TWI430776B - Smart video skin test system and method of the same - Google Patents

Description

Intelligent image skin quality detecting system and method

The invention relates to a smart image skin detecting system and method, which is a standardization process by a skin quality detecting system comprising a pre-processing module, a skin disease measuring feature capturing module and a skin quality analyzing module. Complete skin image detection.

In the past three decades, biotechnology products have become a new direction of consumption, which contains unlimited business opportunities and brings vitality to the present and future markets. Cosmetics and skin care products are industries in the biotechnology industry that can bring beauty, hope and pleasure to women (even men).

According to the survey of the top 50 countries in the major international consumer markets, the market sales of makeup maintenance products reached US$257.5 billion in 2008, and maintained an annual growth rate of 3.6%~7.3%, which is a very important industry. After 30 years of development, guidance and cultivation, China has bred a sales market of 130 billion yuan, and its product exports have also increased year by year. Now it has created 1.1 billion US dollars of foreign exchange, bringing beauty to consumers in 150 countries and regions. .

At present, the sales of cosmetics in China are second only to Japan and the United States, and it is the third largest consumer market in the world. In the dazzling makeup and skin care products, how to choose a cosmetic skin care product suitable for personal skin, to achieve the purpose of beauty and maintenance, has become the current The most important topic. Therefore, if the skin quality information can be provided quickly, safely and accurately, it can not only help consumers to select suitable cosmetic products, but also enhance consumers' willingness to purchase to stimulate buying.

In recent years, the advancement of biotechnology has led to the prevalence of skin texture testing. These contact skin type detectors can detect and judge the skin and age of the subject, but contact the skin of the subject through the probe, except for hygiene and infection. In addition to the problem of the disease, patients with severe wounds are also unable to undergo testing, and the testing personnel must also receive special training to be competent.

The biggest disadvantages of the above-mentioned conventional contact skin testers are:

1. Since the skin texture detector uses a probe to obtain a skin image, only partial images can be obtained at a time.

2. The probe needs to be in contact with the skin of the subject. If the subject suffers from serious skin diseases or wounds, the test cannot be performed, and cross-infection is likely to occur due to hygiene problems.

3. Additional costs are required to maintain the cleaning of the probe.

4. Special training is required before the operator uses the skin tester.

5. The subject needs to clean the face before being tested, so as not to cause the instrument to misjudge.

Therefore, if a digital image capture device such as a digital camera can be used instead of the probe to capture the skin image, and then the image processing technology is used to obtain the skin texture feature, the operation process can be simplified, and the health problem of the contact probe can be solved.

The above mentioned structure about the conventional contact skin type tester, although it can achieve the general basic requirements and effects in the skin quality testing process, the safety and health and education training integration ability in practical application and Cost control and other industries In the case of application specificity, there are many shortcomings and deficiencies, and it is impossible to exert more specific industrial applicability.

In summary, due to the above-mentioned shortcomings and shortcomings due to the structure of the conventional contact skin tester, it is necessary to propose specific improvement plans based on the future trend of industrial progress to meet the needs of industrial progress. Further provide more technical options in the industry.

The invention provides a standardized skin image detection system and process, and uses an automatic face detection technology to locate a face from an image, and then separately takes out characteristics based on the characteristics of the skin lesion in various color spaces and light sources. The skin image feature values such as skin spots, wrinkles, and acne are not directly used to evaluate the quality of the skin because the features taken at this stage are feature images. Therefore, the present invention obtains various skin texture problems by quantifying the feature images. The parameters are evaluated, but since the evaluation parameters cannot clearly define the severity of the skin texture problem, the number of clusters suitable for the data of the present invention is found through subtractive clustering, and then the adaptive network fuzzy inference system is used to obtain the attribution function of the severity of the skin texture problem. Finally, the attribution function can be used to distinguish the severity of the skin texture problem, so the present invention can finally obtain an objective skin quality assessment report.

Furthermore, since the skin images of the present invention are taken by standard processes, and the skin images are obtained by a digital camera in a fixed environment and a fixed distance, the shortcomings of the probes to contact with human skin can be improved, and the skin images can be accurately received. The skin images of the previous shots were compared to further understand the improvement of the skin.

Therefore, regardless of the subjective and objective conditions, the intelligent image quality detecting system and method of the present invention have no relevant technology applied to high performance in domestic and foreign patents. The image quality detection processing frame is built, and has the advantage of the irreplaceable technology in the market. It is very suitable for the equipment market such as the intelligent image skin detection system industry, and it is bound to bring the intelligent image skin quality detection system and its equipment. There are great business opportunities in the production and design manufacturing industry related markets.

The invention completes the skin image detection by a standardized process by using a skin quality detecting system including a pre-processing module, a skin disease measuring feature capturing module and a skin quality analyzing module.

In order to achieve the above objects and functions, a smart image skin quality detecting system comprises a pre-processing module, a skin disease measuring feature capturing module and a skin quality analyzing module, and the specific feasible implementation manners are as follows: The front processing module has a color image input interface, a skin color detection interface, a noise removal interface, a skin color sampling interface and a face positioning interface, and is sequentially connected.

A skin disease photographing feature capture module, which has an automatic face region segmentation interface, a color space interface and a skin lesion feature interface, and is sequentially connected, and the automatic face region segmentation interface is pre-linked. The face positioning interface of the module.

A skin quality analysis module is connected to the dermatological feature extraction module, which has a feature quantization interface, a skin analysis interface, a database, an adaptive network fuzzy inference system, and a skin model. The interface and a skin evaluation report interface, the feature quantization interface is linked to the skin texture analysis interface, and the database is coupled to an adaptive network fuzzy inference system, and the skin texture analysis interface and the adaptive network fuzzy inference system are linked to the skin texture. The model interface, the skin model interface is linked to the skin evaluation report interface, and the feature quantization interface is linked to the skin lesion feature interface of the skin lesion feature extraction module.

The color image input interface of the pre-processing module is a fixed mask type face image capture device for color face image shooting.

The skin color detection interface of the pre-processing module is used as a color candidate block screening for color face images.

The noise removal interface of the pre-processing module is learned by the small-sized removal of the color face image according to the type.

The skin color sampling interface of the pre-processing module is a connected component, and is reserved for the largest full-face skin color block of the color facial image.

The skin lesions feature capture module's automatic face segmentation interface, which separates the forehead, the tail, the cheeks, the nose, the mouth and the chin from the color face image to form six skin detection blocks.

The color space interface of the dermatological feature capture module includes a YcbCr color space and a LAB color space.

The dermatological lesion features a skin lesion characteristic interface of the module, which is characterized by skin lesions of acne, spots and wrinkles.

The feature quantization interface of the skin quality analysis module quantifies the extracted skin texture as a feature value.

The skin analysis interface of the skin analysis module is based on pore size, skin tone uniformity, spot size, wrinkle depth, sebum secretion level, wrinkle density, horniness and acne, forming eight skin evaluation criteria.

The adaptive network fuzzy inference system of the skin quality analysis module compares the database to calculate the degree of severity of the skin texture problem of the skin condition evaluation standard, and classifies Forming a skin model interface.

The skin quality assessment report interface of the skin texture analysis module is based on the skin texture model interface for skin quality score evaluation.

A smart image skin quality detecting method comprising the following steps:

A. Color image input: First, the system reads a color face image taken from a fixed environment.

B. Skin Tone Detection: A candidate block that leaves skin tone after skin color detection.

C. Morphological de-hocing: the use of morphology to remove too small impurities.

D. Connect to take the largest block: use the connected component to leave the largest skin color block, that is, find the face skin color block in the face positioning part.

E. Face Positioning: Take out the complete face color block from the original image and complete the face positioning.

F. Automatic face segmentation: In order to automatically locate the skin mass, the skin segment is located according to the facial skin defined by the beautician and the physician. The six blocks are forehead, eye tail, cheek, Areas such as the nose, mouth and chin.

G. Color space performance: Select the YCbCr color space that is not easily affected by light to detect the skin color area.

H. Characteristics of skin lesions: The main purpose is to extract the characteristics of skin lesions from the face block. In addition to the different manifestations of skin lesions in various color spaces, some lesions appear in humans according to clinical medicine. The characteristics of the specific block on the face to improve the accuracy of feature extraction.

I. Feature Quantization: Due to the single eigenvalues found, it is not possible to quantify all of the eigenvalues found as a measure of skin quality to provide a skin texture analysis.

J. Skin Quality Analysis: The quantitative goal is based on the eight important factors that affect the skin quality set by the beautician and the evaluation criteria of the beautician in evaluating the eight major problems.

K. Skin quality comparison: The Adaptive Network based Fuzzy Inference System (ANFIS) is used to calculate the degree of attribution of the severity of these skin problems, that is, to classify the severity of the skin lesions entered. .

L. Skin Quality Assessment Report: Use the degree of attribution of the severity of the skin condition to score the skin and output a rough skin test score.

The specific features and effects of the intelligent image quality detecting system and method of the present invention are as follows:

1. Skin image analysis of skin images using image processing techniques and adaptive neural networks.

2. Acquire a high-resolution skin image with a stable light source photographic environment, and use image processing techniques such as morphology and connected components to take a complete face image from the photo.

3. Through the skin lesion extraction module, use image processing technology to find the image feature value suitable for determining the skin of the face.

4. Image processing technology can be simple image compared to contact skin tester A pick-up device (such as a digital camera) replaces the probe and does not need to touch the skin of the subject.

5. Automatically remove facial features, leaving only the skin color block of the face to reduce the error of the skin image captured by the face.

6. Using spots, wrinkles and acne characteristics to learn the skin texture characteristics, the results are quite similar to the expert judgment, in line with the general perception of spots, wrinkles and acne.

7. The design method of the evaluation parameters is to determine the skin condition of the subject by combining the ratio of the lesion area to the image area of the skin and the color of the skin surface caused by the lesion.

8. Skin quality evaluation parameters, with certain evaluation ability, combined with subtractive clustering in the adaptive neural network fuzzy inference system to find the number of groups suitable for classifying the severity of skin problems, and classify them through the regression function. The severity of the skin problem.

9. According to the test data, it is possible to recommend suitable cosmetic products for the skin condition of consumers, and further record and study the causes of skin aging or lesions, or the skin care products can use these data to prove to consumers. .

10. The dermatologist or beautician can report to the subject appropriate treatment, advice or provide effective medical beauty products.

11. The overall accuracy of the skin lesion extraction module is about 74.8%, and the skin quality analysis module is about 72.7% accurate.

12. Combine all skin texture characteristics and evaluate the skin condition by evaluating the parameter method to obtain an objective skin quality evaluation standard.

(1) ‧‧‧Pre-processing modules

(11)‧‧‧Color image input interface

(12)‧‧‧ Skin Detection Interface

(13) ‧ ‧ noise removal interface

(14) ‧‧‧ skin color sampling interface

(15)‧‧‧Face location interface

(2) ‧ ‧ ‧ skin lesions feature capture module

(21)‧‧‧Automatic face segmentation interface

(211) ‧ ‧ forehead

(212)‧‧‧ Eye tail

(213) ‧ ‧ cheeks

(214)‧‧‧ nose

(215) ‧ ‧ mouth corner

(216)‧‧‧ chin

(22) ‧‧‧Color Space Interface

(23) ‧‧‧Skin lesions interface

(3) ‧ ‧ skin quality analysis module

(31) ‧‧‧Feature Quantization Interface

(32) ‧‧‧ Skin Analysis Interface

(33) ‧ ‧ database

(34) ‧‧‧Adaptive network fuzzy inference system

(35) ‧‧‧ Skin Model Interface

(36) ‧ ‧ skin quality assessment report interface

First: Schematic diagram of the structure of a skin texture detecting system according to an embodiment of the present invention.

Second: A schematic diagram of a partial skin texture detecting system according to an embodiment of the present invention.

Third: A schematic flow chart of a skin quality detecting step according to an embodiment of the present invention.

Fourth: A schematic diagram of segmentation of a human skin area according to an embodiment of the present invention.

Annex I: Partition map of the six major blocks of human skin.

Annex 2: Image analysis of acne.

Annex III: Speckle image analysis chart.

Annex IV: Image analysis of wrinkles.

Please refer to the first and second figures, respectively, which are schematic diagrams of a skin texture detecting system and a partial skin quality detecting system according to an embodiment of the present invention. The present invention is a smart image skin detecting system, comprising a pre-processing module (1) a dermatological feature capture module (2) and a skin analysis module (3), wherein: a pre-processing module (1), with a color image input interface (11), a skin tone a detection interface (12), a noise removal interface (13), a skin color sampling interface (14), and a face positioning interface (15), and sequentially connected; the color image input interface of the preprocessing module (1) (11) is a fixed mask type face image capture device for color face image shooting; the skin color detection interface (12) of the preprocessing module (1) is used as a color skin candidate block for color face image screening. The pre-processing module (1)'s noise removal interface (13) is based on the type of color face image removal of small noise removal; the pre-processing module (1) skin color sampling interface (14) is a connected component The largest full-face skin color block for color face images is reserved.

A skin disease photographing feature capture module (2), which has an automatic face region segmentation interface (21), a color space interface (22), and a skin lesion feature interface (23), and is sequentially connected. In addition, the automatic face segmentation interface (21) is connected to the face location interface of the pre-processing module (1) (15); the automatic face segmentation interface of the skin lesion feature capture module (2) (21) ), from the color face image segmentation forehead (211), eye tail (212), cheek (213), nose (214), mouth angle (215) and chin (216), forming six skin test blocks, such as The second or fourth figure or reference to the annex 1; the color space interface (22) of the skin disease feature capture module (2) includes the YcbCr color space and the LAB color space, as shown in the second figure; The skin lesion feature interface (23) of the feature extraction module (2) is a skin lesion characteristic of acne, spots and wrinkles, as shown in the second figure or the reference attachments - two to four.

A skin analysis module (3) is connected to the dermatological feature extraction module (2), which has a feature quantization interface (31), a skin analysis interface (32), and a database (33). An adaptive network fuzzy inference system (34), a skin model interface (35), and a skin evaluation report interface (36), the feature quantization interface (31) is linked to the skin texture analysis interface (32), The database (33) is coupled to an adaptive network fuzzy inference system (34), and the skin texture analysis interface (32) and the adaptive network fuzzy inference system (34) are both coupled to the skin model interface (35), the skin The quality model interface (35) links the skin quality assessment report interface (36), and the feature quantization interface (31) links the skin lesion feature extraction module (2) to the skin lesion feature interface (23); skin quality analysis The feature quantization interface (31) of the module (3) quantifies the extracted skin texture value; the skin quality analysis interface (32) of the skin quality analysis module (3) is based on the pore size and skin color uniformity. , spots, wrinkles, sebum lines The degree of secretion, wrinkle density, horniness and acne, form eight skin evaluation criteria, such as the second picture; skin quality analysis module (3) adaptive network fuzzy inference system (34), the comparison database (33) Calculate the degree of attribution of the severity of the skin condition of the skin condition assessment criteria, and classify to form the skin model interface (35); the skin quality assessment report interface (36) of the skin texture analysis module (3), According to the skin model interface (35), the skin quality score was judged.

Please refer to the third figure, which is a schematic flowchart of a skin quality detecting step according to an embodiment of the present invention. When the system structure is actually applied according to the first figure, the procedure for testing the skin quality of the subject is as follows (see also the first and second figures):

A. Color image input: First, the system reads a color face image taken from a fixed environment.

B. Skin Tone Detection: A candidate block that leaves skin tone after skin color detection.

C. Morphological de-hocing: the use of morphology to remove too small impurities.

D. Connect to take the largest block: use the connected component to leave the largest skin color block, that is, find the face skin color block in the face positioning part.

E. Face Positioning: Take out the complete face color block from the original image and complete the face positioning.

F. Automatic face segmentation: In order to automatically locate the skin mass, the skin segment is located according to the facial skin quality defined by the beautician and the physician. The forehead (211) and the end of the eye are respectively (212), cheek (213), nose (214), mouth (215) and chin (216), as shown in Figure 4 or refer to Annex 1.

G. Color space performance: choose YCbCr color space that is not easily affected by light To detect skin color areas.

H. Characteristics of skin lesions: The main purpose is to extract the characteristics of skin lesions from the face block. In addition to the different manifestations of skin lesions in various color spaces, some lesions appear in humans according to clinical medicine. The characteristics of the specific block on the face to improve the accuracy of feature extraction.

I. Feature Quantization: Due to the single eigenvalues found, it is not possible to quantify all of the eigenvalues found as a measure of skin quality to provide a skin texture analysis.

J. Skin Quality Analysis: The quantitative goal is based on the eight important factors that affect the skin quality set by the beautician and the evaluation criteria of the beautician in evaluating the eight major problems.

K. Skin quality comparison: The Adaptive Network based Fuzzy Inference System (ANFIS) is used to calculate the degree of attribution of the severity of these skin problems, that is, to classify the severity of the skin lesions entered. .

L. Skin Quality Assessment Report: Use the degree of attribution of the severity of the skin condition to score the skin and output a rough skin test score.

In view of the above, the present invention is directed to an application technology of a smart image skin detecting system and method, and specifically includes a pre-processing module (1), a skin disease photo feature capturing module (2), and a Skin Quality Analysis Module (3) Skin Quality Detection System uses a standardized skin image detection process to locate human faces from images using automatic face detection technology, and then based on the characteristics of skin lesions in various color spaces. And the characteristics under the light source, individual skin image features such as skin spots, wrinkles, and acne Value, through the method of quantifying the feature image, obtain the evaluation parameters of various skin quality problems, and then use the adaptive network fuzzy inference system to obtain the attribution function of the severity of the skin texture problem, and finally use the attribution function to distinguish the serious skin problem. To the extent that an objective skin quality assessment report is obtained, a non-contact skin texture detection mechanism is achieved, and an optimal improvement and design is achieved, which is the most specific improvement of the intelligent image skin detection system and method of the present invention.

(1) ‧‧‧Pre-processing modules

(11)‧‧‧Color image input interface

(12)‧‧‧ Skin Detection Interface

(13) ‧ ‧ noise removal interface

(14) ‧‧‧ skin color sampling interface

(15)‧‧‧Face location interface

(2) ‧ ‧ ‧ skin lesions feature capture module

(21)‧‧‧Automatic face segmentation interface

(22) ‧‧‧Color Space Interface

(23) ‧‧‧Skin lesions interface

(3) ‧ ‧ skin quality analysis module

(31) ‧‧‧Feature Quantization Interface

(32) ‧‧‧ Skin Analysis Interface

(33) ‧ ‧ database

(34) ‧‧‧Adaptive network fuzzy inference system

(35) ‧‧‧ Skin Model Interface

(36) ‧ ‧ skin quality assessment report interface

Claims (12)

A smart image skin detection system includes: a pre-processing module, a color image input interface, a skin color detection interface, a noise removal interface, a skin color sampling interface, and a face positioning interface, and sequentially Linking a skin lesion feature capture module, having an automatic face segmentation interface, a color space interface, and a skin lesion feature interface, and sequentially connecting, and the automatic face region segmentation interface The face positioning interface of the pre-processing module is connected; a skin quality analysis module is connected to the skin disease feature extraction module, and has a feature quantization interface, a skin analysis interface, a database, and a adaptation. a network fuzzy inference system, a skin model interface and a skin evaluation report interface, the feature quantization interface is linked to a skin analysis interface, the database is coupled with an adaptive network fuzzy inference system, and the skin analysis interface and adaptation The sexual network fuzzy inference system is connected to the skin model interface, and the skin model interface is linked to the skin evaluation report interface, and the feature quantization interface is linked to the skin disease lens. Extracting the skin lesion feature interface of the module; wherein the skin texture analysis module adapts the network fuzzy inference system, compares the database to calculate the degree of skin disease severity of the skin quality assessment standard, and classifies Forming a skin model interface. The intelligent image skin quality detecting system according to claim 1, wherein the color image input interface of the pre-processing module is a fixed mask type image capturing device for color face image capturing. The intelligent image skin quality detecting system according to claim 1 or 2, wherein The skin color detection interface of the pre-processing module is used as a color candidate block screening for color face images. For example, the smart image skin quality detecting system described in claim 1 or 2, wherein the pre-processing module has a noise removal interface, and the small-sized noise removal of the color face image is learned according to the type. The smart image skin quality detecting system according to claim 1 or 2, wherein the skin color sampling interface of the preprocessing module is a connecting component, and the largest full face skin color block of the color face image is reserved. The invention relates to the intelligent image skin detecting system according to the first aspect of the patent application, wherein the skin lesion feature capturing module automatically divides the interface of the face region, and divides the forehead, the tail of the eye, and the cheek from the color face image. , nose, mouth and chin, forming six skin test blocks. The smart image skin quality detecting system according to claim 1, wherein the color space interface of the skin disease feature capturing module comprises a YcbCr color space and a LAB color space. The invention relates to the intelligent image skin quality detecting system according to the first aspect of the patent application, wherein the skin lesion characteristic extracting module has a skin lesion characteristic interface, and is characterized by skin lesions of acne, spots and wrinkles. The smart image skin quality detecting system according to claim 1, wherein the skin coloring module has a feature quantization interface that quantifies the extracted skin texture characteristic value. The invention relates to the intelligent image skin quality detecting system described in claim 1, wherein the skin texture analysis interface of the skin quality analyzing module is based on pore size, skin color uniformity, spot number, wrinkle depth, and sebum line secretion degree. Wrinkle density, horniness and acne, forming eight skin evaluation criteria. The intelligent image skin quality detecting system according to claim 1, wherein the The skin quality assessment report interface of the skin quality analysis module is based on the skin texture model interface for skin quality score evaluation. A smart image skin quality detecting method comprising the following steps: A. color image input: a color face image taken from a fixed environment is read by the system; B. skin color detection: after skin color detection is left Candidate block of lower skin color; C. Morphological de-noise: use type to remove too small noise; D. connect to take the largest block: use the connected component to leave the largest skin color block, that is, in the face positioning part Find the face color block; E. Face positioning: take the complete face color block from the original image to complete the face positioning; F. Automatic face segmentation: according to the beautician and physician defined face Quality detection of six blocks, automatically locate the forehead, eye tail, cheeks, nose, mouth and chin and other skin areas; G. color space performance: select the YCbCr color space that is not susceptible to light to detect skin color areas; H. Characteristics of skin lesions: the characteristics of skin lesions taken from the face of the face, the characteristics of the skin lesions in various color spaces and the characteristics of the specific blocks on the face; I. Feature quantification: for all skins quality The values were quantified to provide skin texture analysis; J. Skin quality analysis: analysis based on eight important factors affecting skin quality and evaluation criteria; K. Skin quality comparison: using adaptive neural network blur Infer the clustering of subtractive systems to find the number of groups suitable for classifying the severity of skin problems, and obtain them through the regression function. Classify the severity of the skin problem and classify the severity of the skin type problem; L. Skin Quality Assessment Report: Skin score using the degree of severity of the skin problem, and output a final rough skin test score .