Sensitive skin automatic discrimination method combining skin color and acne distribution

Abstract

The invention belongs to the technical field of skin detection, and particularly relates to a skin color and acne distribution combined sensitive skin automatic discrimination method comprising the following steps: collecting a user face image through a mobile terminal; identifying a skin color type based on an LAB color space; extracting a red area value and a brightness value of each area of the face by adopting an HSV color model; positioning a face acne area based on the lightweight target detection network; respectively calculating the acne density of each area of the face; calculating an area sensitivity index based on the red area value, the acne density and the skin color type of each area; according to the method, a user face image is collected, a double-color space parallel processing mode is adopted, red area values and skin types of all areas of the face of the user are synchronously output, and then sensitivity indexes capable of reflecting the sensitivity degrees of all the areas of the face of the user are obtained based on calculation of the acne density of all the areas of the face, and the sensitivity index is corrected by the skin correction coefficient corresponding to the skin type, so that missed diagnosis and misdiagnosis caused by skin color difference can be reduced to a certain extent.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of skin detection, and particularly relates to a sensitive skin automatic identification method combining skin color and acne distribution. BACKGROUND

[0002] Reasonable evaluation of facial skin quality plays an important role in the fields of make-up industry, dermatology and plastic surgery. At present, skin testers and other products on the domestic and foreign markets have the shortcomings of high price, large size and inconvenience for carrying. If an artificial evaluation method is used, different people generally have different evaluations on the same skin sample, so it is difficult to achieve consistent evaluation. For a skin surgeon, it is very important to provide objective results by excluding human subjectivity.

[0003] In the prior art, a mobile terminal is used to collect a user's facial image, and then image processing is performed to analyze the sensitive skin condition of the user's face. On this basis, the user can know whether the face is sensitive according to the analysis result, and the specific sensitive position, and the corresponding care according to the sensitive level, and continuously track the skin state after the care, so as to realize the whole process management of skin detection-care-feedback. In the prior art, only a single image processing method is used, such as converting the image into the HSV color space to obtain the red region value corresponding to the user's facial image, and analyzing the sensitive index based on the red region value. However, in the above process, the influence of skin quality on image processing is ignored, and thus the output sensitive index may not truly reflect the real skin sensitive state of the user, resulting in misdiagnosis and missed diagnosis.

[0004] Therefore, the application provides a sensitive skin automatic identification method combining skin color and acne distribution. SUMMARY

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

[0006] The technical scheme adopted by the application to solve the technical problems is that the sensitive skin automatic identification method combining skin color and acne distribution comprises the following steps: S1: collecting a user's facial image through a mobile terminal; S2: identifying the skin color type based on the LAB color space; S3: extracting the red region value of each region of the face by using the HSV color model and the brightness value ; S4: positioning the acne region of the face based on a lightweight target detection network; S5: calculating the acne density of each region of the face ; S6: Red zone values ​​based on each region , acne density and skin color type to calculate regional sensitivity index ; S7: According to the sensitivity index of each region and set threshold The comparison result is output as the sensitivity determination result.

[0007] Preferably, the method for identifying skin color type is: Calculating lightness in LAB color space Value and yellow-blue axis The ratio of values; According to the skin color classification rules: type ; according to , determine the user's skin color type.

[0008] Preferably, in the HSV color model conversion: Red Zone Value Calculated by the pixel ratio of the H channel in the range of 0-10° and 170-180°; Lightness value Used to eliminate light interference, when Start fill light when The facial areas include the face zone, the T zone and the contour zone.

[0009] Preferably, the acne density Calculations include: Each district uses an independent YOLOv5s model for detection: The face area: input size 416×416, confidence level 0.7; The T zone: input size 320×320, confidence level 0.6; The contour area: input size 256×256, confidence level 0.5; According to the formula: ； in, For the region The number of acne; For the region area.

[0010] Preferably, the method for calculating the regional sensitivity index is: Extract the red zone values ​​of each area of ​​the face and acne density ; According to the formula: ; wherein is a red region weight coefficient, is a acne density weight coefficient, is a skin color correction coefficient; wherein, .

[0011] Preferably, the method of outputting the sensitive determination result is: comparing the calculated regional sensitive index with a set threshold value : when , the region is determined to be sensitive, and when the number of sensitive regions , or the cheek region is sensitive, it is determined to be all sensitive; wherein, according to the size of the regional sensitive index , the sensitive level of the corresponding region is output, including: if , the region is determined to be healthy skin; if , the region is determined to be moderately sensitive; if , the region is determined to be severely sensitive; and according to the sensitive level of the region, the face sensitive level is output, including: if any region is severely sensitive, the face is severely sensitive; otherwise, if any region is moderately sensitive, the face is moderately sensitive; otherwise, it is healthy skin.

[0012] Preferably, the set threshold value is a basic threshold value and a result dynamically corrected by the user's skin quality; The user's skin quality is divided according to the moisture value , including: skin type ; wherein, represents the moisture value of any region of the user's face; for dry regions, then ; for dry regions, then ; for neutral regions, then .

[0013] Preferably, it further includes: S8: Based on the determination result and the sensitivity level, a preset solution library is called; the preset solution library stores nursing solutions corresponding to the sensitivity level, including: Identify the user's facial sensitivity level and retrieve the care plan corresponding to the sensitivity level.

[0014] Preferably, it also includes: S9: Establish a user archive to record the user's initial sensitive skin test results and facial sensitivity level; Track and detect users at equal time intervals, and output real-time judgment results and facial sensitivity levels; Output the change curve based on the real-time judgment results and facial sensitivity level.

[0015] The beneficial effects of the present invention are as follows: 1. The present invention describes a method for automatically identifying sensitive skin by combining skin color and acne distribution. By collecting facial images of the user and adopting a dual-color space parallel processing mode, the red zone value and skin type of each area of ​​the user's face are synchronously output. Then, based on the calculation of the acne density in each facial area, a sensitivity index that can reflect the sensitivity of each area of ​​the user's face is obtained. Among them, the sensitivity index is corrected with the skin correction coefficient corresponding to the skin type, which can reduce missed diagnosis and misdiagnosis caused by skin color differences to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0018] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0019] like Figure 1 As shown, the method for automatically identifying sensitive skin by combining skin color and acne distribution according to an embodiment of the present invention includes the following steps: S1: Collect user facial images through mobile terminals; S2: Identify skin color type based on LAB color space; S3: Use the HSV color model to extract the red zone value of each area of ​​the face and brightness value ; S4: Localizing facial acne areas based on lightweight object detection network; S5: Calculate the acne density of each facial area ; S6: Red zone values ​​based on each region , acne density and skin color type to calculate regional sensitivity index ; S7: According to the sensitivity index of each region and set threshold The comparison result is output as the sensitivity determination result.

[0020] In the existing technology, only a single image processing method is relied upon, such as converting the image only into the HSV color space to obtain the red zone value corresponding to the user's facial image, and analyzing the sensitivity index based on the red zone value. However, in the above process, the influence of skin quality on image processing is ignored, which results in the output sensitivity index possibly not being able to truly reflect the user's actual skin sensitivity status, thereby causing misdiagnosis and missed diagnosis.

[0021] In one embodiment of the present invention, when a user performs sensitive skin detection, a facial image can be captured through a mobile terminal. Specifically, the mobile terminal in one embodiment, taking a mobile phone as an example, the user opens the APP to start the detection, the front camera automatically turns on, and captures a high-definition facial image of the user with a resolution of at least 1080P. The system performs dual color space parallel processing, including HSV space processing and LAB space processing, to extract the red zone value of each facial area respectively. and brightness value , and calculate The value and skin color type are determined. The acne area on the face is located through a lightweight target detection network. The acne density of each area on the face is calculated according to the calculation formula. , and finally according to the red zone value of each area and acne density Calculating the regional sensitivity index , and according to Value correction of sensitivity index of each region , through the sensitivity index of each region and set threshold The comparison result can be used to output the sensitive skin judgment result. It can be understood that in the traditional sensitive skin detection process, the missed diagnosis rate of dark-skinned people is higher than that of light-skinned people. This is because the skin of dark-skinned people is interfered with by melanin, which causes the red zone signal to be interfered with, so compensation processing is required. The dermis of light-skinned people is thinner, and the red zone signal is more dominant, so enhanced detection is required. Based on the above, by performing dual-color space parallel processing, the red zone values ​​of various areas of the user's face are identified. and brightness value In addition, based on Calculate the value, identify the user's skin color, and then calculate the sensitivity index of each area of ​​the user's face , based on facial skin color type, correct the sensitivity index of each area , weaken the red signal interference caused by melanin interference problem, so that the sensitive skin detection accuracy of deep skin color population is improved.

[0022] In an embodiment, the method of identifying skin color type is: Calculate the ratio of lightness value to yellow-blue axis value in LAB color space; According to the skin color classification rule: Type ; According to , determine the skin color type of the user.

[0023] After calculating the skin color type of the user through the LAB color space, according to the identification of the user's skin color type, the corresponding skin color correction coefficient is matched for the user, according to the corresponding skin color adjustment coefficient , the missed diagnosis problem caused by skin color in the subsequent calculation process of sensitive index of each region of the user's face can be corrected, in an embodiment, the selection rule of the skin color correction coefficient is as follows: For type - , ; For type - , ; For type - , ; In an embodiment, in the conversion of the HSV color model: The red value is calculated by the pixel ratio in the 0-10° and 170-180° interval of the H channel; The lightness value is used to exclude light interference, and when , the light is turned on; The face region includes face area, T area and contour area.

[0024] In an embodiment, the acne density calculation includes: Each area uses an independent YOLOv5s model for detection: The face area: input size 416x416, confidence 0.7; The T area: input size 320x320, confidence 0.6; The contour area: input size 256x256, confidence 0.5; ​​According to the formula: ； wherein, is the number of acne in the area ; is the area of the area .

[0025] Based on the above, in order to quantify the sensitive index of each area of the user's face, the acne density of each area of the user's face needs to be calculated respectively, and the following will be explained and described with specific data: Suppose the user's skin data is as follows: Skin type is V type; Face area: moisture value 28, 0.85, 0.40; T area: moisture value 65, 0.60, 0.80; Contour area: moisture value 45, 0.30, 0.20; According to the above, the data of each area of the user's face is identified and recorded, including skin type, moisture value of each area, and acne density calculated for each area.

[0026] In an embodiment, the method for calculating the area sensitive index is: Extracting the red area value and acne density of each area of the face; According to the formula: ; wherein is the red area value weight coefficient, is the acne density weight coefficient, is the skin color correction coefficient; wherein, .

[0027] In an embodiment of the present application, specific scenarios are used to explain: Suppose there are user A and user B, wherein: User A is of West African descent, V type skin, ; User B is a Nordic white person, I type skin, ; First, the basic parameters of user A and user B are extracted: The following parameters are obtained: For the face area: User A: 0.85, 0.40; User B: 0.78, 0.35; The skin color correction coefficient is introduced: User A: ; User B: ; The comparison of the sensitive index calculation is as follows: Under the traditional scheme: ; ; After skin color correction: ; ; According to the traditional scheme, the sensitive index of user A and user B is close, but after skin color correction, the sensitive index of user A and user B is quite different, that is, if the threshold is 0.8, according to the traditional scheme, user A and user B are not sensitive skin, but after skin color correction, user B is sensitive skin, and user A is healthy skin. Through skin color correction, the misdiagnosis and missed diagnosis caused by skin color can be reduced. In addition, the face area data of user A obtained in the previous embodiment is as follows: The skin color type is V type; Face area: moisture value 28, 0.85, 0.40; T area: moisture value 65, 0.60, 0.80; Contour area: moisture value 45, 0.30, 0.20; After calculation, the sensitive index of user A in each area of the face is obtained: ; ; ; According to the above calculation, the sensitivity index of each region of the face of user A is obtained, and according to the obtained sensitivity index, the sensitivity of each region of the face of user A is determined by comparing with the set threshold value , and the sensitivity level of the face of user A is determined according to the size of the sensitivity index and the number of sensitive regions.

[0028] In an embodiment, the method of outputting the sensitivity determination result is: The calculated region sensitivity index is compared with the set threshold value . When , it is determined that the skin of the region is sensitive, and when the number of sensitive regions is , it is determined that the whole face is sensitive. Among them, according to the size of the region sensitivity index , the sensitivity level of the corresponding region is output, including: If , it is determined that the region is healthy skin. If , it is determined that the region is moderately sensitive. If , it is determined that the region is severely sensitive. According to the sensitivity level of the region, the face sensitivity level is output, including: If any region is severely sensitive, the face is severely sensitive. Otherwise, if any region is moderately sensitive, the face is moderately sensitive. Otherwise, it is healthy skin.

[0029] As described above, after calculating the sensitivity index of each region of the face of the user, it is necessary to determine whether each region of the face of the user is sensitive according to the set threshold value . The following continues to describe the face data of user A in the above embodiment: From the above content, it can be known that the sensitivity index of each region of the face of user A is as follows: ; ; ; Suppose the set threshold value , then the sensitivity index of each region of the face of user A is lower than the set threshold value . Thus, it can be determined that each region of the face of user A is healthy skin, and when the face sensitivity level is output according to the region sensitivity level, since there is no sensitive region in the face of user A, the face of user A is healthy skin. However, according to the above embodiment, the sensitivity index of the cheek of user B is: ; Thus, it can be determined that the cheek of user B is moderately sensitive in the region, and if the sensitivity index of the T-zone and the contour zone of user B is lower than the set threshold , it can be determined that the face of user B is moderately sensitive, and accordingly, it is output that the face of user A is healthy skin, and the face of user B is moderately sensitive.

[0030] In an embodiment, the set threshold is a basic threshold and a result dynamically corrected according to the skin quality of the user; The skin quality of the user is distinguished according to the moisture value , including: skin quality type ; wherein, represents the moisture value of any region of the face of the user ; For dry regions, then ; For dry regions, then ; For neutral regions, then .

[0031] In an embodiment, when the user uses the mobile phone APP to perform the facial skin sensitivity test, the skin moisture value is obtained based on spectral analysis, and in continuation of the above embodiment, user B obtains the following data through detection: Cheek region: Based on the skin quality type classification rule, the skin quality type of user B is dry, specifically, the cheek skin quality type is dry. Assuming that the basic threshold ; Then, according to the calculation, for user B, the set threshold corresponding to the cheek region is 0.5. Based on the above content, it can be determined that the sensitivity index of the cheek region of user B obtained through detection is as follows: ; And based on the calculation of the set threshold , the determination result can be output, that is, the cheek region of user B is moderately sensitive.

[0032] In an embodiment, it further includes: S8: calling a preset scheme library based on the determination result and the sensitivity level; the preset scheme library stores a care scheme corresponding to the sensitivity level, including: identifying the facial sensitivity level of the user, and calling the care scheme corresponding to the sensitivity level.

[0033] Based on the above, when it is identified that the face of the user B is moderately sensitive, based on the determination result and the sensitive level, a care scheme related to moderate facial sensitivity is retrieved from the preset scheme library, such as the user B's main sensitive area is the cheek, and the corresponding product can be the Centella Asiatica soothing mask, and for the T zone and the contour area, since the above two areas are not sensitive, more basic care is recommended.

[0034] In an embodiment, further comprising: S9: Establish a user profile library, record the sensitive skin determination result and the facial sensitive level of the user in the first detection; Track and detect the user at equal time intervals, and output real-time determination results and facial sensitive levels; According to the real-time determination result and the facial sensitive level, output the change curve.

[0035] In addition to providing the user with the basic function of detecting the facial sensitive level, according to the facial sensitive level obtained after the user is detected, the long-term tracking and early warning of the facial sensitive level of the user can also be realized in an embodiment of the present application. In an embodiment, if the user B is moderately sensitive in the cheek area in the first detection, a period of 30 days is taken, and after 30 days, the user B is actively reminded to detect again whether the face is sensitive. If the secondary detection shows that the sensitive index of the cheek area of the user B is reduced, it means that the care scheme pushed to the user is effective, then the facial sensitive change of the user is tracked, and according to the detection time node, the sensitive index change curve of each area of the face is output. If the sensitive index of the cheek area of the user B is shown to be increased, and the sensitive index of the cheek area of the user B is still not reduced in the next period, according to the change of the sensitive index, the user needs to seek medical treatment in time.

[0036] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for automatically identifying sensitive skin based on skin color and acne distribution, characterized by: The following steps are involved: S1: Collect user facial images through mobile terminals; S2: Identify skin color type based on LAB color space; S3: Use the HSV color model to extract the red zone value of each area of ​​the face and brightness value ; S4: Localizing facial acne areas based on lightweight object detection network; S5: Calculate the acne density of each facial area ; S6: Red zone values ​​based on each region , acne density and skin color type to calculate regional sensitivity index ; S7: According to the sensitivity index of each region and set threshold The comparison result is output as the sensitivity determination result.

2. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 1, characterized in that: The method for identifying skin color type is: Calculating lightness in LAB color space Value and yellow-blue axis The ratio of values; According to the skin color classification rules: type ; according to , determine the user's skin color type.

3. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 2, characterized in that: In the HSV color model conversion: Red Zone Value Calculated by the pixel ratio of the H channel in the range of 0-10° and 170-180°; Lightness value Used to eliminate light interference, when Start fill light when The facial areas include the face zone, the T zone and the contour zone.

4. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 3, characterized in that: The acne density Calculations include: Each district uses an independent YOLOv5s model for detection: The face area: input size 416×416, confidence level 0.7; The T zone: input size 320×320, confidence level 0.6; The contour area: input size 256×256, confidence level 0.5; According to the formula: ； in, For the region The number of acne; For the region area.

5. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 4, characterized in that: The method for calculating the regional sensitivity index is: Extract the red zone values ​​of each area of ​​the face and acne density ; According to the formula: ； in is the red zone value weight coefficient, is the acne density weight coefficient, is the skin color correction coefficient; where, .

6. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 5, characterized in that: The method for outputting the sensitivity determination result is: The calculated regional sensitivity index and set threshold For comparison: when When the skin of the area is judged to be sensitive, and when the number of sensitive areas , or cheek area is sensitive, then it is judged as all sensitive; Among them, according to the regional sensitivity index The size of the output area is the sensitivity level of the corresponding area, including: like , determined to be regional healthy muscle; like , the area was determined to be moderately sensitive; like , it was determined that the region was severely sensitive; Then, based on the regional sensitivity level, the facial sensitivity level is output, including: If any area is severely sensitive, the face is severely sensitive; Otherwise, if any area is moderately sensitive, the face is moderately sensitive; Otherwise, it is healthy muscle.

7. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 6, characterized in that: The set threshold Base threshold And the result after dynamic correction of user's skin type; The user's skin type is determined by the moisture value Distinguish, including: Skin type ; in, Represents any area of ​​the user's face Moisture value; For dry areas, ; For dry areas, ; For the neutral region, .

8. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 7, characterized in that: Also includes: S8: Based on the determination result and sensitivity level, call the preset solution library; The preset solution library stores nursing solutions corresponding to sensitivity levels, including: Identify the user's facial sensitivity level and retrieve the care plan corresponding to the sensitivity level.

9. The method for automatically identifying sensitive skin based on skin color and acne distribution according to claim 8, characterized in that: Also includes: S9: Establish a user archive to record the user's initial sensitive skin test results and facial sensitivity level; Track and detect users at equal time intervals, and output real-time judgment results and facial sensitivity levels; Output the change curve based on the real-time judgment results and facial sensitivity level.

Images