Intelligent control method and system for hair removal device

Abstract

The present application provides an intelligent control method and system for a hair removal device, comprising: obtaining three-dimensional structural data of skin at different depths, and establishing a three-dimensional distribution model of hair follicles at each layer; performing grid division on a hair removal area, and dividing the grid into a dense area and a sparse area; obtaining energy loss characteristics at a division boundary and coupling response characteristics between the vibration of a hair removal head and a division granularity, and establishing a hair removal accuracy and efficiency prediction model; optimizing hair removal parameters using a particle swarm optimization algorithm, and constructing a mapping relationship database between the division granularity and the hair removal parameters; combining adaptability judgment and the mapping relationship database between the division granularity and the hair removal parameters, and calculating hair removal parameters corresponding to different skin characteristic parameters; adjusting hair removal energy output, and adjusting hair removal intensity in real time according to skin deformation data; constructing a hair removal timing prediction model to predict the best hair removal timing, generate a personalized hair removal plan, and adjust and update a user hair removal database in real time.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an intelligent control method and system for a hair removal device. Background Art

[0002] As people's demand for beauty and personal care continues to increase, the demand for hair removal devices in the market is also increasing. Although traditional hair removal methods, such as shaving, plucking and laser hair removal, can meet people's hair removal needs to a certain extent, there are still many shortcomings. These methods are often accompanied by pain during use, the hair removal effect is not lasting, and may cause damage to the skin. These problems have limited the widespread application of traditional hair removal methods and the improvement of user experience to a certain extent. Shaving and plucking are the most common physical hair removal methods, but shaving can only remove surface hair, and the hair will regenerate quickly. Plucking often causes pain and skin irritation due to the direct removal of hair follicles. Laser hair removal, as a more advanced hair removal method, can destroy hair follicles through the thermal effect of lasers, thereby achieving the effect of long-term hair removal. However, laser hair removal has high requirements for the skin, and improper operation may cause side effects such as skin burns and pigmentation. At the same time, the high cost of laser hair removal limits its popularity. The development of modern hair removal technology has significantly improved the accuracy, efficiency and user experience of hair removal devices. However, existing hair removal devices still face some technical challenges in practical applications. There are significant differences in the skin and hair characteristics of different users, including skin thickness, pigmentation, hair density and growth direction. These differences make it difficult for universal hair removal parameter settings to adapt to all users, affecting the hair removal effect and user satisfaction. The problem of uneven energy distribution during hair removal is also prominent. Traditional hair removal devices usually adopt a fixed energy output mode, which is difficult to dynamically adjust according to the specific conditions of different skin areas, resulting in uneven hair removal effects and even possible damage to the skin. In addition, the vibration characteristics of the hair removal head are closely related to the change in skin contact area. The grid division of different granularities has a greater impact on the vibration response. How to optimize the vibration parameters during hair removal to ensure hair removal accuracy and efficiency is an urgent problem to be solved. In the hair removal process, temperature monitoring and energy regulation are also an important technical problem. Existing hair removal devices often lack a real-time temperature monitoring mechanism, making it difficult to dynamically adjust energy output during hair removal, resulting in excessive skin temperature and increasing the risk of skin burns. At the same time, there is a lack of real-time evaluation and adjustment mechanism for skin deformation and user comfort. Users may feel uncomfortable during hair removal, affecting the user experience. In addition, the lack of personalized hair removal solutions is also a major shortcoming of existing hair removal devices. Each user has different skin and hair characteristics, and their hair removal needs and expected effects are also different. It is difficult for existing hair removal devices to generate personalized hair removal solutions based on the user's historical hair removal data, skin characteristics, and feedback data. This means that when using a hair removal device, users often need to adjust parameters multiple times in an attempt to find the hair removal settings that suit them. This process is cumbersome and the effect is not good. Therefore, modern hair removal technology still faces many challenges in improving hair removal accuracy and efficiency, and achieving personalization and safety.How to use advanced technical means to solve problems such as differences in skin and hair characteristics among different users, uneven energy distribution, insufficient temperature monitoring and energy regulation, and lack of personalized hair removal solutions is the key to the current development of hair removal device technology. Summary of the invention

[0003] The present invention provides an intelligent control method for a hair removal device, which mainly includes:

[0004] Optical coherence tomography was used to obtain three-dimensional structural data at different depths of the skin, to establish a three-dimensional distribution model of hair follicles at each layer, and to determine the spatial non-uniformity of the density distribution of hair follicles at each layer.

[0005] According to the spatial non-uniformity of the density distribution of hair follicles in each layer, the hair removal area is divided into grids. By calculating the hair follicle density value of each grid, the grid is divided into dense areas and sparse areas, and a hair removal area division scheme matching the hair follicle distribution is obtained;

[0006] According to the hair removal area division scheme, the energy loss characteristics at the division boundary and the coupling response characteristics of the hair removal head vibration and the division granularity are obtained, and a hair removal accuracy and efficiency prediction model is established to predict the hair removal accuracy and efficiency values;

[0007] According to the generated hair removal area division scheme, the division granularity of each division area and the corresponding initial value of the hair removal parameter are obtained, the hair removal parameter is optimized by using a particle swarm optimization algorithm, and a mapping relationship database between the division granularity and the hair removal parameter is constructed;

[0008] According to the spectral characteristics of skin tissue, the optical characteristic parameters are adjusted in real time and Monte Carlo simulation is performed. The hair removal parameters corresponding to different skin characteristic parameters are calculated by combining the adaptive judgment and the mapping relationship database between the division granularity and the hair removal parameters.

[0009] The infrared thermal imaging camera is used to obtain the temperature distribution image of the hair removal area in real time, and the hair removal energy output is judged and adjusted based on the maximum temperature value and the area of ​​the high-temperature zone. A user comfort score prediction model is built based on the skin deformation data to adjust the hair removal intensity in real time.

[0010] A user hair removal database is built based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data. Deep neural networks and reinforcement learning algorithms are used to extract features and optimize hair removal parameters. A hair removal timing prediction model is built to predict the optimal hair removal time, generate personalized hair removal plans, and adjust and update the user hair removal database in real time.

[0011] Furthermore, the optical coherence tomography method is used to obtain three-dimensional structural data of the skin at different depths, establish a three-dimensional distribution model of hair follicles at each layer, and determine the spatial non-uniformity characteristics of the density distribution of hair follicles at each layer, including:

[0012] An optical coherence tomography system is used to perform multi-level scanning of the target skin area to obtain a three-dimensional structural image from the skin surface to the deep tissue, and a Gaussian filter is used to denoise the three-dimensional structural image; based on the multi-threshold segmentation of the intensity histogram, the three-dimensional structural image is divided into the epidermis, dermis and subcutaneous tissue layer according to the optical properties of the tissue and the preset optical property threshold, and the Canny edge detection algorithm is used to identify the boundaries between the layers to form a three-dimensional digital model of the skin layer. The optical properties include the scattering coefficient and absorption coefficient of the tissue; the Canny edge detection algorithm is applied to the obtained three-dimensional digital model of the skin layer to identify the contour of the hair follicle and calculate the characteristic parameters of the hair follicle, including the length, diameter and tilt angle of the hair follicle. ; According to the characteristic parameters of hair follicles, the K-means clustering algorithm is used for model training to construct a three-dimensional distribution model of hair follicles at each level, and the hair follicles with similarity of hair follicle characteristic parameters greater than the preset similarity threshold are classified into one category; based on the established three-dimensional digital model of skin stratification, the skin stratification model is divided into several equal-volume grid units, the number of hair follicles in each grid unit is calculated, and the hair follicle density is determined; according to the hair follicle density, a hair follicle density heat map is drawn, and different color depths are used to display the spatial distribution of hair follicle density; the hair follicle density heat map is analyzed using the Moran's index calculation method, the global Moran's index and the local Moran's index are calculated, and the spatial autocorrelation of hair follicle distribution is evaluated; among which, the global Moran's index calculation formula is: N is the number of hair follicles, x i and x j is the hair follicle density value, is the average hair follicle density, w ij is the weight matrix; the local Moran index calculation formula is S0 2 is the overall variance; based on the calculation results of the global Moran's index and the local Moran's index, a three-dimensional contour surface map reflecting the spatial distribution law of hair follicle density at each layer of the skin is drawn to show the spatial non-uniformity characteristics of hair follicle distribution.

[0013] Furthermore, the hair removal area is divided into grids according to the spatial non-uniformity characteristics of the hair follicle density distribution of each layer, and the grids are divided into dense areas and sparse areas by calculating the hair follicle density value of each grid, so as to obtain a hair removal area division scheme that matches the hair follicle distribution, including:

[0014] According to the spatial non-uniformity characteristics of the hair follicle density distribution in each layer, based on the preset initial grid size, the quadtree algorithm is used to perform initial grid division on the hair removal area, and the entire hair removal area is divided into n×m initial grids, and each initial grid is assigned a unique identifier; the hair follicle density value in each initial grid is calculated, and it is determined whether the hair follicle density in the grid exceeds the preset density threshold according to the preset density threshold. If it exceeds, the grid is marked as a dense area, otherwise it is marked as a sparse area; for the grid marked as a dense area, it is divided into four sub-grids, and the hair follicle density calculation and area marking are re-performed on each sub-grid ; If the hair follicle density in the sub-grid still exceeds the preset density threshold, continue to divide the sub-grid into smaller sub-grids until the grid size reaches the preset grid size threshold or the hair follicle density in the grid area is lower than the preset density threshold; identify the boundary points of all adjacent grids, for each boundary point, take the hair follicle density value of the four surrounding grid center points, use the bilinear interpolation formula to calculate the interpolation density of the boundary point, adjust the boundary position according to the interpolation density value, and generate a smooth grid boundary; generate a hair removal area division plan based on the smoothed grid boundary, and convert the hair removal area division plan into a vector graphics file in SVG or DXF format.

[0015] Furthermore, according to the hair removal area division scheme, the energy loss characteristics at the division boundary and the coupling response characteristics of the hair removal head vibration and the division granularity are obtained, a hair removal accuracy and efficiency prediction model is established, and the hair removal accuracy and efficiency values ​​are predicted, including:

[0016] According to the hair removal area division plan, the finite element method is used to construct a three-dimensional simulation model, and the grids of different granularities are divided into three sizes, including size A, size B, and size C. The corresponding hair removal parameters are set for each granularity of the grid. The hair removal parameters include hair removal energy density, hair removal head vibration frequency, and hair removal time. Using the Fourier heat conduction equation The hair removal process was numerically simulated, where T is temperature, t is time, and α is the thermal diffusion coefficient. The finite difference method was used to discretize the equation to set boundary conditions and initial conditions. The temperature distribution was calculated using an iterative solver. Based on the temperature gradient and heat flux density at the grid boundary, the boundary energy loss characteristic curve was obtained. The boundary conditions included the temperature and heat flux density of the skin surface, and the initial condition was the initial temperature distribution. ANSYS Mechanical software was used to simulate the vibration characteristics of the hair removal head, establish a geometric model of the hair removal head, set material properties and boundary conditions, and perform modal analysis to calculate the natural frequency and vibration mode of the hair removal head according to the skin contact area and damping coefficient under different division granularity. The frequency response function was analyzed by harmonic response to obtain the vibration frequency, amplitude and phase information, and obtain the coupling response data of the division granularity and the hair removal head vibration. According to the energy loss characteristic curve and the vibration coupling response data, the recurrent neural network algorithm was used for model training to establish a hair removal accuracy and efficiency prediction model. According to the input division granularity, the hair removal accuracy and efficiency values ​​were predicted.

[0017] Furthermore, the method of generating a hair removal area division scheme, obtaining the division granularity of each division area and the corresponding initial value of the hair removal parameter, optimizing the hair removal parameters using a particle swarm optimization algorithm, and constructing a mapping relationship database between the division granularity and the hair removal parameters includes:

[0018] According to the generated hair removal area division scheme, the division granularity of each division area is obtained; according to the granularity information of each division area, the initial value of the hair removal parameter corresponding to the division granularity of the corresponding hair removal area is obtained; according to the initial hair removal parameters of each division area, the particle swarm optimization algorithm is used to optimize the hair removal parameters, and the particle swarm scale is set to use the particle swarm optimization algorithm to solve the nonlinear relationship model, set the population size to a, the maximum number of iterations to b, the inertia weight to c, and the acceleration factor to d, and the hair removal parameters of each division area are optimized to obtain the optimal hair removal parameters of each division area; according to the division granularity of the hair removal area and the corresponding optimal hair removal parameters, a continuous mapping relationship between the division granularity of the hair removal area and the optimal hair removal parameters is established, and a mapping relationship database between the division granularity and the hair removal parameters is constructed.

[0019] Furthermore, the optical characteristic parameters are adjusted in real time according to the spectral characteristics of the skin tissue and Monte Carlo simulation is performed, and the hair removal parameters corresponding to different skin characteristic parameters are calculated by combining the adaptability judgment and the mapping relationship database between the division granularity and the hair removal parameters, including:

[0020] A spectrometer is used to obtain the skin tissue reflectance spectrum in real time, and the sampling wavelength range, sampling interval and sampling frequency are set to obtain the real-time reflectance spectrum data of the skin tissue; based on the reflectance spectrum data and the three-dimensional digital model of skin stratification, the initial optical characteristic parameter value is set, and a forward Monte Carlo simulation is performed to calculate the initial simulated reflectance spectrum; the difference between the simulated reflectance spectrum and the measured reflectance spectrum is compared to calculate the reflectance spectrum error; the optical characteristic parameter value is adjusted according to the magnitude of the reflectance spectrum error, and the forward Monte Carlo simulation is re-performed to update the simulated reflectance spectrum; the forward Monte Carlo simulation is repeated until the error between the simulated reflectance spectrum and the measured reflectance spectrum is less than the preset error threshold, and the optical characteristic parameter value of the skin tissue is determined; according to the skin tissue characteristic parameters and the corresponding adaptive label number According to the data, support vector machine algorithm is used for model training to determine the adaptability of the skin to the current hair removal parameters; if the adaptability of the skin to the current hair removal parameters is low, the hair removal parameters are adjusted, and the current skin characteristic parameters are used as query conditions to find matching hair removal parameters from the database of mapping relationships between division granularity and hair removal parameters; if the same skin characteristic parameters do not exist in the database of mapping relationships between division granularity and hair removal parameters, the Euclidean distance between the current skin characteristic parameters and the sample points in the database is calculated, and K sample points whose Euclidean distance is less than a preset distance threshold are determined, and the hair removal parameters of these K sample points are weighted averaged to calculate the hair removal parameters corresponding to different skin characteristic parameters, and the hair removal parameters corresponding to different skin characteristic parameters are updated to the hair removal device control unit.

[0021] Furthermore, the temperature distribution image of the hair removal area is obtained in real time by an infrared thermal imaging camera, the hair removal energy output is judged and adjusted in combination with the maximum temperature value and the area of ​​the high temperature zone, and a user comfort score prediction model is constructed according to the skin deformation data to adjust the hair removal intensity in real time, including:

[0022] An infrared thermal imaging camera is used, and the acquisition frequency is set to obtain the temperature distribution image of the hair removal area in real time; the Otsu adaptive threshold method is used to segment the temperature distribution image, and the image is divided into a high temperature area and a low temperature area; the maximum temperature value and the area of ​​the high temperature area are extracted to determine whether the maximum temperature value exceeds the preset temperature threshold, and whether the area of ​​the high temperature area exceeds the preset area threshold; if the maximum temperature value or the area of ​​the high temperature area exceeds the preset threshold, the fuzzy control algorithm is used to adaptively adjust the hair removal energy output according to the temperature distribution, and the input variables are set as the maximum temperature and the high temperature area, and the output variable is the energy adjustment coefficient. Fuzzy rules are defined to obtain energy adjustment instructions and dynamically control the hair removal energy; skin deformation data and the corresponding user comfort score are obtained, and a recurrent neural network algorithm is used for model training to establish a user comfort score prediction model; skin deformation data during the hair removal process is obtained through a pressure sensor array, and the user comfort score prediction model is used to predict the user comfort score in real time; if the predicted user comfort score is lower than the preset comfort score threshold, the hair removal intensity is dynamically adjusted according to the user comfort score.

[0023] Furthermore, a user hair removal database is constructed based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data, and a deep neural network and reinforcement learning algorithm are used to extract features and optimize hair removal parameters. A hair removal timing prediction model is constructed to predict the optimal hair removal timing, generate a personalized hair removal plan, and adjust and update the user hair removal database in real time.

[0024] A user hair removal database is built based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data. A distributed storage architecture is adopted, and the data update cycle is set to a fixed time interval. The historical hair removal parameters include the hair removal energy density used, the vibration frequency of the hair removal head, and the hair removal time. The skin characteristics include skin thickness, pigmentation, and skin type. The hair removal feedback data includes the user's pain level, comfort, skin reaction, and hair regeneration speed. The user data is feature extracted and encoded through a deep neural network. The user data includes the user's historical hair removal parameters, skin characteristics and hair removal feedback data. Based on the user data and the current hair removal parameters, a deep Q network is built using a reinforcement learning algorithm. Through multi-layer fully connected network processing, the adjustment value of hair removal parameters is obtained; according to the user's hair growth data, a recurrent neural network is used for model training, and a hair removal timing prediction model is constructed to predict the best hair removal time within a preset time period in the future; according to the prediction results of the deep Q network and the hair removal timing prediction model, a personalized hair removal plan is generated, and hair removal guidance suggestions are provided to the user. The personalized hair removal plan includes hair removal parameters and the best hair removal time; the user hair removal database is continuously updated, the user's hair removal effect and user feedback are recorded, the hair removal parameters and hair removal timing predictions are adjusted in real time based on the user's hair removal effect and feedback, and the recurrent neural network and deep Q network model are retrained regularly.

[0025] The present invention provides an intelligent control system for a hair removal device, which mainly includes:

[0026] A three-dimensional structure data acquisition and analysis module, which is used to obtain skin structure data and analyze hair follicle distribution through optical coherence tomography;

[0027] A hair removal area division module is used to adaptively divide the hair removal area according to the hair follicle density;

[0028] The hair removal area simulation module is used to numerically simulate the hair removal process and predict the hair removal accuracy and efficiency values;

[0029] A hair removal parameter optimization module is used to optimize the hair removal parameters and establish a mapping relationship database between the division granularity and the hair removal parameters;

[0030] A hair removal parameter supplement module is used to calculate the hair removal parameters corresponding to different skin characteristic parameters according to a database of mapping relationships between the division granularity and the hair removal parameters;

[0031] Real-time monitoring and parameter adjustment module, used to monitor skin temperature and pressure, and adjust hair removal parameters to improve safety and comfort;

[0032] The personalized hair removal plan generation module is used to generate a personalized hair removal plan based on user data and predict the optimal time for hair removal.

[0033] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0034] The invention discloses an intelligent control method and system for a hair removal device. The invention acquires three-dimensional structural data of different depths of the skin and establishes a three-dimensional distribution model of hair follicles, which can accurately determine the spatial non-uniformity characteristics of the hair follicle density distribution, thereby finely meshing the hair removal area to ensure that the hair removal process is fully covered without missing any area, thereby improving the accuracy and efficiency of hair removal and avoiding the common problems of poor effect and skin damage in traditional methods. By optimizing the hair removal area division scheme, the coupling response of energy loss characteristics and hair removal head vibration characteristics is obtained, and the hair removal effect can be accurately evaluated before hair removal, ensuring high efficiency and accuracy in the actual hair removal process. The method can adjust the hair removal parameters according to different skin and hair characteristics, and improve the hair removal effect and user comfort. In the actual hair removal process, the optical characteristic parameters can be adjusted in real time according to the spectral characteristics of the skin tissue, ensuring that the hair removal process is always in the best state under different skin characteristics. By acquiring the temperature distribution image of the hair removal area in real time, dynamic adjustment is performed in combination with the highest temperature value and the area of ​​the high temperature zone to avoid skin burns, and a user comfort score is constructed in combination with skin deformation data, and the hair removal intensity is adjusted in real time to improve the user's comfort. The present invention can generate personalized hair removal plans by building a user hair removal database, recording the user's historical hair removal parameters, skin characteristics and hair removal feedback data, and adjusting and updating the database in real time. Each user can obtain the hair removal plan that best suits him / her, significantly improving user satisfaction and hair removal effect. The present invention effectively improves the intelligence level and hair removal effect of the hair removal device, ensuring that the hair removal process is always efficient, safe and comfortable under different users and different skin characteristics, meeting the needs of modern users for efficient, safe and comfortable hair removal. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The present invention is a flow chart of an intelligent control method of a hair removal device.

[0036] Figure 2 The figure is a schematic diagram of an intelligent control system of a hair removal device of the present invention.

[0037] Figure 3 This is another schematic diagram of an intelligent control method and system for a hair removal device according to the present invention. DETAILED DESCRIPTION

[0038] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0039] like Figure 1-3The intelligent control method and system of a hair removal device in this embodiment may specifically include:

[0040] Step S101, using optical coherence tomography to obtain three-dimensional structural data of skin at different depths, establishing a three-dimensional distribution model of hair follicles at each layer, and determining the spatial non-uniformity characteristics of the density distribution of hair follicles at each layer.

[0041] The target skin area was scanned at multiple levels by an optical coherence tomography system to obtain a three-dimensional structural image from the skin surface to the deep tissue, and the three-dimensional structural image was denoised by Gaussian filtering. Based on the multi-threshold segmentation of the intensity histogram, the three-dimensional structural image was divided into the epidermis, dermis and subcutaneous tissue layer according to the optical properties of the tissue and the preset optical property threshold. The Canny edge detection algorithm was used to identify the boundaries between the layers to form a three-dimensional digital model of the skin layer. The optical properties included the scattering coefficient and absorption coefficient of the tissue. The Canny edge detection algorithm was applied to the obtained three-dimensional digital model of the skin layer to identify the contour of the hair follicles and calculate the characteristic parameters of the hair follicles, including the length, diameter and inclination angle of the hair follicles. According to the characteristic parameters of the hair follicles, the K-means clustering algorithm was used for model training to construct a three-dimensional distribution model of the hair follicles at each level, and the hair follicles with a similarity of hair follicle characteristic parameters greater than the preset similarity threshold were classified into one category. Based on the established three-dimensional digital model of the skin layer, the skin layer model was divided into several equal-volume grid units, the number of hair follicles in each grid unit was calculated, and the hair follicle density was determined. According to the hair follicle density, a hair follicle density heat map is drawn, and different color depths are used to show the spatial distribution of hair follicle density. The hair follicle density heat map is analyzed using the Moran index calculation method, and the global Moran index and local Moran index are calculated to evaluate the spatial autocorrelation of hair follicle distribution. Among them, the global Moran index calculation formula is: N is the number of hair follicles, x i and x j is the hair follicle density value, is the average hair follicle density, w ij is the weight matrix. The calculation formula of the local Moran index is S0 2 Based on the calculation results of the global Moran index and the local Moran index, a three-dimensional contour surface map reflecting the spatial distribution law of hair follicle density at each layer of the skin is drawn to show the spatial non-uniformity characteristics of hair follicle distribution.

[0042] For example, an ultra-high resolution optical coherence tomography system is used to perform multi-level scanning of the target skin area, with the scanning range set to 10 mm × 10 mm, the depth to 2 mm, and the resolution to 5 μm. The system collects 1000 cross-sectional images, each containing 2000 × 2000 pixels. The multi-threshold segmentation method based on the intensity histogram sets the scattering coefficient threshold to 0.8 cm according to the optical properties of the tissue. -1, the absorption coefficient threshold is 0.4cm -1 , scattering coefficient>0.8cm -1 And the absorption coefficient is <0.4cm -1 Determined to be the epidermis, 0.4cm -1 <scattering coefficient≤0.8cm -1 And 0.2cm -1 <Absorption coefficient≤0.4cm -1 Determined to be the dermis, scattering coefficient

[0043] ≤0.4cm -1 And the absorption coefficient>0.4cm -1 Determined to be subcutaneous tissue. By performing the above classification on each pixel point, the three-dimensional structural image is divided into the epidermis, dermis and subcutaneous tissue layers, and the Canny edge detection algorithm is used to identify the boundaries between the layers to form a three-dimensional digital model of skin stratification. The Canny edge detection algorithm is applied to the acquired three-dimensional digital model of skin stratification to identify the contours of hair follicles and calculate the characteristic parameters of hair follicles, including the length, diameter and inclination angle of the hair follicles. For example, the length of a hair follicle is 1.2 mm, the diameter is 0.3 mm, and the inclination angle is 15 degrees. According to the characteristic parameters of the hair follicles, the K-means clustering algorithm is used for model training to construct a three-dimensional distribution model of hair follicles at each level, and the hair follicles whose similarity of hair follicle characteristic parameters is greater than the preset similarity threshold are classified into one category. For example, the threshold for the similarity of hair follicle characteristic parameters is set to 0.8. Based on the established three-dimensional digital model of skin stratification, the skin stratification model is divided into several equal-volume grid units, for example, the volume of each grid unit is 1 mm 3 , calculate the number of hair follicles in each grid unit and determine the hair follicle density. For example, if the number of hair follicles in a grid unit is 5, the hair follicle density is 5 hair follicles / mm 3 . According to the hair follicle density, a hair follicle density heat map is drawn, and different color shades are used to show the spatial distribution of hair follicle density. For example, high-density areas are represented by red, and low-density areas are represented by blue. The hair follicle density heat map is analyzed using the Moran's index calculation method, and the global Moran's index and local Moran's index are calculated to evaluate the spatial autocorrelation of the hair follicle distribution. Assume that the calculated global Moran's index I is 0.25, indicating that the hair follicle distribution has a certain degree of spatial aggregation. Among them, the global Moran's index calculation formula is N is the number of hair follicles, x i and x j is the hair follicle density value, is the average hair follicle density, w ij is the weight matrix. The calculation formula of the local Moran index is S0 2Based on the calculation results of the global Moran's index and the local Moran's index, a three-dimensional contour surface map reflecting the spatial distribution law of hair follicle density at each layer of the skin was drawn to show the spatial non-uniformity of hair follicle distribution. The three-dimensional contour surface map showed that the hair follicle density of the epidermis was more evenly distributed, while the hair follicle density of the dermis and subcutaneous tissue layer had obvious spatial non-uniformity.

[0044] Step S102, gridding the hair removal area according to the spatial non-uniformity characteristics of the hair follicle density distribution of each layer, dividing the grid into dense areas and sparse areas by calculating the hair follicle density value of each grid, and obtaining a hair removal area division scheme that matches the hair follicle distribution.

[0045] According to the spatial non-uniformity characteristics of the hair follicle density distribution in each layer, based on the preset initial grid size, the quadtree algorithm is used to perform initial grid division on the hair removal area, and the entire hair removal area is divided into n×m initial grids, and each initial grid is assigned a unique identifier. The hair follicle density value in each initial grid is calculated, and the hair follicle density in the grid exceeds the preset density threshold according to the preset density threshold. If it exceeds, the grid is marked as a dense area, otherwise it is marked as a sparse area. For the grid marked as a dense area, it is divided into four sub-grids, and the hair follicle density and area marking are re-performed on each sub-grid. If the hair follicle density in the sub-grid still exceeds the preset density threshold, the sub-grid is further divided into smaller sub-grids until the grid size reaches the preset grid size threshold or the hair follicle density in the grid area is lower than the preset density threshold. The boundary points of all adjacent grids are identified. For each boundary point, the hair follicle density values ​​of the four surrounding grid center points are taken, and the interpolation density of the boundary point is calculated using the bilinear interpolation formula. The boundary position is adjusted according to the interpolation density value to generate a smooth grid boundary. A hair removal area division scheme is generated according to the smoothed mesh boundary, and the hair removal area division scheme is converted into a vector graphics file in SVG or DXF format.

[0046] For example, according to the spatial non-uniformity of the density distribution of hair follicles in each layer, the quadtree algorithm is used to perform initial grid division on the 100mm×100mm hair removal area. The initial grid size is set to 10mm×10mm, and the entire hair removal area is divided into 10×10 initial grids, for a total of 100. Each initial grid is assigned a unique identifier, such as A1 to J10. The hair follicle density value of each grid is calculated. If the number of hair follicles in grid A1 is 60 and the area is 100mm 2 , then its density is 6 pieces / mm 2 . Set the density threshold to 5 pieces / mm 2, judge that the hair follicle density in the A1 grid exceeds the threshold and mark it as a dense area. For the A1 grid marked as a dense area, divide it into four 5mm×5mm sub-grids A1-1, A1-2, A1-3 and A1-4. Repeat the density calculation and area marking process for these four sub-grids. If the density of A1-1 is 7 / mm 2 , and then divided into four 2.5mm×2.5mm sub-grids. If the density of A1-1-1 is 4 / mm 2 , below the threshold, stop dividing. Perform boundary smoothing on all divided grids, and use bilinear interpolation to calculate the transition area of ​​adjacent grids. Identify the boundary point (5,0) between A1-1 and A1-2, and take the four surrounding grid center points (2.5,2.5), (7.5,2.5), (2.5,7.5),

[0047] The hair follicle density values ​​of (7.5,7.5) are 7, 6, 5, and 4 / mm respectively. 2 The interpolation density of the boundary point (5,0) calculated using the bilinear interpolation formula is 5.5 / mm 2 According to the interpolation density value, the boundary is adjusted by 0.5mm in the direction of A1-2 with lower density to generate a smooth mesh boundary. Finally, the hair removal area division scheme is generated, and the division result is converted into a vector graphics file in SVG or DXF format for precise positioning and energy distribution of subsequent hair removal equipment.

[0048] Step S103, according to the hair removal area division scheme, the energy loss characteristics at the division boundary and the coupling response characteristics of the hair removal head vibration and the division granularity are obtained, and a hair removal accuracy and efficiency prediction model is established to predict the hair removal accuracy and efficiency values.

[0049] According to the hair removal area division plan, the finite element method is used to construct a three-dimensional simulation model, and the grids of different granularities are divided into three sizes, including size A, size B, and size C. The corresponding hair removal parameters are set for each granularity of the grid. The hair removal parameters include hair removal energy density, hair removal head vibration frequency, and hair removal time. Using the Fourier heat conduction equation The hair removal process was numerically simulated, where T is temperature, t is time, and α is the thermal diffusion coefficient. The finite difference method was used to discretize the equation to set boundary conditions and initial conditions, and the iterative solver was used to calculate the temperature distribution. Based on the temperature gradient and heat flux density at the grid boundary, the boundary energy loss characteristic curve was obtained. The boundary conditions included the temperature and heat flux density of the skin surface, and the initial condition was the initial temperature distribution. The ANSYS Mechanical software was used to simulate the vibration characteristics of the hair removal head, establish the geometric model of the hair removal head, set the material properties and boundary conditions, and perform modal analysis to calculate the natural frequency and vibration shape of the hair removal head according to the skin contact area and damping coefficient under different division granularity. The frequency response function was analyzed by harmonic response, and the vibration frequency, amplitude and phase information were obtained to obtain the coupling response data of the division granularity and the hair removal head vibration. According to the energy loss characteristic curve and the vibration coupling response data, the recurrent neural network algorithm was used for model training, and the hair removal accuracy and efficiency prediction model was established. According to the input division granularity, the hair removal accuracy and efficiency values ​​were predicted.

[0050] For example, according to the hair removal area division scheme, a 3D simulation model is constructed using the finite element method, and the grid sizes are divided into size A 1mm×1mm, size B 2mm×2mm and size C 5mm×5mm. Corresponding hair removal parameters are set for each size of grid, where the hair removal energy density of size A is 10J / cm 2 The hair removal head has a vibration frequency of 100Hz and a hair removal time of 5 seconds. The hair removal energy density of size B is 8J / cm 2 The hair removal head has a vibration frequency of 80Hz and a hair removal time of 4 seconds. The hair removal energy density of size C is 6J / cm 2 , the vibration frequency of the hair removal head is 60Hz, and the hair removal time is 3 seconds. Using the Fourier heat conduction equation The hair removal process is numerically simulated, where T is temperature, t is time, α is the thermal diffusion coefficient, is the Laplace operator of temperature T. The spatial part of the heat conduction equation is discretized by the finite difference method, the nodes are calculated using the grid, and the time part is discretized by the implicit time stepping method. The boundary conditions and initial conditions are set, where the temperature of the skin surface is 37°C and the heat flux is 0W / m 2, the initial temperature distribution is uniform at 37℃. The temperature distribution is calculated using an iterative solver, and the boundary energy loss characteristic curve is obtained based on the temperature gradient and heat flux density at the grid boundary. The vibration characteristics of the hair removal head are simulated using ANSYS Mechanical software, the geometric model of the hair removal head is established, the material properties and boundary conditions are set, and the modal analysis is performed to calculate the natural frequency and vibration mode of the hair removal head. If under size A, the natural frequency is 200Hz and the vibration mode is the first mode. Under size B, the natural frequency is 150Hz and the vibration mode is the second mode. Under size C, the natural frequency is 100Hz and the vibration mode is the third mode. Harmonic response analysis is performed to obtain the frequency response function, obtain the vibration frequency, amplitude and phase information, and find that the vibration frequency under size A is 95Hz, the amplitude is 0.02mm, and the phase is 30 degrees. According to the energy loss characteristic curve and vibration coupling response data, the recurrent neural network algorithm is used for model training to establish a hair removal accuracy and efficiency prediction model. If the granularity is size B, the hair removal accuracy and efficiency prediction model predicts that the hair removal accuracy is 85% and the efficiency is 90%. Based on these results, the hair removal parameters can be adjusted to optimize the hair removal effect and efficiency for specific hair removal areas and conditions.

[0051] Step S104, according to the generated hair removal area division scheme, the division granularity of each division area and the corresponding initial value of the hair removal parameter are obtained, the hair removal parameters are optimized using a particle swarm optimization algorithm, and a mapping relationship database between the division granularity and the hair removal parameters is constructed.

[0052] According to the generated hair removal area division scheme, the division granularity of each division area is obtained. According to the granularity information of each division area, the initial value of the hair removal parameter corresponding to the division granularity of the corresponding hair removal area is obtained. According to the initial hair removal parameters of each division area, the particle swarm optimization algorithm is used to optimize the hair removal parameters, and the particle swarm scale is set to use the particle swarm optimization algorithm to solve the nonlinear relationship model, set the population scale to a, the maximum number of iterations to b, the inertia weight to c, and the acceleration factor to d. The hair removal parameters of each division area are optimized to obtain the optimal hair removal parameters of each division area. According to the division granularity of the hair removal area and the corresponding optimal hair removal parameters, a continuous mapping relationship between the division granularity of the hair removal area and the optimal hair removal parameters is established, and a mapping relationship database between the division granularity and the hair removal parameters is constructed.

[0053] For example, according to the generated hair removal area division scheme, the division granularity of each division area is obtained, and a hair removal area is divided into three grids with different granularity of 1mm×1mm, 2mm×2mm and 5mm×5mm. According to the granularity information of each division area, the corresponding initial hair removal parameters are obtained, wherein the initial hair removal energy density is 10J / cm 2 , 8J / cm 2 , 6J / cm2 The vibration frequencies of the hair removal heads are 100Hz, 80Hz, and 60Hz, respectively, and the hair removal times are 5 seconds, 4 seconds, and 3 seconds, respectively. Based on these initial hair removal parameters, the particle swarm optimization algorithm is used to optimize the hair removal parameters. The population size is set to 50, the maximum number of iterations is 1000, the inertia weight is 0.9, and the acceleration factor is 2.0. The hair removal parameters of each divided area are optimized. After optimization, in the area with a divided granularity of 1mm×1mm, the optimal hair removal parameter is a hair removal energy density of 12J / cm 2 The hair removal head has a vibration frequency of 110Hz and a hair removal time of 4.5 seconds. In an area with a particle size of 2mm×2mm, the optimal hair removal parameter is a hair removal energy density of 10J / cm 2 The hair removal head has a vibration frequency of 90Hz and a hair removal time of 3.8 seconds. In an area with a particle size of 5mm×5mm, the optimal hair removal parameter is a hair removal energy density of 7J / cm 2 , the vibration frequency of the epilator head is 70Hz, and the epilator time is 3.2 seconds. Based on these optimal epilator parameters, a continuous mapping relationship between the division granularity of the epilator area and the optimal epilator parameters is established, and a database of the mapping relationship between the division granularity and the epilator parameters is constructed. Therefore, in an area with a granularity of 1mm×1mm, the optimal epilator energy density of 12J / cm 2 , the vibration frequency is 110Hz, and the time is 4.5 seconds, and these parameters are applied to the actual hair removal process to ensure the best hair removal effect.

[0054] Step S105, according to the spectral characteristics of skin tissue, the optical characteristic parameters are adjusted in real time and Monte Carlo simulation is performed, and the hair removal parameters corresponding to different skin characteristic parameters are calculated by combining adaptability judgment and a database of mapping relationships between division granularity and hair removal parameters.

[0055] The spectrometer is used to obtain the skin tissue reflectance spectrum in real time, and the sampling wavelength range, sampling interval and sampling frequency are set to obtain the real-time reflectance spectrum data of the skin tissue. Based on the reflectance spectrum data and the three-dimensional digital model of skin stratification, the initial optical characteristic parameter value is set, and the forward Monte Carlo simulation is performed to calculate the initial simulated reflectance spectrum. The difference between the simulated reflectance spectrum and the measured reflectance spectrum is compared to calculate the reflectance spectrum error. The optical characteristic parameter value is adjusted according to the magnitude of the reflectance spectrum error, and the forward Monte Carlo simulation is performed again to update the simulated reflectance spectrum. The forward Monte Carlo simulation is repeated until the error between the simulated reflectance spectrum and the measured reflectance spectrum is less than the preset error threshold, and the optical characteristic parameter value of the skin tissue is determined. According to the skin tissue characteristic parameters and the corresponding adaptability label data, the support vector machine algorithm is used for model training to determine the adaptability of the skin to the current hair removal parameters. If the adaptability of the skin to the current hair removal parameters is low, the hair removal parameters are adjusted, and the current skin characteristic parameters are used as query conditions to find matching hair removal parameters from the mapping relationship database between the division granularity and the hair removal parameters. If the same skin characteristic parameters do not exist in the mapping relationship database between the division granularity and the hair removal parameters, the Euclidean distance between the current skin characteristic parameters and the sample points in the database is calculated, and K sample points whose Euclidean distances are less than a preset distance threshold are determined. The hair removal parameters of these K sample points are weighted averaged to obtain the hair removal parameters corresponding to different skin characteristic parameters, and the hair removal parameters corresponding to different skin characteristic parameters are updated to the hair removal device control unit.

[0056] For example, an OceanOpticsHR4000 high-precision spectrometer is used to collect the skin tissue reflectance spectrum in real time, and the sampling wavelength range is set to 400-1000nm, the sampling interval is 1nm, and the sampling frequency is 10Hz to obtain the real-time reflectance spectrum data of the skin tissue. Based on the reflectance spectrum data and the three-dimensional digital model of the skin layer, the initial optical characteristic parameter values ​​are set, such as the absorption coefficient is 0.1cm -1 , the scattering coefficient is 1.5cm -1 , perform forward Monte Carlo simulation, calculate the initial simulated reflection spectrum, compare the difference between the simulated reflection spectrum and the measured reflection spectrum, and calculate the reflection spectrum error to be 0.05. Adjust the optical characteristic parameter value according to the reflection spectrum error, such as adjusting the absorption coefficient to 0.12cm -1 , the scattering coefficient is adjusted to 1.6cm -1 , re-perform forward Monte Carlo simulation and update the simulated reflection spectrum. Repeat the forward Monte Carlo simulation until the error between the simulated reflection spectrum and the measured reflection spectrum is less than the preset error threshold of 0.01, and finally determine the optical characteristic parameter value of skin tissue to be the absorption coefficient of 0.11cm -1 , scattering coefficient 1.55cm -1According to the skin tissue characteristic parameters and the corresponding adaptability label data, the support vector machine algorithm is used for model training to determine the adaptability of the skin to the current hair removal parameters. If the current hair removal parameter is an energy density of 12J / cm 2 , vibration frequency 100Hz, hair removal time 5s, model predicts low adaptability. If the skin's adaptability to the current hair removal parameters is low, the hair removal parameters are adjusted, and the current skin characteristic parameters are used as query conditions to find matching hair removal parameters from the database of mapping relationships between division granularity and hair removal parameters. If the same skin characteristic parameters do not exist in the database, the Euclidean distance between the current skin characteristic parameters and the sample points in the database is calculated, and the five sample points whose Euclidean distance is less than the preset distance threshold are determined. The hair removal parameters of these five sample points are weighted averaged, and the hair removal parameters corresponding to different skin characteristic parameters are calculated to be an energy density of 10.5J / cm 2 , vibration frequency 95Hz, hair removal time 4.8s, and the hair removal parameters corresponding to different skin characteristic parameters are updated to the hair removal device control unit.

[0057] Step S106, using an infrared thermal imaging camera to obtain a temperature distribution image of the hair removal area in real time, combining the maximum temperature value and the area of ​​the high temperature zone to determine and adjust the hair removal energy output, and building a user comfort score prediction model based on skin deformation data to adjust the hair removal intensity in real time.

[0058] An infrared thermal imaging camera is used to set the acquisition frequency to obtain the temperature distribution image of the hair removal area in real time. The temperature distribution image is segmented by the Otsu adaptive threshold method to divide the image into a high temperature area and a low temperature area. The maximum temperature value and the area of ​​the high temperature area are extracted to determine whether the maximum temperature value exceeds the preset temperature threshold and whether the area of ​​the high temperature area exceeds the preset area threshold. If the maximum temperature value or the area of ​​the high temperature area exceeds the preset threshold, the fuzzy control algorithm is used to adaptively adjust the hair removal energy output according to the temperature distribution. The input variables are set as the maximum temperature and the high temperature area, and the output variable is the energy adjustment coefficient. Fuzzy rules are defined to obtain energy adjustment instructions to dynamically control the hair removal energy. Skin deformation data and the corresponding user comfort score are obtained, and a recurrent neural network algorithm is used for model training to establish a user comfort score prediction model. The skin deformation data during the hair removal process is obtained through a pressure sensor array, and the user comfort score prediction model is used to predict the user comfort score in real time. If the predicted user comfort score is lower than the preset comfort score threshold, the hair removal intensity is dynamically adjusted according to the user comfort score.

[0059] For example, a FL IR A655sc infrared thermal imaging camera is used to collect temperature distribution images of the hair removal area in real time, and the acquisition frequency is set to 30 Hz, the spatial resolution is 640×480 pixels, and the temperature accuracy is

[0060] ±0.1℃, to obtain real-time temperature data of the skin surface. Based on the temperature data, the cv2.threshold function in the OpenCV library is used to implement the Otsu adaptive threshold method to segment the temperature distribution image, dividing the image into high temperature area and low temperature area, extracting the highest temperature value and area of ​​the high temperature area. The highest temperature value is 45℃, and the area of ​​the high temperature area is 20cm 2 , determine whether the maximum temperature value exceeds the preset temperature threshold of 40°C, and whether the high temperature area exceeds the preset area threshold of 15cm 2 . Since both the maximum temperature value and the high temperature area exceed the preset threshold, the fuzzy control algorithm is used to adaptively adjust the hair removal energy output according to the temperature distribution. The input variables are set as the maximum temperature and the high temperature area, and the output variable is the energy adjustment coefficient. The fuzzy rules are defined. If the maximum temperature is high and the high temperature area is large, the energy adjustment coefficient is greatly reduced. If the maximum temperature is medium and the high temperature area is small, the energy adjustment coefficient is slightly reduced. Through fuzzy reasoning to process these rules, the energy adjustment instructions are obtained, and it is decided to reduce the hair removal energy output by 20%. At the same time, the skin deformation data and the corresponding user comfort score are obtained. If the comfort score corresponding to the data of a pressure sensor array is 0.7 in the training data. The recurrent neural network algorithm is used for model training to establish a user comfort score prediction model. The skin deformation data of the hair removal process is obtained in real time through the pressure sensor array. For example, in a hair removal process, the skin deformation data recorded by the pressure sensor is 0.5mm. The user comfort score prediction model is used to predict the user comfort score to be 0.5. Since the predicted user comfort score is lower than the preset comfort score threshold of 0.6, the hair removal intensity is dynamically adjusted according to the user comfort score, and the hair removal intensity is reduced by 10%. Through the above process, it is ensured that the hair removal energy and intensity are dynamically adjusted according to real-time data during the hair removal process to optimize the hair removal effect and user comfort.

[0061] Step S107, constructing a user hair removal database based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data, using deep neural network and reinforcement learning algorithm to extract features and optimize hair removal parameters, and constructing a hair removal timing prediction model to predict the best hair removal timing, generate a personalized hair removal plan, and adjust and update the user hair removal database in real time.

[0062] A user hair removal database is constructed based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data. A distributed storage architecture is adopted, and the data update cycle is set to a fixed time interval. The historical hair removal parameters include the hair removal energy density used, the vibration frequency of the hair removal head, and the hair removal time. The skin characteristics include skin thickness, pigmentation, and skin type. The hair removal feedback data includes the user's pain level, comfort, skin reaction, and hair regeneration speed. The user data is feature extracted and encoded through a deep neural network. The user data includes the user's historical hair removal parameters, skin characteristics, and hair removal feedback data. According to the user data and the current hair removal parameters, a deep Q network is constructed using a reinforcement learning algorithm, and the hair removal parameter adjustment value is obtained through processing through a multi-layer fully connected network. According to the user's hair growth data, a recurrent neural network is used for model training to construct a hair removal timing prediction model to predict the best hair removal time in the future preset time period. According to the prediction results of the deep Q network and the hair removal timing prediction model, a personalized hair removal plan is generated, and hair removal guidance suggestions are provided to the user. The personalized hair removal plan includes hair removal parameters and the best hair removal time. Continuously update the user hair removal database, record the user's hair removal effects and user feedback, adjust the hair removal parameters and hair removal timing prediction in real time based on the user's hair removal effects and feedback, and regularly retrain the recurrent neural network and deep Q network models.

[0063] Exemplarily, the Apache Cassandra distributed storage system is used to construct a user hair removal database, and a 10-node cluster is set up, each node is equipped with a 16-core CPU, 64GB memory and 2TB SSD storage. Historical hair removal parameters include the hair removal energy density used, the vibration frequency of the hair removal head, and the hair removal time. Skin characteristics include skin thickness, pigmentation, and skin type. Hair removal feedback data includes the user's pain level, comfort, skin reaction, and hair regeneration rate. User data is feature extracted and encoded through a deep neural network. User data includes the user's historical hair removal parameters, skin characteristics, and hair removal feedback data. Based on the user data and the current hair removal parameters, a deep Q network is constructed using a reinforcement learning algorithm, and processed through a multi-layer fully connected network to obtain the hair removal parameter adjustment value, for example, adjusting the energy density by 10J / cm 2, the vibration frequency is adjusted to 90Hz, and the hair removal time is adjusted to 4.5 minutes. According to the user's hair growth data, a recurrent neural network is used for model training to build a hair removal timing prediction model to predict the best hair removal time in the preset time period in the future, such as predicting that the next best hair removal time is 3 weeks later. Based on the prediction results of the deep Q network and the hair removal timing prediction model, a personalized hair removal plan is generated, and hair removal guidance suggestions are provided to the user. The personalized hair removal plan includes adjusted hair removal parameters and the best hair removal time is 3 weeks later. Continuously update the user hair removal database to record the user's hair removal effect and user feedback, such as recording user feedback as a pain level of 0.2, a comfort level of 0.8, a skin reaction of no redness and swelling, and a hair regeneration rate of 0.8mm per week. Adjust the hair removal parameters and hair removal timing prediction in real time based on the user's hair removal effect and feedback, and regularly retrain the recurrent neural network and deep Q network models to ensure the optimization and effectiveness of the personalized hair removal plan.

[0064] What is disclosed above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. A person skilled in the art can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. An intelligent control system for a hair removal device, characterized in that: The system comprises: The hair follicle characteristic parameter acquisition and analysis module is used to identify the skin layer boundary by optical coherence tomography method, obtain the hair follicle characteristic parameters of different skin layers, establish the three-dimensional distribution model of hair follicles in each layer, and determine the spatial non-uniformity characteristics of the density distribution of hair follicles in each layer; The hair removal area division module is used to divide the hair removal area into grids according to the spatial non-uniformity characteristics of the hair follicle density distribution at each skin layer, and obtain a hair removal area division scheme that matches the hair follicle distribution by calculating the hair follicle density value of each grid; The hair removal area simulation module is used to obtain the energy loss characteristics at the division boundary and the coupling response characteristics of the hair removal head vibration and the division granularity according to the hair removal area division plan, establish a hair removal accuracy and efficiency prediction model, and predict the hair removal accuracy and efficiency values; The hair removal parameter optimization module is used to obtain the division granularity of each division area and the corresponding initial value of the hair removal parameter according to the generated hair removal area division scheme, and to build a mapping relationship database between the division granularity and the hair removal parameter; A hair removal parameter supplement module is used to calculate the hair removal parameters corresponding to different skin characteristic parameters based on the reflectance spectrum data of skin tissue, combined with adaptability judgment and a database of mapping relationships between the division granularity and hair removal parameters; Real-time monitoring and parameter adjustment module, used to obtain the temperature distribution image of the hair removal area, judge and adjust the hair removal energy output based on the maximum temperature value and the area of ​​the high-temperature zone, and build a user comfort score prediction model based on skin deformation data to adjust the hair removal intensity in real time; The personalized hair removal plan generation module is used to build a user hair removal database based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data, perform feature extraction and hair removal parameter optimization, and build a hair removal timing prediction model to predict the optimal hair removal timing and generate a personalized hair removal plan.

2. The system according to claim 1, wherein: The hair follicle characteristic parameter acquisition and analysis module is used to identify the skin layer boundary by using the optical coherence tomography method, and obtain the hair follicle characteristic parameters of different skin layers, establish a three-dimensional distribution model of hair follicles in each layer, and determine the spatial non-uniformity characteristics of the hair follicle density distribution in each layer, including: An optical coherence tomography system is used to perform multi-level scanning of the target skin area to obtain a three-dimensional structural image from the skin surface to the deep tissue, and a Gaussian filter is used to perform denoising on the three-dimensional structural image; Based on the multi-threshold segmentation of intensity histogram, the three-dimensional structural image is divided into the epidermis, dermis and subcutaneous tissue layer according to the optical properties of the tissue and the preset optical property threshold. The Canny edge detection algorithm is used to identify the boundaries between the layers to form a three-dimensional digital model of the skin layer. The optical properties include the scattering coefficient and absorption coefficient of the tissue. The Canny edge detection algorithm is applied to the obtained three-dimensional digital model of the skin layer to identify the contour of the hair follicle and calculate the characteristic parameters of the hair follicle, including the length, diameter and inclination angle of the hair follicle. According to the characteristic parameters of the hair follicle, the K-means clustering algorithm is used for model training to construct each The three-dimensional distribution model of hierarchical hair follicles is established, and the hair follicles whose hair follicle characteristic parameter similarity is greater than the preset similarity threshold are classified into one category; based on the established three-dimensional digital model of skin stratification, the skin stratification model is divided into several equal-volume grid units, the number of hair follicles in each grid unit is calculated, and the hair follicle density is determined; according to the hair follicle density, a hair follicle density heat map is drawn, and different color depths are used to display the spatial distribution of hair follicle density; the hair follicle density heat map is analyzed using the Moran index calculation method, and the global Moran index and local Moran index are calculated to evaluate the spatial autocorrelation of hair follicle distribution; among which, the global Moran index calculation formula is: , N is the number of hair follicles, x i and x j is the hair follicle density value, is the average hair follicle density, w ij is the weight matrix; the local Moran index calculation formula is , S0 2 is the overall variance; based on the calculation results of the global Moran's index and the local Moran's index, a three-dimensional contour surface map reflecting the spatial distribution law of hair follicle density at each layer of the skin is drawn to show the spatial non-uniformity characteristics of hair follicle distribution.

3. The system according to claim 1, wherein: The hair removal area division module is used to divide the hair removal area into grids according to the spatial non-uniformity characteristics of the hair follicle density distribution at each skin layer, and obtain a hair removal area division scheme that matches the hair follicle distribution by calculating the hair follicle density value of each grid, including: According to the spatial non-uniformity of the hair follicle density distribution at each skin layer, based on the preset initial grid size, the quadtree algorithm is used to perform initial grid division on the hair removal area, and the entire hair removal area is divided into n×m initial grids, and each initial grid is assigned a unique identifier; Calculate the hair follicle density value in each initial grid, and determine whether the hair follicle density in the grid exceeds the preset density threshold according to the preset density threshold. If it exceeds, mark the grid as a dense area, otherwise mark it as a sparse area; For the grid marked as a dense area, it is divided into four sub-grids, and the hair follicle density is recalculated and the area is marked for each sub-grid; if the hair follicle density in the sub-grid still exceeds the preset density threshold, the sub-grid is continued to be divided into smaller sub-grids until the grid size reaches the preset grid size threshold or the hair follicle density in the grid area is lower than the preset density threshold; the boundary points of all adjacent grids are identified, and for each boundary point, the hair follicle density value of the four surrounding grid center points is taken, and the interpolation density of the boundary point is calculated using a bilinear interpolation formula, and the boundary position is adjusted according to the interpolation density value to generate a smooth grid boundary; based on the smoothed grid boundary, a hair removal area division scheme is generated, and the hair removal area division scheme is converted into a vector graphics file in SVG or DXF format.

4. The system according to claim 1, wherein: The depilatory area simulation module is used to obtain the energy loss characteristics at the division boundary and the coupling response characteristics of the depilatory head vibration and the division granularity according to the depilatory area division scheme, establish a depilatory accuracy and efficiency prediction model, and predict the depilatory accuracy and efficiency values, including: according to the depilatory area division scheme, use the finite element method to build a three-dimensional simulation model, divide the grids of different granularities, including three sizes of size A, size B, and size C, set corresponding depilatory parameters for each granularity of the grid, and the depilatory parameters include depilatory energy density, depilatory head vibration frequency and depilatory time; use the Fourier heat conduction equation to calculate the depilatory energy density, the depilatory head vibration frequency and the depilatory time; , the hair removal process is numerically simulated, where T is temperature, t is time, and α is the thermal diffusion coefficient; the finite difference method is used to discretize the equation to set the boundary conditions and initial conditions, and the iterative solver is used to calculate the temperature distribution. Based on the temperature gradient and heat flux density at the grid boundary, the boundary energy loss characteristic curve is obtained. The boundary conditions include the temperature and heat flux density of the skin surface, and the initial condition is the initial temperature distribution; ANSYS Mechanical software is used to simulate the vibration characteristics of the hair removal head, establish the geometric model of the hair removal head, set the material properties and boundary conditions, and perform modal analysis to calculate the natural frequency and vibration mode of the hair removal head according to the skin contact area and damping coefficient under different division granularity. The frequency response function is performed to obtain the vibration frequency, amplitude and phase information, and obtain the coupling response data of the division granularity and the hair removal head vibration; According to the energy loss characteristic curve and the vibration coupling response data, the recurrent neural network algorithm is used for model training, and a hair removal accuracy and efficiency prediction model is established. According to the input division granularity, the hair removal accuracy and efficiency values ​​are predicted.

5. The system according to claim 1, wherein: The hair removal parameter optimization module is used to obtain the division granularity of each divided area and the corresponding initial value of the hair removal parameter according to the generated hair removal area division scheme, and to construct a mapping relationship database between the division granularity and the hair removal parameter, including: According to the generated hair removal area division scheme, the division granularity of each division area is obtained; according to the granularity information of each division area, the initial value of the hair removal parameter corresponding to the division granularity of the corresponding hair removal area is obtained; according to the initial hair removal parameters of each division area, the particle swarm optimization algorithm is used to optimize the hair removal parameters, and the particle swarm scale is set to use the particle swarm optimization algorithm to solve the nonlinear relationship model, set the population size to a, the maximum number of iterations to b, the inertia weight to c, and the acceleration factor to d, and the hair removal parameters of each division area are optimized to obtain the optimal hair removal parameters of each division area; according to the division granularity of the hair removal area and the corresponding optimal hair removal parameters, a continuous mapping relationship between the division granularity of the hair removal area and the optimal hair removal parameters is established, and a mapping relationship database between the division granularity and the hair removal parameters is constructed.

6. The system according to claim 1, wherein: The hair removal parameter supplement module is used to calculate the hair removal parameters corresponding to different skin characteristic parameters according to the reflectance spectrum data of the skin tissue, combined with the adaptability judgment and the mapping relationship database between the division granularity and the hair removal parameters, including: A spectrometer is used to obtain the reflectance spectrum of skin tissue in real time, and a sampling wavelength range, a sampling interval and a sampling frequency are set to obtain real-time reflectance spectrum data of skin tissue; based on the reflectance spectrum data and a three-dimensional digital model of skin stratification, an initial optical characteristic parameter value is set, and a forward Monte Carlo simulation is performed to calculate an initial simulated reflectance spectrum; the difference between the simulated reflectance spectrum and the measured reflectance spectrum is compared to calculate the reflectance spectrum error; the optical characteristic parameter value is adjusted according to the magnitude of the reflectance spectrum error, and the forward Monte Carlo simulation is performed again to update the simulated reflectance spectrum; the forward Monte Carlo simulation is repeated until the error between the simulated reflectance spectrum and the measured reflectance spectrum is less than a preset error threshold, and the optical characteristic parameter value of the skin tissue is determined; According to the skin tissue characteristic parameters and the corresponding adaptability label data, the support vector machine algorithm is used to perform model training to determine the adaptability of the skin to the current hair removal parameters; if the adaptability of the skin to the current hair removal parameters is low, the hair removal parameters are adjusted, and the current skin characteristic parameters are used as query conditions to find matching hair removal parameters from the database of mapping relationships between division granularity and hair removal parameters; if the same skin characteristic parameters do not exist in the database of mapping relationships between division granularity and hair removal parameters, the Euclidean distance between the current skin characteristic parameters and the sample points in the database is calculated, and K sample points whose Euclidean distances are less than a preset distance threshold are determined, and the hair removal parameters of these K sample points are weighted averaged to calculate the hair removal parameters corresponding to different skin characteristic parameters, and the hair removal parameters corresponding to different skin characteristic parameters are updated to the hair removal device control unit.

7. The system according to claim 1, wherein: The real-time monitoring and parameter adjustment module is used to obtain the temperature distribution image of the hair removal area, judge and adjust the hair removal energy output in combination with the maximum temperature value and the area of ​​the high temperature zone, and build a user comfort score prediction model based on the skin deformation data to adjust the hair removal intensity in real time, including: An infrared thermal imaging camera is used to set the acquisition frequency to obtain the temperature distribution image of the hair removal area in real time; the temperature distribution image is segmented by the Otsu adaptive threshold method to divide the image into a high temperature area and a low temperature area; the maximum temperature value and the area of ​​the high temperature area are extracted to determine whether the maximum temperature value exceeds the preset temperature threshold and whether the area of ​​the high temperature area exceeds the preset area threshold; If the maximum temperature value or the area of ​​the high-temperature zone exceeds the preset threshold, the fuzzy control algorithm is used to adaptively adjust the hair removal energy output according to the temperature distribution. The input variables are set as the maximum temperature and the high-temperature area, and the output variable is the energy adjustment coefficient. Fuzzy rules are defined to obtain energy adjustment instructions and dynamically control the hair removal energy. The skin deformation data and the corresponding user comfort score are obtained, and the recurrent neural network algorithm is used for model training to establish a user comfort score prediction model. The skin deformation data during the hair removal process is obtained through the pressure sensor array, and the user comfort score prediction model is used to predict the user comfort score in real time. If the predicted user comfort score is lower than the preset comfort score threshold, the hair removal intensity is dynamically adjusted according to the user comfort score.

8. The system according to claim 1, wherein: The personalized hair removal scheme generation module is used to build a user hair removal database based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data, perform feature extraction and hair removal parameter optimization, and build a hair removal timing prediction model to predict the best hair removal timing and generate a personalized hair removal scheme, including: A user hair removal database is built based on the user's historical hair removal parameters, skin characteristics and hair removal feedback data. A distributed storage architecture is adopted, and the data update cycle is set to a fixed time interval. The historical hair removal parameters include the hair removal energy density used, the vibration frequency of the hair removal head, and the hair removal time. The skin characteristics include skin thickness, pigmentation, and skin type. The hair removal feedback data includes the user's pain level, comfort, skin reaction, and hair regeneration speed. The user data is feature extracted and encoded through a deep neural network. The user data includes the user's historical hair removal parameters, skin characteristics and hair removal feedback data. Based on the user data and the current hair removal parameters, a deep Q network is built using a reinforcement learning algorithm. Through multi-layer fully connected network processing, the adjustment value of hair removal parameters is obtained; according to the user's hair growth data, a recurrent neural network is used for model training, and a hair removal timing prediction model is constructed to predict the best hair removal time within a preset time period in the future; according to the prediction results of the deep Q network and the hair removal timing prediction model, a personalized hair removal plan is generated, and hair removal guidance suggestions are provided to the user. The personalized hair removal plan includes hair removal parameters and the best hair removal time; the user hair removal database is continuously updated, the user's hair removal effect and user feedback are recorded, the hair removal parameters and hair removal timing predictions are adjusted in real time based on the user's hair removal effect and feedback, and the recurrent neural network and deep Q network model are retrained regularly.

Images