Skin detection method and apparatus based on multi-spectral polarized light

Abstract

Disclosed in the present invention are a skin detection method and apparatus based on multi-spectral polarized light. The method comprises: a light source alternately emitting first polarized light and second polarized light to illuminate skin, and collecting corresponding images, wherein the polarization direction of the first polarized light is different from the polarization direction of the second polarized light; extracting pixel information of the images; performing down-sampling processing on the pixel information, and acquiring a difference matrix of the pixel information; and performing visual transformation on the difference matrix, so as to obtain a skin heat map. Compared with existing PPGI-based living skin detection methods, the skin detection method in the present invention uses polarization to filter out reflected light on the surface of skin, so as to provide deeper and more accurate information for analysis, thereby improving the accuracy and real-time performance of living-skin detection and physiological monitoring.

Description

Skin detection method and device based on multispectral polarized light Technical Field

[0001] The present invention relates to the fields of computer vision and image processing, and in particular to a skin detection method and device based on multi-spectral polarized light. Background Art

[0002] The human body's reflection, refraction, and absorption of light allow many physiological parameters to be acquired in a non-contact manner. Currently, vital sign cameras have been used to detect physiological signals from the human face or body, and their concept and feasibility have been proven in real hospital scenarios such as ICUs, NICUs, and sleep medicine centers. However, the monitoring of physiological signals by vital sign cameras usually requires the selection of a skin area for physiological signal extraction. The selected skin area is the RoI (Region of Interest). Traditional machine learning-based methods will include areas with no physiological significance, such as hair and eyebrows, when selecting skin, and cannot guarantee the physiological characteristics of the detection area. Therefore, methods for selecting RoI based on live skin detection have attracted widespread attention.

[0003] Among the current methods, the skin perfusion mapping method for live skin detection based on photoplethysmography (PPGI) is widely used. This method can not only effectively determine live skin and assist in selecting RoI, but its liveness detection capability can also be used to help improve the success rate of anti-fraud systems. However, during implementation, this method requires continuous measurement of stable PPG signals for at least 2-3 cardiac cycles to obtain a signal for analysis, which limits the deployment of practical applications. For example, active newborns in the NICU find it difficult to remain still for a long time to ensure continuous and effective physiological monitoring, and automatic identification at airports cannot require passengers to wait for a considerable period of time for measurement and identification. Therefore, new methods are urgently needed to ensure real-time and accurate physiological detection and liveness detection, and can also be truly deployed in practical applications.

[0004] Therefore, the current methods for living skin detection still need to be further improved.

[0005] Summary of the Invention

[0006] The present invention proposes a skin detection method and device based on multispectral polarized light, which solves the technical problems of existing live skin detection methods based on PPGI, which are not real-time and cannot fully confirm that the subject is alive due to PPG signal interference when recording videos of living subjects.

[0007] In order to solve the above problems, the present invention proposes the following technical solutions:

[0008] The present invention provides a skin detection method and device based on multi-spectral polarized light.

[0009] In one aspect, the present invention provides a skin detection method based on multispectral polarized light, comprising:

[0010] Alternately emitting a first polarized light and a second polarized light from a light source to illuminate the skin, and collecting a first image of the skin illuminated by the first polarized light and a second image of the skin illuminated by the second polarized light, wherein the first polarized light and the second polarized light have different polarization directions;

[0011] extracting first pixel information of the first image and second pixel information of the second image;

[0012] Downsampling the first pixel information and the second pixel information, and obtaining a difference matrix between the first pixel information and the second pixel information;

[0013] The difference matrix is ​​visually transformed to obtain a skin thermal map.

[0014] Based on this technical solution, it is further preferred that the polarization direction of the first polarized light is perpendicular to the polarization direction of the second polarized light.

[0015] Based on this technical solution, further preferably, the step of alternately emitting first polarized light and second polarized light to illuminate the skin through a light source, and collecting a first image of the skin illuminated by the first polarized light and a second image of the skin illuminated by the second polarized light, includes:

[0016] Alternatingly emitting the first polarized light and the second polarized light toward the skin with an emission time of 1:1;

[0017] Images of the first polarized light and the second polarized light irradiated on the skin are captured to acquire the first image and the second image.

[0018] Based on this technical solution, further preferably, the extracting the first pixel information of the first image and the second pixel information of the second image includes: obtaining the RGB channel pixel values ​​of the first image as the first pixel information, and obtaining the RGB channel pixel values ​​of the second image as the second pixel information.

[0019] Based on this technical solution, further preferably, downsampling the first pixel information and the second pixel information and obtaining a difference matrix between the first pixel information and the second pixel information includes:

[0020] Processing the first pixel information and the second pixel information at a downsampling ratio of 0.1, and performing downsampling processing on the first pixel information and the second pixel information using a box kernel as an interpolation kernel to obtain processed first pixel information and processed second pixel information;

[0021] Obtaining a ratio of an R channel pixel value to a G channel pixel value in the processed first pixel information as a first R / G value, and obtaining a ratio of an R channel pixel value to a G channel pixel value in the processed second pixel information as a second R / G value;

[0022] The first R / G value is subtracted from the second R / G value to obtain the difference matrix, where the R channel is the red channel and the G channel is the green channel.

[0023] Based on this technical solution, further preferably, the visual conversion of the difference matrix to obtain a skin thermal map includes:

[0024] Setting the area with a value less than 0 in the difference matrix to 0 to obtain a clipped difference matrix;

[0025] Performing Gaussian filtering on the clipped difference matrix with a standard deviation of 1 to obtain a processed difference matrix;

[0026] Based on the RGB channel pixel values ​​in the processed difference matrix, the skin thermal map is obtained by visual conversion.

[0027] On the basis of this technical solution, it is further preferred that the skin area in the first image and the second image is determined based on the first pixel information and the second pixel information, and the skin thermodynamic map of the skin area is obtained through the difference matrix.

[0028] In a second aspect, the present invention further provides a skin detection device based on multispectral polarized light, the skin detection device based on multispectral polarized light comprising:

[0029] an image acquisition module, configured to alternately emit a first polarized light and a second polarized light through a light source to illuminate the skin, and to acquire a first image of the skin illuminated by the first polarized light and a second image of the skin illuminated by the second polarized light, wherein the first polarized light is parallel polarized light and has a different polarization direction from the second polarized light, which is cross-polarized light;

[0030] an information extraction module, configured to extract first pixel information of the first image and second pixel information of the second image;

[0031] an image processing module, configured to perform downsampling processing on the first pixel information and the second pixel information, and obtain a difference matrix between the first pixel information and the second pixel information;

[0032] An image generation module is used to visually transform the difference matrix to obtain a skin thermal map.

[0033] Based on this technical solution, it is further preferred that the image acquisition module includes the light source, the light-collecting device and the polarizer, wherein the light source is an LED light source provided by time-division multiplexed multi-channel LEDs, the light-collecting device is a camera, and the polarizer is arranged between the light-collecting device and the skin.

[0034] Based on this technical solution, further preferably, the image processing module includes a downsampling unit, an acquisition unit and a calculation unit.

[0035] The downsampling unit performs downsampling processing on the first pixel information and the second pixel information using a box kernel as an interpolation kernel; the acquisition unit acquires the ratio of the R channel pixel value to the G channel pixel value in the processed first pixel information as a first R / G value, and acquires the ratio of the R channel pixel value to the G channel pixel value in the processed second pixel information as a second R / G value; the calculation unit subtracts the first R / G value from the second R / G value to obtain the difference matrix.

[0036] Compared with the prior art, the present invention can achieve the following technical effects:

[0037] The present invention combines multispectral technology and polarized light technology, not only obtaining the optical properties of objects in different spectral bands, but also analyzing the impact of polarized light on the spectrum, using polarization to filter out light reflected from the skin surface, providing deeper and more accurate information for analysis, thereby improving the accuracy and real-time performance of living skin detection and physiological monitoring, and proposing a novel and simple algorithm to complete living skin detection.

[0038] The skin detection device based on multispectral polarized light proposed in the present invention utilizes the wavelength dependence of the depolarization effect and multi-wavelength skin depolarization. The MSD algorithm only requires two images generated under parallel and cross polarization as input. Under a 20 frame / second camera, the total duration of live skin detection can in principle be less than 0.1 second, and the application is simple. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] FIG1 is a flow chart of the skin detection method based on multispectral polarized light according to the present invention;

[0040] FIG2 is a schematic diagram of the results of living skin detection based on multispectral polarized light according to Example 1 of the present invention;

[0041] FIG3 is a diagram of a skin detection device based on multi-spectral polarized light according to Example 2 of the present invention;

[0042] FIG4 is a schematic diagram of the in vivo skin detection results based on PPGI in Comparative Example 1 of the present invention;

[0043] FIG5 is a schematic diagram of live skin segmentation of premature infants in a neonatal intensive care unit according to Example 2 of the present invention and Comparative Example 2;

[0044] FIG6 is a schematic diagram of living skin segmentation of ICU patients according to Example 2 of the present invention and Comparative Example 2. DETAILED DESCRIPTION

[0045] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field based on the technology to which the claims of the present invention belong without making any creative work are within the scope of protection of the present invention.

[0046] It should be understood that the terms used in this description of the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the embodiments of the present invention. As used in the description of the embodiments of the present invention and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0047] In light-based remote sensing, whether for monitoring human physiological parameters or detecting liveness, there are issues with artifacts and noisy data caused by light reflected from the skin. Furthermore, current physiological monitoring and liveness detection methods are difficult to translate into practical applications due to their non-real-time nature. However, the use of two crossed polarizers can effectively remove specular reflections from the skin surface, thereby significantly improving the quality of PPG extracted by vital sign cameras. Assuming depolarization of skin tissue, the light sensitivity of each channel can be shifted to the longer wavelength portion of its sensing range, resulting in changes in skin color and blood volume pulse characteristics.

[0048] Based on the above findings and assumptions, the present invention proposes a method and application of live skin detection based on multi-spectral polarized light. According to the different skin penetration of photons of different wavelengths, the different degrees of depolarization of skin tissue at R, G, and B wavelengths are used to distinguish skin and non-skin pixels. In addition, a new algorithm segmentation core algorithm (MSD) is proposed. Only by relying on two pictures in parallel polarization state and cross polarization state, the R / G contrast of live skin pixels can be analyzed to achieve live skin detection. Live skin detection based on the above-mentioned MSD algorithm can help to select the appropriate RoI more accurately and in real time when extracting physiological signals. Removing the mirror reflection of the skin by polarized light can also effectively reduce motion artifacts and improve the robustness of physiological signals.

[0049] FIG1 is a flow chart of the skin detection method based on multispectral polarized light. Therefore, this embodiment 1 provides a skin detection method based on multispectral polarized light, which specifically includes:

[0050] Step 1: Alternately emit first polarized light and second polarized light through a light source to illuminate the skin, and collect a first image of the skin illuminated by the first polarized light and a second image of the skin illuminated by the second polarized light, wherein the polarization directions of the first polarized light and the second polarized light are different.

[0051] Step 2: extract first pixel information of the first image and second pixel information of the second image.

[0052] Step 3: downsample the first pixel information and the second pixel information, and obtain a difference matrix between the first pixel information and the second pixel information.

[0053] Step 4: Visually transform the difference matrix to obtain a skin thermal map.

[0054] The polarization direction of the first polarized light is perpendicular to the polarization direction of the second polarized light, and the polarization direction of the first polarized light is parallel to the direction of the polarizer.

[0055] The step of alternately emitting a first polarized light and a second polarized light through a light source to illuminate the skin, and collecting a first image of the skin illuminated by the first polarized light and a second image of the skin illuminated by the second polarized light, comprises:

[0056] The first polarized light and the second polarized light are alternately emitted toward the skin with an emission time of 1:1.

[0057] Images of the first polarized light and the second polarized light irradiated on the skin are captured to acquire the first image and the second image.

[0058] The extracting the first pixel information of the first image and the second pixel information of the second image includes: obtaining RGB channel pixel values ​​of the first image as the first pixel information, and obtaining RGB channel pixel values ​​of the second image as the second pixel information.

[0059] The downsampling the first pixel information and the second pixel information and obtaining a difference matrix between the first pixel information and the second pixel information includes:

[0060] The first pixel information and the second pixel information are processed at a downsampling ratio of 0.1, and the first pixel information and the second pixel information are downsampled using a box kernel as an interpolation kernel to obtain processed first pixel information and processed second pixel information; wherein the downsampling preprocessing includes using the box kernel as an interpolation kernel, which not only reduces the influence of noise in the pixel information, but also helps to improve the processing efficiency of the pixel information.

[0061] The ratio of the R channel pixel value to the G channel pixel value in the processed first pixel information is obtained as a first R / G value, and the ratio of the R channel pixel value to the G channel pixel value in the processed second pixel information is obtained as a second R / G value.

[0062] The first R / G value is subtracted from the second R / G value to obtain the difference matrix, where the R channel is the red channel and the G channel is the green channel.

[0063] The visual conversion of the difference matrix to obtain a skin thermal map includes:

[0064] The area with a value less than 0 in the difference matrix is ​​set to 0 to obtain a clipped difference matrix.

[0065] The clipped difference matrix is ​​subjected to Gaussian filtering with a standard deviation of 1 to obtain a processed difference matrix.

[0066] Based on the RGB channel pixel values ​​in the processed difference matrix, the heat map is obtained by visual conversion.

[0067] In a preferred embodiment, the above steps 3 and 4 also include another embodiment, that is, the MSD algorithm, specifically further including:

[0068] S1, by switching between parallel polarization mode and cross polarization mode, acquiring their respective images, using 0.1 as the downsampling ratio and box kernel as the interpolation method, respectively, to perform downsampling preprocessing, thereby obtaining two downsampled images;

[0069] S2, subtract the R / G values ​​of the downsampled image in the cross-polarization mode from the R / G values ​​of the downsampled image in the parallel polarization mode, to obtain a matrix H of the difference between the two images in units of pixel values;

[0070] S3, set the area smaller than 0 in the matrix H to 0, and obtain the cropped matrix H;

[0071] S4, the cropped H is subjected to Gaussian filtering with a standard deviation of 1;

[0072] S5, the final H is displayed in the form of a heat map.

[0073] That is,

[0074] Input: Get I for parallel polarization mode and cross polarization mode para and I cros , r = 0.1, σ = 1 (default);

[0075] 1:I para =imresize(I para ,r',box'→box downsampling;

[0076] 2:I cros =imresize(I cros ,r',box'→box downsampling;

[0077] 3: H=I cros (R) / I cros (GI para (R) / I para (G);

[0078] 4: H(H<0=0→cropping difference matrix;

[0079] 5: H = imgaussfilt(H, σ→spatial smoothing);

[0080] Output: Skin heatmap.

[0081] Among them, I para is a parallel image, I cros is the cross image, r is the reduction ratio, σ is the standard deviation of the Gaussian filter, and box' is the box kernel.

[0082] Finally, after processing with a Gaussian filter, the visualization transformation is performed, so that the skin heat map is smoother.

[0083] According to the above method, 10 adult subjects with different skin colors were selected as test subjects. The test subjects sat on a chair, 3 meters away from the camera, and located in a parallel position in front of it. At the same time, a doll face and doll torso that could deceive the system were placed next to the test subjects, with similar skin tone and appearance.

[0084] Results: This embodiment uses a multispectral polarized light-based physiological signal extraction process. When polarization switches from parallel to crossed, specular reflection from the skin surface is significantly reduced, the skin's chromaticity changes, making the face appear "redder," the pulsation intensity of the RGB signal (AC / DC) increases, and the relative PPG amplitudes between the RGB channels change. For example, the blood volume pulsation characteristic changes from [0.4, 0.5, 0.6] to [0.3, 0.8, 0.5]. Because the PPG signal is obtained by filtering and synthesizing the original signal, the enhancement of the original signal and the contrast between the RGB channels facilitate the acquisition of a more robust PPG signal.

[0085] As can be seen from Figure 2, the skin detection method based on multispectral polarized light of Example 1 is able to identify dolls and doll torsos that deceive the system, indicating that the method of this embodiment effectively distinguishes skin (test object) and non-skin pixels (i.e., doll and doll torso) by using the different degrees of depolarization of skin tissue at R, G, and B wavelengths; and relying solely on two images in parallel and perpendicular polarization states, it is possible to analyze the R / G contrast of live skin pixels, thereby achieving live skin detection and more accurate ROI extraction.

[0086] Example 2

[0087] A skin detection device based on multispectral polarized light, as shown in FIG3 , includes an image acquisition module 10 , an information extraction module 20 , an image processing module 30 , and an image generation module 40 , and specifically also includes:

[0088] An image acquisition module is configured to alternately emit first polarized light and second polarized light through a light source to illuminate the skin, and to capture a first image of the skin illuminated by the first polarized light and a second image of the skin illuminated by the second polarized light, wherein the first polarized light is parallel polarized light and has a different polarization direction from the second polarized light, which is cross-polarized light.

[0089] The image acquisition module includes the light source, a light-collecting device and a polarizer, wherein the light source is an LED light source provided by time-division multiplexing multi-channel LEDs, the light-collecting device is a camera, and the polarizer is arranged between the light-collecting device and the skin; the light source emits any monochromatic laser and illuminates the surface of the living body.

[0090] To facilitate the continuous emission of different types of polarized light, this embodiment uses a time-division multiplexed multi-channel LED board to provide an LED light source based on the existing light source. The time-division multiplexed multi-channel LED board sequentially emits two types of polarized light, thereby irradiating the surface of living skin within a relatively short time interval. The facial skin recording of each subject lasts for 2 minutes, with 1 minute of parallel polarization recording and 1 minute of cross-polarization recording. That is, the camera polarizer is rotated 90° to switch the polarization mode during the recording process. The recording environment is carried out under ambient lighting conditions, without any control of background lighting. The polarized light is mixed with indoor light. In this way, the skin detection device of this embodiment can also be implemented in everyday environments.

[0091] An information extraction module is used to extract first pixel information of the first image and second pixel information of the second image.

[0092] An image processing module is used to downsample the first pixel information and the second pixel information and obtain a difference matrix between the first pixel information and the second pixel information; the image processing module includes a downsampling unit, an acquisition unit and a calculation unit.

[0093] The downsampling unit performs downsampling processing on the first pixel information and the second pixel information using a box kernel as an interpolation kernel; the acquisition unit acquires the ratio of the R channel pixel value to the G channel pixel value in the processed first pixel information as a first R / G value, and acquires the ratio of the R channel pixel value to the G channel pixel value in the processed second pixel information as a second R / G value; the calculation unit subtracts the first R / G value from the second R / G value to obtain the difference matrix.

[0094] An image generation module is used to visually transform the difference matrix to obtain a skin thermal map.

[0095] The device described in this embodiment was used to test 20 patients, including 10 patients from the NICU with birth ages ranging from 10 minutes to 8 days, and 10 patients from the ICU with ages ranging from 5 months to 88 years.

[0096] Both the NICU and ICU recordings were conducted under ambient lighting conditions, with no control for background lighting, resulting in the mixing of polarized light with room light. In the NICU, the infant lay in an incubator with their eyes covered. The recording setup (camera and light source) was placed above the incubator, filming the infant from above. Physiological monitoring sessions were conducted under ambient hospital lighting conditions, while in the ICU, the patient lay in bed, recorded from above by the setup. The ICU recording method was similar to that of the NICU, but the patient was in bed.

[0097] Comparative Example 1

[0098] A living skin detection method, which differs from Example 1 in that it uses existing PPGI-based living skin detection, specifically includes the following steps:

[0099] Firstly, the Lucas-Kanade optical flow method is used to dynamically track the feature points of the image sequence, and the image is corrected through affine transformation to reduce motion artifacts and improve the IPPG signal quality.

[0100] Then, a sliding window was used to traverse the image to obtain the Spearman correlation coefficient between the spatial pixel average signal of each window and the spatial pixel average signal of the entire image, and correlation topography imaging was performed to obtain the skin blood perfusion distribution image.

[0101] Results: Figures 2 and 4 are the detection results of the living skin detection method of Example 1 and Comparative Example 1, respectively. The MSD (R / G) in the last column of Figure 2 is the method adopted in Example 1. Figure 4 is the living skin heat map of 10 subjects obtained by PPGI and MSD (R / B or R / G) in Comparative Example 1. The brighter colors in the heat map represent the skin area.

[0102] It can be seen that compared with the PPGI of Comparative Example 1, the skin detection method based on multispectral polarized light of Example 1 can process the live skin pixels with R / G contrast by relying only on two images in the parallel polarization state and the vertical polarization state. The RGB channel pixel values ​​in the difference matrix after the processing are visualized and converted to obtain the skin heat map, thereby realizing live skin detection and more accurate ROI extraction.

[0103] Comparative Example 2

[0104] A living skin detection device based on PPGI is different from Example 2 in that the existing PPGI detection device is used to perform living skin detection.

[0105] Results: Figures 5-6 show the results of live skin segmentation using the live skin detection devices of Example 2 and Comparative Example 2, respectively. Figure 5 shows the effects of PPGI and MSD (R / G) on premature infants in the neonatal intensive care unit. It can be seen that in Example 2, live skin detection based on the MSD (R / G) algorithm can help more accurately and in real time select the appropriate ROI when extracting physiological signals. Removing skin specular reflections using polarized light can also effectively reduce motion artifacts and improve the robustness of physiological signals.

[0106] Figure 6 shows the evaluation of PPGI and MSD (R / G) for patients with worsening conditions in the ICU. It can be seen that in Example 2, the existing method for extracting physiological signals includes areas with no physiological significance, such as hair and eyebrows, when selecting the ROI of the skin, and the physiological characteristics of the detection area cannot be guaranteed. In contrast, Example 2 uses polarization to filter out light reflected from the skin surface, providing deeper and more accurate information for analysis, thereby improving the accuracy and real-time performance of physiological monitoring.

[0107] In summary, the present invention provides a method and device for live skin detection based on multispectral polarized light. By combining multispectral and polarized light technologies, the varying degrees of depolarization of skin tissue at RGB wavelengths are utilized to distinguish skin from non-skin pixels. The proposed MSD algorithm is then used to detect skin pixels based on the R / G contrast between parallel and cross polarization. The MSD algorithm selects ROIs more efficiently and in real time, and the resulting PPG signal is more robust because it removes the influence of specular light reflected from the skin and captures more deep skin information. Furthermore, the MSD algorithm requires only two images generated under parallel and cross polarization as input. In principle, the total duration of live skin detection can be less than 0.1 seconds with a 20-frame-per-second camera, making it simple to use.

[0108] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A skin detection method based on multispectral polarized light, characterized in that, Including: The skin is irradiated by alternately emitting a first polarized light and a second polarized light from a light source, and a first image of the first polarized light irradiated on the skin and a second image of the second polarized light irradiated on the skin are collected, wherein the polarization directions of the first polarized light and the second polarized light are different; Extract the first pixel information of the first image and the second pixel information of the second image; Perform downsampling processing on the first pixel information and the second pixel information, and obtain a difference matrix of the first pixel information and the second pixel information; Visually transform the difference matrix to obtain a skin heat map.

2. The skin detection method based on multi-spectral polarized light according to claim 1, wherein The polarization direction of the first polarized light is perpendicular to the polarization direction of the second polarized light.

3. The skin detection method based on multi-spectral polarized light according to claim 2, characterized in that, The step of irradiating the skin by alternately emitting the first polarized light and the second polarized light from the light source and collecting the first image of the first polarized light irradiated on the skin and the second image of the second polarized light irradiated on the skin includes: Alternately emit the first polarized light and the second polarized light to the skin at an emission time ratio of 1:1; Take images of the first polarized light and the second polarized light irradiated on the skin to obtain the first image and the second image.

4. The skin detection method based on multi-spectral polarized light according to claim 2, characterized in that The step of extracting the first pixel information of the first image and the second pixel information of the second image includes: obtaining the RGB channel pixel values of the first image as the first pixel information, and obtaining the RGB channel pixel values of the second image as the second pixel information.

5. The skin detection method based on multi-spectral polarized light according to claim 4, characterized in that, The step of performing downsampling processing on the first pixel information and the second pixel information and obtaining a difference matrix of the first pixel information and the second pixel information includes: Process the first pixel information and the second pixel information at a downsampling ratio of 0.1, and use a box kernel as an interpolation kernel to perform downsampling processing on the first pixel information and the second pixel information to obtain processed first pixel information and processed second pixel information; Obtain the ratio of the R-channel pixel value to the G-channel pixel value in the processed first pixel information as the first R / G value, and obtain the ratio of the R-channel pixel value to the G-channel pixel value in the processed second pixel information as the second R / G value; Subtract the first R / G value from the second R / G value to obtain the difference matrix, where the R channel is the red channel and the G channel is the green channel.

6. The skin detection method based on multi-spectral polarized light according to claim 5, wherein The step of visually transforming the difference matrix to obtain a skin heat map includes: Set the area in the difference matrix less than 0 to 0 to obtain a cropped difference matrix; Perform Gaussian filtering on the cropped difference matrix with a standard deviation of 1 to obtain a processed difference matrix; Based on the RGB channel pixel values in the processed difference matrix, visually transform to obtain the skin heat map.

7. The skin detection method based on multi-spectral polarized light according to claim 6, characterized in that, It also includes: Based on the first pixel information and the second pixel information, determine the skin areas in the first image and the second image, and obtain the skin heat map of the skin area through the difference matrix.

8. A skin detection device based on multispectral polarized light, characterized in that, The skin detection device based on multi-spectral polarized light includes: An image acquisition module, which is used to irradiate the skin by alternately emitting a first polarized light and a second polarized light through a light source, and acquire a first image of the first polarized light irradiating on the skin and a second image of the second polarized light irradiating on the skin, wherein the first polarized light is a parallel polarized light, and the polarization direction is different from that of the second polarized light which is a cross polarized light; An information extraction module, which is used to extract the first pixel information of the first image and the second pixel information of the second image; An image processing module, which is used to perform downsampling processing on the first pixel information and the second pixel information, and obtain a difference matrix of the first pixel information and the second pixel information; An image generation module, which is used to visually transform the difference matrix to obtain a skin heat map.

9. The skin detection device based on multi-spectral polarized light according to claim 8, characterized in that, The image acquisition module includes the light source, a light collecting device and a polarizer, wherein the light source is an LED light source provided by a time division multiplexing multi-channel LED, the light collecting device is a camera, and the polarizer is arranged between the light collecting device and the skin.

10. The skin detection device based on multi-spectral polarized light according to claim 8, characterized in that, The image processing module includes a downsampling unit, an acquisition unit and a calculation unit; The downsampling unit performs downsampling processing on the first pixel information and the second pixel information by using a box kernel as an interpolation kernel; The acquisition unit obtains the ratio of the R-channel pixel value to the G-channel pixel value in the processed first pixel information as a first R / G value, and obtains the ratio of the R-channel pixel value to the G-channel pixel value in the processed second pixel information as a second R / G value; The calculation unit subtracts the first R / G value from the second R / G value to obtain the difference matrix.

Images