before the element does not exclude the inclusion of "multiple" elements. The multiple elements stated in the product claim can also be implemented by one element through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any specific order.

Claims

1. A skin oil evaluation method based on test paper, characterized in that: include: Step (1): collecting skin oil through a test paper and obtaining an image of the test paper with skin oil; Step (2): collecting feature information of the irregular oil region in the test paper image; wherein the feature information of the irregular oil region includes area feature and light feature; wherein the area feature includes a first area feature and a second area feature; The first area feature is obtained by the following steps: Step (21): detecting the contour of the irregular grease region in the center of the test paper image, and calculating the centroid of the irregular grease region based on the contour of the irregular grease region; Step (22): intercepting the irregular grease area based on the coordinates of the centroid; Step (23): calculating the area of ​​the irregular oil region and using it as the first area feature input into the skin oil evaluation model; The second area feature is obtained by the following steps: Step (24): performing image opening operation on the irregular grease area to obtain a plurality of small contour areas; Step (25): sorting the areas of the small contours from large to small, and summing them up in bins to serve as the second area feature input into the skin oil evaluation model; The optical characteristics are obtained by the following steps: Step (26): converting the image corresponding to the irregular grease area into the HSI color space, separating the I channel image, and calculating the optical density integral value of the irregular grease area in the I channel image; Step (27): binning and summing the optical density integral values ​​of the irregular oil region in the I channel image as the optical features input into the skin oil evaluation model; Step (3): input the characteristic information of the irregular oily area into a skin oil evaluation model, and output the oily condition of the skin.

2. The skin oil evaluation method based on test paper according to claim 1, characterized in that: The step (21) uses an edge detection operator to detect the contour of the irregular grease area in the center of the test paper image.

3. The skin oil evaluation method based on test paper according to claim 1, characterized in that: The step (1) specifically comprises: pressing the test paper on the skin for a preset time to obtain the test paper with skin oil, photographing the test paper with skin oil, and obtaining an image of the test paper with skin oil.

4. The skin oil evaluation method based on test paper according to claim 1, characterized in that: The step (3) also includes: training the skin oil evaluation model by inputting a plurality of skin oil images, wherein each skin oil image corresponds to a Sebumeter value, and continuously optimizing the skin oil evaluation model by comparing the output value of each skin oil image with the Sebumeter value corresponding to each skin oil image.

5. The skin oil evaluation method based on test paper according to claim 1, characterized in that: The skin oil evaluation model in step (3) is a GBDT model, an XGBoost model, a LinearModels model or a SVM model.