CN113679355A - Skin complexion measuring device and method for intense pulse light xerophthalmia treatment - Google Patents

Skin complexion measuring device and method for intense pulse light xerophthalmia treatment Download PDF

Info

Publication number
CN113679355A
CN113679355A CN202111092569.1A CN202111092569A CN113679355A CN 113679355 A CN113679355 A CN 113679355A CN 202111092569 A CN202111092569 A CN 202111092569A CN 113679355 A CN113679355 A CN 113679355A
Authority
CN
China
Prior art keywords
module
data
skin
class
dry eye
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111092569.1A
Other languages
Chinese (zh)
Inventor
李�杰
邱选兵
李传亮
李昌
尚建平
秦少春
陈利强
王美婷
程莎莎
李春生
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanxi Ruihao Biotechnology Co ltd
Original Assignee
Shanxi Ruihao Biotechnology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanxi Ruihao Biotechnology Co ltd filed Critical Shanxi Ruihao Biotechnology Co ltd
Priority to CN202111092569.1A priority Critical patent/CN113679355A/en
Publication of CN113679355A publication Critical patent/CN113679355A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/44—Detecting, measuring or recording for evaluating the integumentary system, e.g. skin, hair or nails
    • A61B5/441—Skin evaluation, e.g. for skin disorder diagnosis
    • A61B5/443—Evaluating skin constituents, e.g. elastin, melanin, water
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
    • A61B5/0075—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence by spectroscopy, i.e. measuring spectra, e.g. Raman spectroscopy, infrared absorption spectroscopy
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
    • A61B5/0077—Devices for viewing the surface of the body, e.g. camera, magnifying lens
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Veterinary Medicine (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Artificial Intelligence (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Dermatology (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The invention relates to the technical field of skin detection, in particular to a skin complexion measuring device and method for intense pulse light dry eye treatment. The invention has reasonable structural design, firstly, an LED supplementary lighting instruction is sent out at the touch screen end of the display touch screen module, the annular LED supplementary lighting lamp module is controlled to emit light to irradiate the surface of the skin of human eyes, then, the multispectral digital sensor module is used for collecting the spectral data characteristics of a plurality of wave bands of the light reflected by the skin, finally, the data is input into a trained lightweight convolution neural network model, and a multi-classification support vector machine classifier is used for predicting and classifying the skin color to obtain the result of identifying the skin color.

Description

Skin complexion measuring device and method for intense pulse light xerophthalmia treatment
Technical Field
The invention relates to the technical field of skin detection, in particular to a skin complexion measuring device and method for intense pulse light dry eye treatment.
Background
In skin color skin division, the invention is used for carrying out skin color division on the basis of Fitzpatrick model skin color classification, and the Fitzpatrick model is divided into six different types of skin colors. Type I: white skin is not sunburn and is always easily burned; type II: white skin, light tan, always easy to burn; type III: light brown skin, moderate burn; and IV, type: medium brown skin, tan, slight burn; type V: dark brown skin, heavy tanning, few burns; type VI: skin melanin is deposited and never burned.
The multispectral sensing technology integrates a high-precision optical filter on standard CMOS silicon through a nano optical deposition interference filter technology, and the main system structure comprises an optical part and a control display part. An optical path including imaging optics and spectrally separated optics. The housing of the sensor houses an entrance aperture for receiving the measured radiation. In the radiation path behind the entrance aperture, the optical device splits the beam of human radiation into a number of different spectral transmission range filters. And a radiation sensitive element is arranged behind the optical filter, and a detection signal is transmitted through a sensor sensitive element. Identification acquisition including spectral data may be implemented, among other things.
The lightweight convolutional neural network is a feedforward deep network model with a small quantity of parameter models and containing convolutional nerves, can reduce the dimension of a large number of data pictures into a small data quantity and simultaneously reserve picture features, and can effectively extract shallow and deep features of the data.
In the aspect of skin color identification, the traditional skin color identification mainly adopts color rule judgment or classifier training, the problems that human factors subjectively extract features and deep features are difficult to extract and the identification accuracy is not high enough exist, and therefore the skin color measurement device and the method for the intense pulse light dry eye treatment are provided.
Disclosure of Invention
The invention aims to overcome the problems in the prior art and provides a skin complexion measuring device and method for intense pulse light dry eye treatment.
In order to achieve the technical purpose and achieve the technical effect, the invention is realized by the following technical scheme:
the utility model provides a skin complexion measuring device for intense pulsed light xerophthalmia treatment, includes handle grip module, the top of handle grip module is provided with the shell, the front side of shell is provided with the external component of surveying, the rear side of shell is provided with the display touch screen module, the inside of shell be provided with the external component matched with modular component of surveying.
Preferably, a top cover is arranged at the top of the shell, and the top cover is connected with the shell through a tenon structure.
Based on the technical characteristics, the shell is convenient to detach, and the detailed construction and maintenance in the shell are checked.
Preferably, survey the subassembly outward and include annular LED light filling lamp subassembly module, incident aperture module, power indicator module and operation pilot lamp module, the incident aperture module is located the inboard of annular LED light filling lamp subassembly module, power indicator module and operation pilot lamp module are located the below of annular LED light filling lamp subassembly module, and power indicator module is located the left side of operation pilot lamp module.
Based on the technical characteristics, the power indicator light module and the operation indicator light module are mainly used for prompting whether the operation is normal or not.
Preferably, the LED fill light module adopts 16 white light LED lamps.
Based on the technical characteristics, the annular LED light supplementing lamp component module is used for supplementing light to the skin.
Preferably, the module assembly comprises a power module, an MCU control module, a multispectral digital sensor module and an LED lamp panel module.
Based on above-mentioned technical characteristics, LED lamp plate module mainly plays the effect of shining for shine people's eye skin, carry out the light filling for skin.
Preferably, the bottom of the handle grip module is provided with a 5-core aviation socket, and the interfaces of the five-core aviation socket are a power supply 12V interface, a power supply GND interface, a protection grounding interface, an I2C SCL interface and an I2C SDA interface respectively.
Based on the above technical features, the I2C SCL interface and the I2C SDA interface are used to transmit data to and exchange data with the device.
A skin complexion measuring method for intense pulse light dry eye treatment comprises the following specific steps:
after the device is started and powered on, the display touch screen module firstly controls the annular LED light supplementing lamp assembly module to emit light to irradiate the skin to be detected through the MCU control module;
detecting the light after emitting the skin through a multi-spectral digital sensor module scintillation channel, and then acquiring multi-spectral data through a multi-channel sensor;
step three, after the multispectral data is collected, storing the transmission data into an RAM memory;
step four, transmitting the data processing of the step three to an MCU control module through a transmission protocol;
and fifthly, performing principal component analysis and preprocessing on the data, inputting the data into a trained convolutional neural network recognition model configured in the MCU control module to perform skin color feature extraction, finally obtaining a skin color recognition result by using multi-classification SVM classification, and displaying the transmitted data and the spectrum data recognition skin color result on the display touch screen module.
Preferably, before identifying the skin color, performing principal component analysis and dimension reduction processing on the collected multispectral data; performing dimensionality reduction on the acquired spectral data, wherein the specific calculation steps are as follows:
step 1: firstly, standardizing multispectral original data, unifying scalar quantities to a fixed range value, facilitating convergence training and calculation of a model, and ensuring that each collected spectral data can be extracted equivalently; the calculation formula is as follows:
Figure BDA0003268089020000041
step 2: and (3) calculating a covariance matrix of the spectral data, wherein the calculation formula is as follows:
Figure BDA0003268089020000042
Figure BDA0003268089020000043
wherein x isi,xjRepresenting skin color spectrum data corresponding to the ith class and the jth class, wherein A is a covariance matrix;
and step 3: parameters for calculating the covariance of the spectral data: eigenvectors and eigenvalues; the calculation formula is as follows:
Figure BDA0003268089020000044
A＝QλQ-1
wherein A is a covariance matrix, Q is a characteristic vector, and lambda is a characteristic value;
and 4, step 4: selecting principal component eigenvectors according to the sequence of the eigenvalues from big to small to form a new matrix;
and 5: carrying out mapping projection on the spectrum data, mapping the spectrum sample data onto the characteristic vector, namely finding a new set of orthogonal bases { u }1,u2,...,ukAnd after the spectral data are projected on a plane formed by orthogonal bases, the distance and the variance J between the datajThe maximum is reached, namely:
Figure BDA0003268089020000045
wherein m is the number of spectrum samples, and the orthogonal base is ujData point at xiA projected distance on the substrate of
Figure BDA0003268089020000046
Preferably, the measuring method is mainly embodied in the aspect of a convolutional neural network model, for a lightweight convolutional neural network, a deformable convolutional neural network is added behind a convolutional layer to enhance the extraction of the characteristic quantity, and in the aspect of classification, a multi-classification SVM classifier pairwise classification method is adopted to classify skin color results.
Preferably, the multi-classification SVM classifier classified in pairs identifies and classifies by constructing a binary SVM between every two skin color spectrum data classes; then, for classifying the ith and jth spectrum data, a binary SVM is trained, i.e. the quadratic programming problem needs to be solved:
Figure BDA0003268089020000051
Figure BDA0003268089020000052
Figure BDA0003268089020000053
in the formula, phi represents the nonlinear mapping of input and feature space, subscript t represents the union sample serial number of the ith class and the jth class, and i and j represent the parameters of the ith class and the jth class respectively; the decision function between class i and class j is therefore:
Figure BDA0003268089020000054
and classifying the multispectral new data by adopting a voting mechanism: each binary SVM processes the new data X according to its decision functionnewThere is a prediction if X is pairednewThe prediction of (1) is class i, and the number of tickets belonging to class i is increased by 1; otherwise, increasing the number of the j tickets by 1, and finally, obtaining the class with the most tickets, namely XnewThe result of the recognition of (1).
Compared with the prior art, the invention has the following beneficial effects:
the invention has reasonable structural design, and in the aspect of the device: the design structure is light, small, simple and portable; in terms of recognition accuracy: the skin color identification result obtained by combining skin color spectral characteristic data obtained by adopting the multispectral digital sensor module and the back of the deep learning model is more accurate.
Of course, it is not necessary for any product in which the invention is practiced to achieve all of the above-described advantages at the same time.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a block diagram of the overall structure of the skin tone measuring device for intense pulse light dry eye treatment of the present invention;
FIG. 2 is a schematic diagram of a touch screen interface module of the skin tone measurement device for intense pulse light dry eye treatment of the present invention;
FIG. 3 is a schematic structural view of a 5-core aerial socket at the bottom of the skin complexion measuring device for intense pulse light dry eye treatment of the present invention;
FIG. 4 is a schematic view of the internal structure of the skin complexion measuring device for intense pulse light dry eye treatment of the present invention;
FIG. 5 is a measurement flow diagram of the skin complexion measuring device for intense pulse light dry eye treatment of the present invention;
FIG. 6 is a flow chart of the data recognition process of the skin complexion measuring device for intense pulse light dry eye treatment of the present invention;
FIG. 7 is a flow chart of the Principal Component Analysis (PCA) processing of the skin complexion measuring device for intense pulse light dry eye treatment of the present invention;
FIG. 8 is a detection principle block diagram of the skin complexion measuring device for intense pulse light dry eye treatment of the present invention;
FIG. 9 is a schematic diagram of a skin tone measurement algorithm for the skin tone measurement device for intense pulse light dry eye treatment of the present invention;
FIG. 10 is a diagram of the MobileNet V3 network plus Deformable Convolution (DCN) architecture of the present invention;
FIG. 11 is a block diagram of the inner structure Bneck in combination with the Deformable Convolution (DCN) of MobileNet V3 according to the present invention;
FIG. 12 is a graph of the pairwise classification results of the Support Vector Machine (SVM) classifier of the present invention;
in the drawings, the components represented by the respective reference numerals are listed below:
the LED light supplement lamp comprises a top cover 1, a 2-annular LED light supplement lamp component module, a 3-incident aperture module, a 4-power indicator lamp module, a 5-operation indicator lamp module, a 6-handle grip module, a 7-touch screen interface module, an 8-power 12V interface, a 9-power GND interface, a 10-protection grounding interface, an 11-I2C SCL interface, a 12-I2C SDA interface, a 13-power module, a 14-MCU control module, a 15-multispectral digital sensor module and a 16-LED lamp panel module.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
A skin complexion measuring device for intense pulsed light xerophthalmia treatment comprises a handle grip module 6, which is a device convenient for a user to operate and use, wherein the top of the handle grip module 6 is provided with a shell, the top of the shell is provided with a top cover 1, and the top cover 1 is connected with the shell by adopting a tenon structure, so that the device is convenient to disassemble, and the detailed construction and maintenance in the shell are checked;
an external measurement component is arranged on the front side of the shell and comprises an annular LED light supplement lamp component module 2, an incident aperture module 3, a power indicator lamp module 4 and an operation indicator lamp module 5, the incident aperture module 3 is positioned on the inner side of the annular LED light supplement lamp component module 2, the power indicator lamp module 4 and the operation indicator lamp module 5 are positioned below the annular LED light supplement lamp component module 2, the power indicator light module 4 is positioned on the left side of the operation indicator light module 5, the annular LED light supplement light module 2 is used for performing light supplement operation on skin, the incident aperture module 3 is mainly used for receiving detected radiation and detecting light reflected by human skin, and the power indicator light module 4 and the operation indicator light module 5 are mainly used for prompting whether the operation is normal or not (see figure 1 in the attached drawing of the specification);
the inside of the shell is provided with a module component matched with the external measurement component, the rear side of the shell is provided with a display touch screen module 7, the module component comprises a power supply module 13, an MCU control module 14, a multispectral digital sensor module 15 and an LED lamp panel module 16, the power supply module 13 is used for supplying power to the LED lamp panel module 16, the multispectral digital sensor module 15 and the MCU control module 14 to ensure the stability of the power consumption of the device, the multispectral digital sensor module 15 collects light reflected by skin through the multispectral digital sensor, the light enters the spectrum sensor after being divided into different sensors to obtain spectrum data of different wave bands, further the collection and extraction of the multispectral data reflected by the skin are realized, the purpose of collecting skin color data is achieved, the skin color characteristic quantity of the skin color is more various and effective, a trained convolutional neural network identification model is configured in the MCU control module 14, the method mainly comprises the steps of identifying and detecting collected multispectral data, obtaining a display result on a display touch screen module 7, further performing overall skin color detection control through the display touch screen module 7, enabling an LED lamp panel module 16 to play a role of irradiation and be used for irradiating human skin and supplementing light for the skin, enabling the size of a screen of the display touch screen module 7 to be 5.5 inches, enabling a user to complete control of a supplementary lighting lamp, collection of multispectral data and identification of skin color through operation of the display touch screen module 7 in a key-pressing mode, controlling skin color detection through the display touch screen, and finally displaying the skin color type result identified and detected on the touch display screen (see an upper graph of a figure 2 and a figure 4 in an instruction book attached diagram);
the bottom of the handle grip module 6 is provided with a 5-core aviation socket, the interfaces of the five-core aviation socket are a power supply 12V interface 8, a power supply GND interface 9, a protection grounding interface 10, an I2C SCL interface 11 and an I2C SDA interface 12 respectively, the power supply 12V interface 8 and the power supply GND interface 9 are used for connecting and supplying power to the device, and the I2C SCL interface 11 and the I2C SDA interface 12 are used for transmitting data to the device and exchanging the data (refer to fig. 3 in the attached drawings of the specification).
A skin complexion measuring method for intense pulse light dry eye treatment (as shown in a measuring flow chart of figure 5) comprises the following specific steps:
firstly, after the device is started and powered on, the display touch screen module 7 controls the annular LED supplementary lighting lamp assembly module 2 to emit light to irradiate the skin to be detected through the MCU control module 14;
secondly, the light after emitting the skin is detected through a scintillation channel of the multi-spectral digital sensor module 15, and then multi-spectral data is collected through a multi-channel sensor;
step three, after the multispectral data is collected, storing the transmission data into an RAM memory;
step four, transmitting the data processing of the step three to the MCU control module 14 through a transmission protocol;
and step five, as shown in the data identification processing flow of fig. 6, preprocessing the data by Principal Component Analysis (PCA), inputting the preprocessed data into a trained convolutional neural network identification model configured in the MCU control module 14 to extract skin color features, classifying the skin color features by using a multi-classification Support Vector Machine (SVM), and displaying the skin color results of the data and the spectrum data transmitted on the display touch screen module 7.
Specifically, before identifying skin color, performing Principal Component Analysis (PCA) dimensionality reduction processing on the collected multispectral data; the collected spectral data is subjected to dimensionality reduction treatment, and the specific calculation steps (shown in a flow chart in fig. 7) are as follows:
step 1: firstly, standardizing multispectral original data, unifying scalar quantities to a fixed range value, facilitating convergence training and calculation of a model, and ensuring that each collected spectral data can be extracted equivalently; the calculation formula is as follows:
Figure BDA0003268089020000091
step 2: and (3) calculating a covariance matrix of the spectral data, wherein the calculation formula is as follows:
Figure BDA0003268089020000092
Figure BDA0003268089020000093
wherein x isi,xjIndicates that the ith class corresponds to the jth classThe skin color spectrum data A is a covariance matrix;
and step 3: parameters for calculating the covariance of the spectral data: eigenvectors and eigenvalues; the calculation formula is as follows:
Figure BDA0003268089020000101
A＝QλQ-1
wherein A is a covariance matrix, Q is a characteristic vector, and lambda is a characteristic value;
and 4, step 4: selecting principal component eigenvectors according to the sequence of the eigenvalues from big to small to form a new matrix;
and 5: carrying out mapping projection on the spectrum data, mapping the spectrum sample data onto the characteristic vector, namely finding a new set of orthogonal bases { u }1,u2,...,ukAnd after the spectral data are projected on a plane formed by orthogonal bases, the distance and the variance J between the datajThe maximum is reached, namely:
Figure BDA0003268089020000102
wherein m is the number of spectrum samples, and the orthogonal base is ujData point at xiA projected distance on the substrate of
Figure BDA0003268089020000103
In the technical scheme, the measuring method is mainly embodied in the aspect of a convolutional neural network model, a deformable convolutional neural network is added behind a convolutional layer for a lightweight convolutional neural network to enhance the extraction of the characteristic quantity, and in the aspect of classification, a multi-classification SVM classifier is adopted for classifying skin color results in a pairwise classification method.
In the technical scheme, the multi-classification SVM classifier classified in pairs identifies and classifies by respectively constructing a dichotomous SVM between every two skin color spectrum data classes; then, for classifying the ith and jth spectrum data, a binary SVM is trained, i.e. the quadratic programming problem needs to be solved:
Figure BDA0003268089020000104
Figure BDA0003268089020000105
Figure BDA0003268089020000106
in the formula, phi represents the nonlinear mapping of input and feature space, subscript t represents the union sample serial number of the ith class and the jth class, and i and j represent the parameters of the ith class and the jth class respectively; the decision function between class i and class j is therefore:
Figure BDA0003268089020000111
and classifying the multispectral new data by adopting a voting mechanism: each binary SVM processes the new data X according to its decision functionnewThere is a prediction (vote) if X is pairednewThe prediction of (1) is class i, and the number of tickets belonging to class i is increased by 1; otherwise, increasing the number of the j tickets by 1, and finally, obtaining the class with the most tickets, namely XnewThe result of the recognition of (1).
Example 2
Based on embodiment 1, the present invention provides another technical solution: AS shown in a detection principle framework diagram of a skin color measuring device in fig. 8, the data acquisition module comprises an MCU control module, an LED lamp panel module and a multispectral filter sensor module. The MCU control module firstly controls the LDE lamp panel to emit light to irradiate the surface of human skin color, then the light emitted by the human skin color is detected and received by the multispectral digital sensor, and spectral data of different wave bands are formed by sensors of different wave bands, wherein the MCU control module comprises 10 data acquisition channels: f1(405-425nm), F2(435-455nm), F3(470-490nm), F4(505-525nm), F5(545-565nm), F6(580-600nm), F7(620-640nm), F8(670-690nm), NIR (850-1050nm) and no filter channel.
The skin complexion measuring device for intense pulse light dry eye treatment of the embodiment collects spectral data of 10 channels through the multispectral digital sensor module AS7341 when collecting images of skin, samples the data of the skin, and stores and transmits the sampled skin data through data storage. The multispectral digital sensor module for collecting data is not limited to this, and may also include other multi-channel spectral digital sensors.
Example 3
Based on embodiment 1, the present invention provides another technical solution: the invention discloses a skin complexion measuring device and method for intense pulse light dry eye treatment, which are used for identifying the skin complexion of dry eye skin of people. On the basis of the existing lightweight MobileNet V3, a deformable convolution structure and a pair-classified SVM classifier are combined to train the model. Skin classification adopts Fitzpatrick model skin classification, totally classified into type I-type VI, the embodiment mainly collects yellow skin color, namely type IV-type V skin color corresponding to Fitzpatrick, classifies the yellow skin color into 6 types according to the characteristics of the yellow skin color, collects thousands of skin spectrum data by using a skin color measuring device, then preprocesses the data, carries the data into a model for training, extracts the shallow layer characteristic and the deep layer characteristic of the skin color, gradually optimizes the model parameters, and finally obtains higher skin color identification accuracy.
For skin color grading selection, the classification of the skin of yellow race selected by the embodiment into 6 classes is limited to the embodiment, and other skin color classes can be further classified.
As shown in the processing flow of fig. 9, for the detected skin color data, the collected multispectral data is subjected to Principal Component Analysis (PCA) dimensionality reduction; performing dimensionality reduction on the acquired spectral data, wherein the specific calculation steps are as follows:
step 1: firstly, standardizing multispectral original data, unifying scalar quantities to a fixed range value, facilitating convergence training and calculation of a model, and ensuring that each collected spectral data can be extracted equivalently; the calculation formula is as follows:
Figure BDA0003268089020000121
for each spectral data, the original value is subtracted first by the average of the data
Figure BDA0003268089020000122
Then, the standard deviation σ is divided to obtain normalized data.
Step 2: and (3) calculating a covariance matrix of the spectral data, wherein the calculation formula is as follows:
Figure BDA0003268089020000123
Figure BDA0003268089020000131
wherein x isi,xjThe skin color spectrum data corresponding to the ith class and the jth class are shown, and A is a covariance matrix.
And step 3: parameters for calculating data covariance: eigenvectors and eigenvalues. The calculation formula is as follows:
Figure BDA0003268089020000132
A＝QλQ-1
wherein A is a covariance matrix, Q is an eigenvector, and lambda is an eigenvalue.
And 4, step 4: and selecting principal component eigenvectors according to the sequence of the eigenvalues from big to small to form a new matrix.
And 5: performing mapping projection of data, and mapping the sample data toOn the feature vector, a new set of orthogonal bases u is found1,u2,...,ukAnd after the spectral data are projected on a plane formed by orthogonal bases, the distance and the variance J between the datajThe maximum is reached, namely:
Figure BDA0003268089020000133
wherein m is the number of spectrum samples, and the orthogonal base is ujData point at xiA projected distance on the substrate of
Figure BDA0003268089020000134
After the principal component analysis and pretreatment is finished, the data is brought into a trained lightweight convolution neural network model for prediction, and finally an output skin color recognition result is obtained.
In the skin color model, a model formed by adding a Deformable Convolution (DCN) combination on the basis of MobileNet V3 is adopted to train and predict a skin color result, and finally a Support Vector Machine (SVM) classifier classified in pairs is used for classification, wherein the MobileNet V3 structure adopts a large parameter structure and comprises a convolution layer, an inverted residual block, an average pooling layer, a full-link layer, a flattening layer, a classifier and the like, and is specifically shown in FIG. 10.
In the lightweight MobileNet V3 recognition model, as shown in fig. 11, based on the existing MobileNet V3, a deformable convolutional network structure is combined, a deformable convolutional network framework is added behind the convolutional layer of the first layer, and a deformable convolutional neural network is added in the inverted residual error block.
And finally classifying the skin color model, and identifying the skin color by adopting a SVM classifier method of paired classification. The specific algorithm process is as follows:
the SVM classifier based on paired classification is used for identifying and classifying by respectively constructing a dichotomous SVM between every two skin color spectrum data classes. Then, for classifying the ith and jth spectrum data, a binary SVM is trained, i.e. the quadratic programming problem needs to be solved:
Figure BDA0003268089020000141
Figure BDA0003268089020000142
Figure BDA0003268089020000143
in the formula, phi represents the nonlinear mapping of the input and the feature space, the subscript t represents the union sample number of the ith class and the jth class, and i and j represent the parameters of the ith class and the jth class respectively. The decision function between class i and class j is therefore:
Figure BDA0003268089020000144
and classifying the multispectral new data by adopting a voting mechanism: each binary SVM processes the new data X according to its decision functionnewThere is a prediction (vote) if X is pairednewThe prediction of (1) is class i, and the number of tickets belonging to class i is increased by 1; otherwise, increasing the number of the j tickets by 1, and finally, obtaining the class with the most tickets, namely XnewThe result of the classification is shown in fig. 12.
All the descriptions presented above are of the basic principles, principal features and advantages of the invention. The invention is not limited by the above embodiments, and the embodiments described in the above description are only preferred embodiments of the invention, so that the specific contents of the invention can be clearly illustrated and presented, and the invention is not limited by them. Without departing from the spirit and scope of the present invention, there may be some variations and modifications which may follow, and these are intended to be included within the scope of the present invention. The scope of the invention herein claimed is defined by the appended claims and equivalents thereof.

Claims (10)

1. A skin complexion measuring device for intense pulse light dry eye treatment, includes handle grip module (6), its characterized in that: the handle comprises a handle grip module (6) and is characterized in that a shell is arranged at the top of the handle grip module (6), an external measurement component is arranged on the front side of the shell, a display touch screen module (7) is arranged on the rear side of the shell, and a module component matched with the external measurement component is arranged in the shell.
2. A skin complexion measuring device for intense pulse light dry eye treatment according to claim 1, characterized in that: the top of shell is provided with top cap (1), adopt the tenon structure to be connected between top cap (1) and the shell.
3. A skin complexion measuring device for intense pulse light dry eye treatment according to claim 1, characterized in that: survey the subassembly outward and include annular LED light filling banks spare module (2), incident aperture module (3), power indicator module (4) and operation pilot lamp module (5), incident aperture module (3) are located the inboard of annular LED light filling banks spare module (2), power indicator module (4) and operation pilot lamp module (5) are located the below of annular LED light filling banks spare module (2), and power indicator module (4) are located the left side of operation pilot lamp module (5).
4. A skin complexion measuring device for intense pulse light dry eye treatment according to claim 3, characterized in that: the LED light supplementing lamp assembly module (2) adopts white light LED lamps, and the number of the white light LED lamps is 16.
5. A skin complexion measuring device for intense pulse light dry eye treatment according to claim 1, characterized in that: the module assembly comprises a power supply module (13), an MCU control module (14), a multispectral digital sensor module (15) and an LED lamp panel module (16).
6. A skin complexion measuring device for intense pulse light dry eye treatment according to claim 1, characterized in that: the bottom of handle grip module (6) is provided with 5 core aviation sockets, the interface of five core aviation sockets is power 12V interface (8), power GND interface (9), protection ground connection interface (10), I2C SCL interface (11) and I2C SDA interface (12) respectively.
7. A method of skin complexion measurement for intense pulse light dry eye treatment, characterized by: the skin complexion measuring method for the intense pulse light dry eye treatment comprises the following specific steps:
firstly, after the device is started and powered on, the display touch screen module (7) controls the annular LED supplementary lighting lamp assembly module (2) to emit light to irradiate the skin to be detected through the MCU control module (14);
secondly, the light after emitting the skin is detected through a scintillation channel of a multi-spectral digital sensor module (15), and then multi-spectral data is acquired through a multi-channel sensor;
step three, after the multispectral data is collected, storing the transmission data into an RAM memory;
step four, the data processing in the step three is transmitted to an MCU control module (14) through a transmission protocol;
and fifthly, performing principal component analysis and preprocessing on the data, inputting the data into a trained convolutional neural network recognition model configured in the MCU control module (14) to perform skin color feature extraction, finally obtaining a skin color recognition result by using multi-classification SVM classification, and displaying the transmitted data and the spectrum data recognition skin color result on the display touch screen module (7).
8. A skin complexion measuring method for intense pulse light dry eye treatment according to claim 7, characterized in that: before identifying skin color, performing principal component analysis and dimension reduction processing on the collected multispectral data; performing dimensionality reduction on the acquired spectral data, wherein the specific calculation steps are as follows:
step 1: firstly, standardizing multispectral original data, unifying scalar quantities to a fixed range value, facilitating convergence training and calculation of a model, and ensuring that each collected spectral data can be extracted equivalently; the calculation formula is as follows:
Figure FDA0003268089010000021
step 2: and (3) calculating a covariance matrix of the spectral data, wherein the calculation formula is as follows:
Figure FDA0003268089010000022
Figure FDA0003268089010000023
wherein x isi,xjRepresenting skin color spectrum data corresponding to the ith class and the jth class, wherein A is a covariance matrix;
and step 3: parameters for calculating the covariance of the spectral data: eigenvectors and eigenvalues; the calculation formula is as follows:
Figure FDA0003268089010000031
A＝QλQ-1
wherein A is a covariance matrix, Q is a characteristic vector, and lambda is a characteristic value;
and 4, step 4: selecting principal component eigenvectors according to the sequence of the eigenvalues from big to small to form a new matrix;
and 5: carrying out mapping projection on the spectrum data, mapping the spectrum sample data onto the characteristic vector, namely finding a new set of orthogonal bases { u }1,u2,...,ukAnd after the spectral data are projected on a plane formed by orthogonal bases, the distance and the variance J between the datajThe maximum is reached, namely:
Figure FDA0003268089010000032
wherein m is the number of spectrum samples, and the orthogonal base is ujData point at xiA projected distance on the substrate of
Figure FDA0003268089010000033
9. A skin complexion measuring method for intense pulse light dry eye treatment according to claim 7, characterized in that: the measuring method is mainly embodied in the aspect of a convolutional neural network model, for a lightweight convolutional neural network, a deformable convolutional neural network is added behind a convolutional layer to enhance the extraction of the characteristic quantity, and in the aspect of classification, a multi-classification SVM classifier pairwise classification method is adopted to classify skin color results.
10. A skin complexion measuring method for intense pulse light dry eye treatment according to claim 9, characterized in that: the multi-classification SVM classifier with paired classification is used for identifying and classifying by respectively constructing a dichotomous SVM between every two skin color spectrum data classes; then, for classifying the ith and jth spectrum data, a binary SVM is trained, i.e. the quadratic programming problem needs to be solved:
Figure FDA0003268089010000041
Figure FDA0003268089010000042
Figure FDA0003268089010000043
in the formula, phi represents the nonlinear mapping of input and feature space, subscript t represents the union sample serial number of the ith class and the jth class, and i and j represent the parameters of the ith class and the jth class respectively; the decision function between class i and class j is therefore:
Figure FDA0003268089010000044
and classifying the multispectral new data by adopting a voting mechanism: each binary SVM processes the new data X according to its decision functionnewThere is a prediction if X is pairednewThe prediction of (1) is class i, and the number of tickets belonging to class i is increased by 1; otherwise, increasing the number of the j tickets by 1, and finally, obtaining the class with the most tickets, namely XnewThe result of the recognition of (1).
CN202111092569.1A 2021-09-17 2021-09-17 Skin complexion measuring device and method for intense pulse light xerophthalmia treatment Pending CN113679355A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111092569.1A CN113679355A (en) 2021-09-17 2021-09-17 Skin complexion measuring device and method for intense pulse light xerophthalmia treatment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111092569.1A CN113679355A (en) 2021-09-17 2021-09-17 Skin complexion measuring device and method for intense pulse light xerophthalmia treatment

Publications (1)

Publication Number Publication Date
CN113679355A true CN113679355A (en) 2021-11-23

Family

ID=78586615

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111092569.1A Pending CN113679355A (en) 2021-09-17 2021-09-17 Skin complexion measuring device and method for intense pulse light xerophthalmia treatment

Country Status (1)

Country Link
CN (1) CN113679355A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114618090A (en) * 2022-03-14 2022-06-14 重庆大学 Energy control method of strong pulse laser xerophthalmia therapeutic instrument
CN115429293A (en) * 2022-11-04 2022-12-06 之江实验室 Sleep type classification method and device based on impulse neural network

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114618090A (en) * 2022-03-14 2022-06-14 重庆大学 Energy control method of strong pulse laser xerophthalmia therapeutic instrument
CN115429293A (en) * 2022-11-04 2022-12-06 之江实验室 Sleep type classification method and device based on impulse neural network
CN115429293B (en) * 2022-11-04 2023-04-07 之江实验室 Sleep type classification method and device based on impulse neural network

Similar Documents

Publication Publication Date Title
CN113679355A (en) Skin complexion measuring device and method for intense pulse light xerophthalmia treatment
JP6375557B2 (en) System and method for measuring and categorizing surface colors and spectra
CN101839979B (en) Method and device for measuring index number of canopy vegetation of crops
CN103268499B (en) Human body skin detection method based on multispectral imaging
CN201383157Y (en) Image acquisition device of finger vein certification system
CN107346434A (en) A kind of plant pest detection method based on multiple features and SVMs
CN106290238A (en) A kind of apple variety method for quick identification based on high light spectrum image-forming
CN103489006A (en) Computer vision-based rice disease, pest and weed diagnostic method
CN202887210U (en) Multispectral face recognition system
CN107123227A (en) A kind of embedded image flame detector and its recognition methods based on two waveband
US20220156479A1 (en) Spectral imaging chip and apparatus, information processing method, fingerprint living body identification device and fingerprint module
WO2016202165A1 (en) Multispectral banknote identification apparatus and method
CN105866050A (en) Low-cost lossless and fast detecting equipment for apple moldy core
CN112115838A (en) Thermal infrared image spectrum fusion human face classification method
KR102036043B1 (en) Diagnosis Device of optical skin disease based Smartphone
CN113080848B (en) Near infrared light skin detection device and method
CN105550668A (en) Apparatus for collecting biological features of living body and method for identifying biological features of living body
Robila Toward hyperspectral face recognition
CN109544627B (en) Intelligent hair analysis device and analysis method
CN113807320B (en) Human skin color recognition method and system based on multispectral imaging technology
CN111368801A (en) True and false fingerprint identification device and identification method thereof
CN106991371A (en) A kind of face type collection and analysis system based on traditional Chinese medical science five-element people's classification of TCM constitution
CN108965636A (en) A kind of flatbed scanner and its application
WO2022196999A1 (en) Fruit quantity measurement system and method therefor
US20230148269A1 (en) System for skin and hair measurement and analysis

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination