Skin quality detection method, skin quality detection device, and computer-readable storage medium

Abstract

The present application provides a kind of skin quality detection method, skin quality detection device and computer readable storage medium.The skin quality detection method includes: obtaining the spectral image data and standard instrument detection data of skin sample;The spectral image data of skin sample is preprocessed;According to standard instrument detection data and the spectral image data of skin sample after preprocessing, construct skin quality detection model;Obtain the spectral image data of skin to be detected;Skin spectral image data to be detected is input into skin quality detection model, and multi-dimensional skin quality detection result is output.The skin quality detection method of the present application can accurately, efficiently and reliably detect the skin quality of real skin in depth and multidimension, support non-contact detection, support whole or local detection, and can improve user experience.

Description

Technical Field

[0001] This invention relates generally to the field of spectroscopy, and more particularly to a skin texture detection method, a skin texture detection device, and a computer-readable storage medium. Background Technology

[0002] Traditional skin texture analysis methods can detect both real and fake skin, or provide a rough assessment of skin condition, but they struggle to achieve accurate, reliable, and efficient results, let alone in-depth, multi-dimensional analysis. Therefore, how to accurately, reliably, and efficiently perform in-depth, multi-dimensional skin texture analysis on real skin is the technical problem this invention aims to solve.

[0003] The content of the background section is merely the technology known to the inventor and does not necessarily represent the prior art in this field. Summary of the Invention

[0004] In view of one or more of the problems existing in the prior art, the present invention provides a skin texture detection method, a skin texture detection device, and a computer-readable storage medium, which can accurately, efficiently, and reliably perform in-depth and multi-dimensional detection of the skin texture of real skin, support non-contact detection, support overall or partial detection, and improve user experience.

[0005] A first aspect of the present invention provides a skin texture detection method. The skin texture detection method includes: acquiring spectral image data of a skin sample and standard instrument detection data; preprocessing the spectral image data of the skin sample; constructing a skin texture detection model based on the standard instrument detection data and the preprocessed spectral image data of the skin sample; acquiring spectral image data of the skin to be detected; inputting the spectral image data of the skin to be detected into the skin texture detection model; and outputting multi-dimensional skin texture detection results.

[0006] Optionally, the spectral image data includes data from multiple spectral channels, which cover the visible light band and the near-infrared band.

[0007] Optionally, preprocessing the spectral image data of the skin sample includes performing one or more operations on the spectral image data of the skin sample, such as background subtraction, shadow removal, interpolation, or noise reduction.

[0008] Optionally, the skin samples include a preset number of skin samples of different genders, ages, and skin types; the multi-dimensional skin quality detection results include quantitative or qualitative assessment results of at least two skin quality indicators among skin color, pigmentation, erythema, sebum secretion level, hydration status, sensitive skin, skin texture, elasticity, or pore characteristics.

[0009] Optionally, constructing a skin texture detection model based on the standard instrument detection data and preprocessed skin sample spectral image data includes: determining an effective region; performing spectral inversion on the spectral image data of the skin sample in the effective region to obtain the spectral image of the effective region; obtaining spectral images of multiple preset bands based on the spectral image of the effective region; acquiring standard instrument detection data of the effective region to obtain multiple preset component data; and establishing the correlation between the spectral image of each preset band and the multiple preset components.

[0010] Optionally, establishing the association between the spectral image of each preset band and the multiple preset components includes: calculating the correlation coefficient between the spectral image data of each preset band and the data of the multiple preset components; determining the weight of each preset component in the spectral image data of each preset band based on the correlation coefficient; using the weight to perform weighted fusion of the spectral image data of each preset band to form fused feature data; and training a machine learning model based on the standard instrument detection data and the fused feature data to obtain the skin texture detection model.

[0011] Optionally, the machine learning model includes at least one of the following: support vector machine, decision tree, random forest, or neural network model.

[0012] Optionally, establishing the association between the spectral image of each preset band and the multiple preset components further includes: calculating the gender and age correlation coefficients between the gender and age information of the spectral image data of each preset band and the multiple preset component data; determining the gender and age weights of each preset component in the spectral image data of each preset band based on the gender and age correlation coefficients; weighting the spectral image data of each preset band using the gender and age weights to form the fused feature data; and training the machine learning model based on the standard instrument detection data and the fused feature data to obtain the skin texture detection model.

[0013] Optionally, the correlation coefficient is calculated based on the Pearson algorithm.

[0014] Optionally, the preset ingredients include at least one of melanin, hemoglobin, collagen, oil, or water.

[0015] Optionally, the skin texture detection method further includes: preprocessing the spectral image data of the skin to be detected; determining the target effective region; performing spectral inversion on the spectral image data of the skin to be detected in the target effective region to obtain the spectral image of the target effective region; obtaining spectral images of multiple preset bands of the target effective region based on the spectral image of the target effective region; and establishing the correlation between the spectral image of each preset band of the target effective region and the multiple preset components.

[0016] Optionally, establishing the association between the spectral image of each preset band of the target effective region and the multiple preset components includes: calculating the correlation coefficient between the spectral image data of each preset band of the target effective region and the data of the multiple preset components; determining the weight of each preset component in the spectral image data of each preset band according to the correlation coefficient; and using the weight to perform weighted fusion of the spectral image data of each preset band to form the fused feature data to be detected.

[0017] Optionally, inputting the skin spectral image data to be detected into the skin texture detection model and outputting multi-dimensional skin texture detection results includes: inputting the fused feature data to be detected into the skin texture detection model and outputting the multi-dimensional skin texture detection results.

[0018] A second aspect of the present invention provides a skin texture detection device. The skin texture detection device includes a processor configured to perform the skin texture detection method as described above.

[0019] Optionally, the skin texture detection device further includes: a microlens array, a filter unit array, and a photoelectric sensor array coupled to the processor, arranged along the optical path, wherein each filter unit includes a filter sub-unit array, the filter sub-unit array being configured to form multiple spectral channels; the photoelectric sensor is configured to convert incident light signals into electrical signals; the processor is configured to generate spectral image data of the skin sample or spectral image data of the skin to be detected based on the electrical signals, the spectral image data including data from the multiple spectral channels; the multiple spectral channels covering the visible light band and the near-infrared band.

[0020] A third aspect of the present invention provides a computer-readable storage medium. The computer-readable storage medium includes computer-executable instructions stored thereon, which, when executed by a processor, implement the skin texture detection method as described above.

[0021] The skin texture detection method of the present invention can accurately, efficiently and reliably perform in-depth and multi-dimensional detection of the skin texture of real skin, supports non-contact detection, supports overall or partial detection, and can improve user experience.

[0022] The skin texture detection device of the present invention covers the visible light band and the near-infrared band, covers the broadband band, and covers the characteristic bands of real skin and various preset components. It can simultaneously acquire spectral image data of multiple spectral channels in a single shot, and can realize deep, multi-dimensional, high-precision, non-contact skin texture detection of the whole or local area of ​​real skin. Moreover, it has fast detection speed, high efficiency, and small size, which helps to improve the user experience. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the following description of the embodiments will be provided as examples. The drawings described below are merely embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation of the present invention.

[0024] Figure 1 A schematic flowchart of a skin texture detection method according to some embodiments of the present invention is shown.

[0025] Figure 2 A schematic diagram of a skin texture detection device according to some embodiments of the present invention is shown.

[0026] Figure 3 A schematic diagram of a skin texture detection device according to some embodiments of the present invention is shown.

[0027] Figure 4 A flowchart illustrating step S13 according to some embodiments of the present invention is shown.

[0028] Figure 5 A flowchart illustrating sub-step S135 according to some embodiments of the present invention is shown.

[0029] Figure 6 A schematic diagram of step S15 according to some embodiments of the present invention is shown.

[0030] Figure 7 A schematic diagram of step S15 according to some embodiments of the present invention is shown. Detailed Implementation

[0031] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0032] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0033] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "coupling" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection, an electrical connection, or a connection that allows for communication; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0034] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0035] The following provides many different embodiments or examples for implementing various structures of the invention. To simplify the invention, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. In addition, examples of various specific processes and materials are provided in this invention, but those skilled in the art will recognize the application of other processes and / or the use of other materials.

[0036] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0037] This invention provides a skin texture detection method. The method includes: acquiring spectral image data of a skin sample and standard instrument detection data; preprocessing the spectral image data of the skin sample; constructing a skin texture detection model based on the standard instrument detection data and the preprocessed spectral image data of the skin sample; acquiring spectral image data of the skin to be detected; inputting the spectral image data of the skin to be detected into the skin texture detection model; and outputting multi-dimensional skin texture detection results. This invention's skin texture detection method can accurately, efficiently, and reliably perform in-depth, multi-dimensional detection of real skin texture, supports non-contact detection, and supports overall or partial detection, thus improving the user experience.

[0038] Figure 1 A schematic flowchart of a skin texture detection method 10 according to some embodiments of the present invention is shown. Figure 1 As shown, skin texture detection method 10 includes steps S11 to S15. Step S11: Acquire spectral image data of a skin sample and standard instrument detection data. Step S12: Preprocess the spectral image data of the skin sample. Step S13: Construct a skin texture detection model based on the standard instrument detection data and the preprocessed spectral image data of the skin sample. Step S14: Acquire spectral image data of the skin to be detected. Step S15: Input the spectral image data of the skin to be detected into the skin texture detection model and output multi-dimensional skin texture detection results.

[0039] In some embodiments, the skin texture detection method 10 may be executed by a processor. The processor may be located on an electronic device such as a skin texture detection device, a mobile phone, a tablet computer, a laptop computer, a wearable device, an in-vehicle device, an augmented reality (AR) / virtual reality (VR) device, an ultramobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), or a smart home device.

[0040] In some embodiments, the processor may include processing circuitry, a central processing unit (CPU), a microcontroller unit (MCU), a digital signal processor (DSP), a graphics processing unit (GPU), an accelerator, a neural processing unit (NPU), a tensor processing unit (TPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, gate devices, or transistor logic devices, or similar devices.

[0041] The present invention also provides a skin texture detection device. Figure 2 A schematic diagram of a skin texture detection device 20 according to some embodiments of the present invention is shown. Figure 2 As shown, the skin texture detection device 20 includes a processor 21. The processor 21 can execute the skin texture detection method 10.

[0042] Figure 3 A schematic diagram of a skin texture detection device 20 according to some embodiments of the present invention is shown. Figure 3 As shown, the skin texture detection device 20 also includes a microlens array 22, a filter unit array 23, and a photoelectric sensor array 24 coupled to the processor 21, arranged along the optical path. Although not shown in the figure, the upstream of the optical path of the microlens array 22 may include a lens or lens group. Although not shown in the figure, the skin texture detection device 20 may include a display screen for visualizing the skin texture detection results.

[0043] In some embodiments, the microlens array 22 may include a plurality of microlenses. The plurality of microlenses may be arranged in a one-dimensional array or a two-dimensional array. The microlenses may focus the incident light L onto the filter unit array 23. In some embodiments, the incident surface of the microlenses may include an antireflection coating to improve light throughput, reduce reflection, and suppress noise.

[0044] In some embodiments, the filter unit array 23 is located downstream of the optical path of the microlens array 22. The filter unit array 23 includes multiple filter units. The multiple filter units can be arranged in a one-dimensional or two-dimensional array. Each filter unit can include an array of filter sub-units. The filter sub-unit array can be a one-dimensional or two-dimensional array. For example, the filter sub-units can be arranged in an n*n two-dimensional array or an m*n two-dimensional array, where n and m are positive integers, and n≠m. For example, the filter sub-unit array can be arranged in 2*2, 3*3, 4*4, 2*3, 3*4 arrays, etc. Each filter sub-unit can have a specific transmittance curve. Each filter sub-unit can form a spectral channel to allow light of a preset wavelength band to pass through. The operating wavelength bands of the filter sub-units in each filter unit are not exactly the same. For example, the operating wavelength bands of the filter sub-units in each filter unit can be different or partially the same. The filter sub-unit array in each filter unit can form multiple spectral channels to allow light of multiple preset wavelength bands to pass through. In some embodiments, multiple spectral channels can cover the visible light band and the near-infrared band. Preset bands may include the visible light band and the near-infrared band. The skin texture detection method 10 and skin texture detection device 20 of the present invention operate in the visible light band and the near-infrared band, supporting broadband detection. Compared to some schemes that only use red light for skin texture detection, the present invention can cover the characteristic bands of real skin and more components, enabling the detection of more components and thus achieving multi-dimensional skin texture detection. This helps improve the reliability and dimensionality of the detection results and enhances the user experience.

[0045] In some embodiments, the filter subunit may include a filter film, filter, nanoarray, grating, or similar device to achieve wavelength selection. It should be noted that this invention does not limit the number, arrangement, transmittance, number of channels, device type, or other parameters of the filter subunits in each filter unit; in practical applications, these parameters can be configured according to requirements.

[0046] In some embodiments, the photoelectric sensor array 24 is located downstream of the optical path of the filter unit array 23. The photoelectric sensor array 24 includes multiple photoelectric sensors. The multiple photoelectric sensors can be arranged in a one-dimensional or two-dimensional array. The photoelectric sensors can convert incident light signals into electrical signals. In some embodiments, the photoelectric sensor can include one or more pixels. The multiple pixels can be arranged in a one-dimensional or two-dimensional array. The pixel can serve as the smallest photosensitive unit of the photoelectric sensor array, converting incident light signals into electrical signals. In some embodiments, the pixel can include photosensitive elements such as photodiodes or phototransistors, for example, charge-coupled devices (CCDs) or complementary metal-oxide-semiconductor (CMOS) devices.

[0047] In some embodiments, processor 21 is coupled to photoelectric sensor array 24. Processor 21 can generate spectral image data of a skin sample or spectral image data of the skin to be detected based on the electrical signals output by photoelectric sensor array 24. The spectral image data includes data from multiple spectral channels. These multiple spectral channels cover the visible light band and the near-infrared band. The spectral image includes multiple pixels. Each pixel includes three-dimensional data (x, y, λ). Here, x and y represent the spatial dimensions, indicating the horizontal and vertical positions of the pixel in the image, and represent the pixel coordinates. λ represents the spectral dimension, indicating the wavelength. Multispectral images can reflect the reflection, absorption, or emission characteristics of an object at different wavelengths. Multispectral images include data from multiple spectral channels.

[0048] It should be noted that the present invention does not limit the correspondence between microlenses, filter units, filter sub-units, photoelectric sensors, pixels, and image pixels. It can be one-to-many or many-to-one. In practical applications, it can be configured according to requirements.

[0049] In some embodiments, the skin texture detection device 20 may further include a memory coupled to the processor 21. The memory may store information such as spectral image data of the skin sample, spectral image data of the skin to be detected, standard instrument test data, and program instructions. The processor 21 may retrieve the spectral image data of the skin sample, the spectral image data of the skin to be detected, standard instrument test data, and program instructions from the memory.

[0050] In some embodiments, the memory may include random access memory (RAM) or non-volatile memory (NVM). Further, the memory may include at least one of phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), read-only memory (ROM), and electrically erasable programmable read-only memory (EEPROM). In some embodiments, the memory may include cloud storage.

[0051] In some embodiments, the microlens array 22, the filter unit array 23, and the photoelectric sensor array 24 can be integrated onto a single spectral chip. In some embodiments, the processor 21 can be disposed on a separate chip. In some embodiments, the processor 21, microlens array 22, filter unit array 23, and photoelectric sensor array 24 can be integrated onto a single spectral chip, resulting in high integration, small size, and low cost. In practical applications, the configuration can be tailored to specific requirements.

[0052] In some embodiments, the skin texture detection device 20 may include a display module for visualizing the skin texture detection results. The display module may include a display screen. The display screen may include at least one of an LCD (liquid-crystal display), an LED (light emitting diode), and an OLED (organic light emitting diode).

[0053] In some embodiments, the skin texture detection device 20 may include a spectral camera. A single shot from the spectral camera can acquire image data from multiple spectral channels, simultaneously obtaining information on various components. This results in high detection efficiency, facilitating multi-dimensional simultaneous analysis, and supports non-contact detection, ensuring cleanliness and hygiene, thus enhancing the user experience.

[0054] The skin texture detection device 20 of the present invention is highly efficient, small in size, and low in cost, making it easy to integrate into various electronic devices. For example, it can be integrated into electronic devices in medical aesthetic institutions, hospitals, beauty salons, and other similar locations. It can also be integrated into mobile phones, tablets, laptops, wearable devices, in-vehicle devices, augmented reality / virtual reality devices, super mobile personal computers, netbooks, personal digital assistants, smart home devices, and other electronic devices. It is understood that the skin texture detection device 20 of the present invention can exist independently without integration into other electronic devices. For example, the skin texture detection device 20 may include a skin texture analyzer, which consumers can use themselves, making it very convenient.

[0055] The following describes each step using processor 21 executing skin texture detection method 10 as an example. It is understood that examples of processors in other electronic devices executing skin texture detection method 10 are similar.

[0056] In step S11, the processor 21 can acquire spectral image data and standard instrument detection data of the skin sample. In some embodiments, the skin sample may include a preset number of skin samples of different genders, ages, and skin types. For example, different genders may include men and women. For example, different ages may include people in age ranges such as 18-55 years old, 18-60 years old, and 18-70 years old. For example, different skin types may include dry, oily, combination, sensitive skin, acne lesions, acne scars, and pigmentation. For example, the preset number may be no less than 40, 50, 60, 100, or 200. It should be noted that this is only an illustrative example, and the present invention is not limited thereto. In practical applications, it can be flexibly adjusted according to needs.

[0057] In some embodiments, the processor 21 can communicate with a standard instrument. The standard instrument can detect skin sample data and communicate this data to the processor 21. Thus, the processor 21 can acquire the detection data from the standard instrument. In some embodiments, the skin sample data detected by the standard instrument can be stored in the cloud, and the processor 21 can retrieve the standard instrument detection data from the cloud.

[0058] In some embodiments, the standard instrument can detect at least one preset component from the skin sample, namely melanin, hemoglobin, collagen, sebum, or moisture. In other words, the standard instrument detection data may include data on at least one preset component from the standard instrument, namely melanin, hemoglobin, collagen, sebum, or moisture. One or more of these preset components can be used to analyze multidimensional skin texture test results. Multidimensional skin texture test results may include quantitative or qualitative assessments of at least two skin texture indicators, including skin color, pigmentation, erythema, sebum secretion level, hydration status, sensitive skin, skin texture, elasticity, or pore characteristics.

[0059] In some embodiments, standard instrument detection data for skin samples may include standard instrument detection data for effective areas. For example, effective areas may include the forehead, corners of the eyes, cheeks, chin, nose, etc. An effective area may include one or more key points, each including coordinate information. For example, different effective areas may correspond to different body parts. For example, one effective area may correspond to different body parts. In some embodiments, standard instrument detection data for skin samples may include standard instrument detection data for different preset components within the same effective area. In some embodiments, standard instrument detection data for skin samples may include standard instrument detection data for the same preset component in different effective areas. For example, the standard instrument may include a Corneometer CM825 probe, which can detect the water content (skin moisture) of the effective area. For example, the standard instrument may include a Vapometer probe, which can detect the water loss rate (water loss rate) of the effective area. For example, the standard instrument may include a Sebumeter@SM815, which can detect sebum secretion (sebum secretion) in the effective area. For example, the standard instrument can detect dermal thickness, epidermal density, and the upper and lower dermis (collagen). For example, a standard instrument may include a Vplus probe, which can detect melanin and hemoglobin in the effective area, performing pigmentation analysis, acne identification analysis, and large pore identification analysis. It should be noted that these effective areas and standard instruments are merely illustrative examples, and the invention is not limited thereto. In practical applications, appropriate effective areas or instruments can be selected according to requirements. In some embodiments, a standard instrument can support one or more detection functions; by adjusting the probe, multiple corresponding parameters or component detections can be achieved.

[0060] In some embodiments, the skin texture detection device 20 can acquire spectral data of skin samples. In some embodiments, the spectral data of the skin sample may include spectral image data of the effective area. For example, the effective area may include the forehead, corner of the eye, cheek, chin, nose, etc. The spectral image data of the skin sample corresponds to the effective area of ​​the standard instrument detection data. In some embodiments, the processor 21 can label the spectral data and standard instrument detection data of the corresponding effective area of ​​the skin sample, associate the spectral data and standard instrument detection data of the same effective area, and associate the coordinate information, spectral data, and standard instrument detection data of the same key point. For example, the processor 21 can label the spectral data and standard instrument detection data of the forehead with label1 and label1'. For example, the processor 21 can label the spectral data and standard instrument detection data of the corner of the eye with label2 and label2'. For example, the processor 21 can label the spectral data and standard instrument detection data of the cheek with label3 and label3'. For example, the processor 21 can label the spectral data and standard instrument detection data of the chin with label4 and label4'. For example, processor 21 can label the spectral data and standard instrument detection data of the nose with labels 5 and 5'. The labels on the spectral data and standard instrument detection data of the effective area can include information such as gender and age. It is understood that the spectral data of the effective area can include information on multiple components of real skin. The standard instrument detection data of the effective area can also include information on multiple components of real skin.

[0061] In some embodiments, the spectral image data includes data from multiple spectral channels. These multiple spectral channels can be formed using an array of filter sub-units. For example, the filter sub-unit array may include 2*2, 2*3, 3*3, 3*4, and 4*4 arrays, forming 4, 6, 9, 12, and 16 spectral channels respectively. The skin texture detection device 20 can simultaneously acquire spectral image data from multiple spectral channels in a single shot, enabling simultaneous acquisition of information on various components of real skin, improving skin texture detection efficiency, achieving multi-dimensional synchronous analysis, and enhancing the user experience.

[0062] In some embodiments, multiple spectral channels cover the visible and near-infrared bands, supporting broadband detection. For example, multiple spectral channels can cover 350–1100 nm, covering the characteristic wavelengths of real skin and various preset components, enabling the detection of multiple preset components and achieving multi-dimensional skin texture detection while ensuring the reliability of the detection results. In some embodiments, preset components may include at least one of melanin, hemoglobin, collagen, oil, or water.

[0063] For example, melanin absorbs strongly in the 600-700nm wavelength range, which can be used to assess pigmentation, UV damage, and skin tone evenness. Other wavelengths that may detect melanin are 400nm, 450nm, 490nm, and 700nm. In other words, the characteristic wavelength ranges of melanin include 600-700nm, 400nm, 450nm, 490nm, and 700nm.

[0064] For example, the visible spectrum of hemoglobin exhibits characteristic peaks at 540 nm and 580 nm. In other words, the characteristic wavelength range of hemoglobin includes 540 nm and 580 nm.

[0065] For example, collagen exhibits predominantly scattering properties in the visible light region, and its correlation with dermal structure and tissue density occurs in the 400–420 nm and near-infrared 850–950 nm bands, which can be used to reflect skin elasticity and aging trends. In other words, the characteristic wavelengths of collagen include 400–420 nm and 850–950 nm.

[0066] For example, water exhibits a significant absorption characteristic around 970 nm, and within the upper limit of this system's wavelength range, it can effectively respond to the skin's hydration state. In other words, the characteristic wavelength range of water includes 970 nm.

[0067] For example, sebum absorbs light weakly in the visible light region, and its sebum secretion and skin texture differences are indirectly characterized by changes in reflection and absorption in the near-infrared 850–950 nm band. In other words, the characteristic wavelength range of sebum includes 850–950 nm.

[0068] It should be noted that these preset components and characteristic bands are merely illustrative examples, and the present invention is not limited thereto. These characteristic bands can have a certain degree of flexibility. It is understood that the characteristic bands of the preset components may vary slightly under different conditions or instrument measurements.

[0069] In some embodiments, the multidimensional skin texture detection results may include quantitative or qualitative assessments of at least two skin texture indicators selected from skin tone, pigmentation, erythema, sebum secretion level, hydration status, sensitive skin, skin texture, elasticity, or pore characteristics. It should be noted that the preset ingredients and skin texture indicators described herein are merely illustrative and are not intended to limit the scope of this invention.

[0070] Some traditional skin texture detection technologies are limited to red light band detection, making it difficult or impossible to detect multiple preset components reliably, and even more difficult or impossible to achieve efficient multi-dimensional detection results. Unlike traditional solutions, this invention features multiple spectral channels covering the visible and near-infrared bands, providing a broad spectrum and covering the characteristic bands of real skin and multiple preset components. This enables the detection of multiple preset components and achieves multi-dimensional skin texture detection. The skin texture detection device 20 can simultaneously acquire spectral image data from multiple spectral channels in a single shot, achieving efficient and simultaneous multi-dimensional detection. Its compact size also contributes to a better user experience.

[0071] In some embodiments, the skin texture detection device 20 can acquire spectral image data of skin samples under a preset light source environment. For example, the processor 21 can acquire spectral image data of skin samples under a white light source of 350~1100nm, which can achieve a uniform lighting environment, eliminate shadows on the face, reduce shadow noise, and improve the accuracy of skin texture detection results. It should be noted that this is only an illustrative example, and the present invention does not limit the wavelength of the white light source, but depends on the actual situation.

[0072] In step S12, the processor 21 can preprocess the spectral image data of the skin sample. In some embodiments, preprocessing the spectral image data of the skin sample includes one or more operations such as background subtraction, shadow removal, interpolation, or noise reduction. For example, the processor 21 can use an image segmentation algorithm to segment the spectral image to achieve background subtraction and shadow removal, which helps to remove noise and improve accuracy. In some embodiments, the image segmentation algorithm may include end-to-end algorithms such as UYOLO and U-NET, but the present invention is not limited thereto. For example, the processor 21 can perform interpolation processing through nearest neighbor interpolation, bilinear interpolation, bicubic interpolation, etc., but the present invention is not limited thereto. Through interpolation processing, each pixel of the spectral image can include data from multiple spectral channels. For example, the processor 21 can perform mean processing on the corresponding channel data of each pixel of the spectral image data of the skin sample to achieve noise reduction. Through preprocessing, noise can be removed, the amount of data can be reduced, and computational power can be saved to achieve efficient and accurate detection.

[0073] In step S13, the processor 21 can construct a skin texture detection model based on standard instrument detection data and preprocessed skin sample spectral image data. In some embodiments, the standard instrument detection data includes standard instrument detection data of multiple preset components in one or more effective regions of the skin sample. In some embodiments, the preprocessed skin sample spectral image data includes spectral image data of multiple preset components in one or more effective regions. Based on the standard instrument detection data of multiple preset components in one or more effective regions and the preprocessed skin sample spectral image data, the processor 21 can construct a skin texture detection model to achieve the detection of multiple preset components and realize multi-dimensional skin texture detection.

[0074] Figure 4 A flowchart illustrating step S13 according to some embodiments of the present invention is shown. Figure 4 As shown, step S13 includes sub-steps S131 to S135.

[0075] In sub-step S131, processor 21 can determine the valid area. In some embodiments, processor 21 can receive user input instructions and determine the valid area based on the user input instructions. For example, user input instructions may include at least one of voice instructions, action instructions, gesture instructions, image instructions, or text instructions.

[0076] In sub-step S132, processor 21 can perform spectral inversion on the spectral image data of the skin sample in the effective area to obtain the spectral image of the effective area. In some embodiments, processor 21 can acquire the spectral image data of the effective area based on a label. After acquiring the spectral image data of the effective area, processor 21 can perform spectral inversion on the spectral image data of the effective area to obtain the spectral image of the effective area. For example, processor 21 can acquire the spectral data of the forehead through label 1, perform spectral inversion on the spectral image data of the skin sample in the forehead to obtain the spectral image IMG1 of the forehead region. For example, processor 21 can acquire the spectral data of the corner of the eye through label 2, perform spectral inversion on the spectral image data of the skin sample in the corner of the eye to obtain the spectral image IMG2 of the corner of the eye region. For example, processor 21 can acquire the spectral data of the cheek through label 3, perform spectral inversion on the spectral image data of the skin sample in the cheek to obtain the spectral image IMG3 of the cheek region. For example, processor 21 can acquire the spectral data of the chin through label 4, perform spectral inversion on the spectral image data of the skin sample in the chin to obtain the spectral image IMG4 of the chin region. For example, processor 21 can obtain the spectral data of the nose through label 5, perform spectral inversion on the spectral image data of the skin sample of the nose, and obtain the spectral image IMG5 of the nose region. The spectral data of each effective region can include information on multiple components of the real skin.

[0077] In sub-step S133, processor 21 obtains spectral images of multiple preset bands based on the spectral image of the effective region. The spectral image of the effective region covers the visible light and near-infrared bands, covering 350~1100nm. Processor 21 can divide the spectral image of the effective region into bands to obtain spectral images of multiple preset bands. For example, the multiple preset bands may include a first preset band B1, a second preset band B2, a third preset band B3, a fourth preset band B4, and a fifth preset band B5. In some embodiments, the first preset band B1 can be 350~500nm. In some embodiments, the second preset band B2 can be 500~650nm. In some embodiments, the third preset band B3 can be 650~750nm. In some embodiments, the fourth preset band B4 can be 750~900nm. In some embodiments, the fifth preset band B5 can be 900~1100nm. Preset bands B1~B5 may correspond to one or more preset components such as melanin, hemoglobin, collagen, oil, and water. In some embodiments, the processor 21 can divide the spectral image IMG1 of the forehead region, the spectral image IMG2 of the corner of the eye region, the spectral image IMG3 of the cheek region, the spectral image IMG4 of the chin region, and the spectral image IMG5 of the nose region into preset bands B1 to B5 and output them. The spectral data of each preset band in each effective region can include information on multiple components of real skin. It should be noted that the band division strategy for the spectral images of different effective regions can be the same or different. In practical applications, a suitable division method can be selected according to the requirements.

[0078] In sub-step S134, processor 21 can acquire standard instrument detection data of the effective area to obtain multiple preset component data. In some embodiments, processor 21 can acquire standard instrument detection data of the effective area based on labels. For example, processor 21 can acquire standard instrument detection data of the forehead through label1'. For example, processor 21 can acquire standard instrument detection data of the corner of the eye through label2'. For example, processor 21 can acquire standard instrument detection data of the cheek through label3'. For example, processor 21 can acquire standard instrument detection data of the chin through label4'. For example, processor 21 can acquire standard instrument detection data of the nose through label5'. The standard instrument detection data of each effective area may include detection data of multiple components of real skin.

[0079] In sub-step S135, processor 21 can establish a correlation between the spectral image of each preset band and multiple preset components. For example, processor 21 can establish a correlation between the spectral image of the first preset band B1 and multiple preset components. For example, processor 21 can establish a correlation between the spectral image of the second preset band B2 and multiple preset components. For example, processor 21 can establish a correlation between the spectral image of the third preset band B3 and multiple preset components. For example, processor 21 can establish a correlation between the spectral image of the fourth preset band B4 and multiple preset components. For example, processor 21 can establish a correlation between the spectral image of the fifth preset band B5 and multiple preset components. In this way, processor 21 can establish a correlation between the spectral image of each preset band in the spectral image of each effective region and multiple preset components.

[0080] Figure 5 A flowchart illustrating sub-step S135 according to some embodiments of the present invention is shown. Figure 5 As shown, sub-step S135 includes operations OP1~OP4.

[0081] By operating OP1, processor 21 can calculate the correlation coefficient between the spectral image data of each preset band and the data of multiple preset components. Processor 21 can also calculate the correlation coefficient between the spectral image data of each preset band and multiple preset components in each effective region.

[0082] For each effective region's first preset band B1, processor 21 can calculate the correlation coefficient between the spectral image data of the first preset band B1 and data of one or more preset components, including melanin, hemoglobin, collagen, lipids, and water. For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the first preset band B1 and melanin data. B1,Mel For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the first preset band B1 and the hemoglobin data. B1,Hem For example, processor 21 can calculate the correlation coefficient r between spectral image data of the first preset band B1 and collagen data. B1,Col For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the first preset band B1 and the oil data. B1,Oil For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the first preset band B1 and the moisture data. B1,Moi .

[0083] For each effective region's second preset band B2, processor 21 can calculate the correlation coefficient between the spectral image data of the second preset band B2 and data of one or more preset components, including melanin, hemoglobin, collagen, lipids, and water. For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the second preset band B2 and melanin data. B2,Mel For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the second preset band B2 and the hemoglobin data. B2,Hem For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the second preset band B2 and the collagen data. B2,Col For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the second preset band B2 and the oil data. B2,Oil For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the second preset band B2 and the moisture data. B2,Moi .

[0084] For each effective region's third preset band B3, processor 21 can calculate the correlation coefficient between the spectral image data of the third preset band B3 and data of one or more preset components, including melanin, hemoglobin, collagen, lipids, and water. For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the third preset band B3 and melanin data. B3,Mel For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the third preset band B3 and the hemoglobin data. B3,Hem For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the third preset band B3 and the collagen data. B3,Col For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the third preset band B3 and the oil data. B3,Oil For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the third preset band B3 and the moisture data. B3,Moi .

[0085] For each effective region, the processor 21 can calculate the correlation coefficient between the spectral image data of the fourth preset band B4 and the data of one or more preset components, such as melanin, hemoglobin, collagen, lipids, and water. For example, the processor 21 can calculate the correlation coefficient r between the spectral image data of the fourth preset band B4 and the melanin data. B4,Mel For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fourth preset band B4 and the hemoglobin data. B4,HemFor example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fourth preset band B4 and the collagen data. B4,Col For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fourth preset band B4 and the oil data. B4,Oil For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fourth preset band B4 and the moisture data. B4,Moi .

[0086] For each effective region, the processor 21 can calculate the correlation coefficient between the spectral image data of the fifth preset band B5 and the data of one or more preset components, such as melanin, hemoglobin, collagen, lipids, and water. For example, the processor 21 can calculate the correlation coefficient r between the spectral image data of the fifth preset band B5 and the melanin data. B5,Mel For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fifth preset band B5 and the hemoglobin data. B5,Hem For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fifth preset band B5 and the collagen data. B5,Col For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fifth preset band B5 and the oil data. B5,Oil For example, processor 21 can calculate the correlation coefficient r between the spectral image data of the fifth preset band B5 and the moisture data. B5,Moi .

[0087] The foregoing embodiments exemplify an example of how the processor 21 calculates the correlation coefficients between the spectral image data of the first preset band B1 to the fifth preset band B5 of each effective region and the data of one or more preset components, including melanin, hemoglobin, collagen, oil, and water. It should be noted that these preset bands and preset components are merely illustrative examples, and the present invention is not limited thereto. In practical applications, they can be appropriately adjusted according to requirements.

[0088] In some embodiments, the processor 21 can calculate the correlation coefficients based on the Pearson algorithm. That is, the correlation coefficients can be Pearson correlation coefficients, with values ​​ranging from -1 to 1.

[0089] In some embodiments, the processor 21 can calculate the correlation coefficient between the spectral image data of the preset band and the preset component data based on the covariance between the spectral image data of the preset band and the preset component data, the standard deviation of the spectral image data of the preset band, and the standard deviation of the preset component data.

[0090] In some embodiments, the processor 21 can calculate the correlation coefficient between the spectral image data of the preset band and the preset component data based on (Equation 1).

[0091] (Equation 1).

[0092] in, This represents the correlation coefficient between the spectral image data of the j-th preset band in each effective region and the data of the preset component k. The preset component k can represent one of the following: melanin, hemoglobin, collagen, lipids, or water. This represents the covariance between the spectral image data of the j-th preset band and the preset component k data for each effective region. This represents the standard deviation of the spectral image data for the j-th preset band in each effective region. This represents the standard deviation of the preset component k data for each effective region.

[0093] It should be noted that the above embodiments use Pearson correlation coefficient as an example for illustrative purposes. This invention does not limit the calculation method of correlation coefficient. In practical applications, an appropriate calculation method can be selected according to the requirements.

[0094] By operating OP2, processor 21 can determine the weight of each preset component in the spectral image data of each preset band based on the correlation coefficient. Processor 21 can calculate the weight of each preset component in the spectral image data of each preset band in multiple preset bands of each effective region.

[0095] For the first preset band B1 of each effective region, for example, processor 21 can base it on the correlation coefficient r. B1,Mel Determine the weight w of melanin in the spectral image data of the first preset band B1. B1,Mel For example, processor 21 can be based on the correlation coefficient r. B1,Hem Determine the weight w of hemoglobin in the spectral image data of the first preset band B1. B1,Hem For example, processor 21 can be based on the correlation coefficient r. B1,Col Calculate the weight w of collagen in the spectral image data of the first preset band B1. B1,Col For example, processor 21 can be based on the correlation coefficient r. B1,Oil Calculate the weight w of oil in the spectral image data of the first preset band B1. B1,Oil For example, processor 21 can be based on the correlation coefficient r. B1,Moi Calculate the weight w of water content in the spectral image data of the first preset band B1. B1,Moi .

[0096] For the second preset band B2 of each effective region, for example, processor 21 can base it on the correlation coefficient r. B2,Mel Determine the weight w of melanin in the spectral image data of the second preset band B2. B2,Mel For example, processor 21 can be based on the correlation coefficient r. B2,Hem Determine the weight w of hemoglobin in the spectral image data of the second preset band B2. B2,Hem For example, processor 21 can be based on the correlation coefficient r. B2,Col Calculate the weight w of collagen in the spectral image data of the second preset band B2. B2,Col For example, processor 21 can be based on the correlation coefficient r. B2,Oil Calculate the weight w of oil in the spectral image data of the second preset band B2. B2,Oil For example, processor 21 can be based on the correlation coefficient r. B2,Moi Calculate the weight w of water content in the spectral image data of the second preset band B2. B2,Moi .

[0097] For the third preset band B3 of each effective region, for example, processor 21 can base it on the correlation coefficient r B3,Mel Determine the weight w of melanin in the spectral image data of the third preset band B3. B3,Mel For example, processor 21 can be based on the correlation coefficient r. B3,Hem Determine the weight w of hemoglobin in the spectral image data of the third preset band B3. B3,Hem For example, processor 21 can be based on the correlation coefficient r. B3,Col Calculate the weight w of collagen in the spectral image data of the third preset band B3. B3,Col For example, processor 21 can be based on the correlation coefficient r. B3,Oil Calculate the weight w of oil in the spectral image data of the third preset band B3. B3,Oil For example, processor 21 can be based on the correlation coefficient r. B3,Moi Calculate the weight w of water content in the spectral image data of the third preset band B3. B3,Moi .

[0098] For the fourth preset band B4 of each effective region, for example, processor 21 can base it on the correlation coefficient r. B4,Mel Determine the weight w of melanin in the spectral image data of the fourth preset band B4. B4,Mel For example, processor 21 can be based on the correlation coefficient r. B4,Hem Determine the weight w of hemoglobin in the spectral image data of the fourth preset band B4. B4,Hem For example, processor 21 can be based on the correlation coefficient r. B4,ColCalculate the weight w of collagen in the spectral image data of the fourth preset band B4. B4,Col For example, processor 21 can be based on the correlation coefficient r. B4,Oil Calculate the weight w of oil in the spectral image data of the fourth preset band B4. B4,Oil For example, processor 21 can be based on the correlation coefficient r. B4,Moi Calculate the weight w of water content in the spectral image data of the fourth preset band B4. B4,Moi .

[0099] For the fifth preset band B5 of each effective region, for example, processor 21 can base it on the correlation coefficient r. B5,Mel Determine the weight w of melanin in the spectral image data of the fifth preset band B5. B5,Mel For example, processor 21 can be based on the correlation coefficient r. B5,Hem Determine the weight w of hemoglobin in the spectral image data of the fifth preset band B5. B5,Hem For example, processor 21 can be based on the correlation coefficient r. B5,Col Calculate the weight w of collagen in the spectral image data of the fifth preset band B5. B5,Col For example, processor 21 can be based on the correlation coefficient r. B5,Oil Calculate the weight w of oil in the spectral image data of the fifth preset band B5. B5,Oil For example, processor 21 can be based on the correlation coefficient r. B5,Moi Calculate the weight w of water content in the spectral image data of the fifth preset band B5. B5,Moi .

[0100] In some embodiments, the weight of each preset component in the spectral image data of each preset band can be the same as the correlation coefficient. For example, the processor 21 can use the correlation coefficient as the weight. In some embodiments, the weight of each preset component in the spectral image data of each preset band can be different from the correlation coefficient. For example, the processor 21 can adjust the weight based on the correlation coefficient. In practical applications, this can be set according to requirements.

[0101] By operating OP3, processor 21 can use weights to perform weighted fusion of spectral image data for each preset band, forming fused feature data. For multiple preset bands in each effective region, processor 21 can use the weights of various preset components in the spectral image data of each preset band to perform weighted fusion of the spectral image data of each preset band, forming fused feature data for each preset band.

[0102] For the first preset band B1 of each effective region, processor 21 can utilize the weight w of melanin. B1,Mel Weight of hemoglobin wB1,Hem The weight of collagen w B1,Col The weight of oils w B1,Oil and the weight of water content w B1,Moi The spectral image data of the first preset band B1 are weighted and fused to form the fused feature data FF1 of the first preset band B1 for each effective region.

[0103] FF1=[w B1,Mel *B1 sep , w B1,Hem *B1 sep , w B1,Col *B1 sep , w B1,Oil *B1 sep , w B1,Moi *B1 sep ].

[0104] Among them, B1 sep This represents the spectral image data of the first preset band B1.

[0105] For the second preset band B2 of each effective region, processor 21 can utilize the weight w of melanin. B2,Mel Weight of hemoglobin w B2,Hem The weight of collagen w B2,Col The weight of oils w B2,Oil and the weight of water content w B2,Moi The spectral image data of the second preset band B2 are weighted and fused to form the fused feature data FF2 of the second preset band B2 for each effective region.

[0106] FF2=[w B2,Mel *B2 sep , w B2,Hem *B2 sep , w B2,Col *B2 sep , w B2,Oil *B2 sep , w B2,Moi *B2 sep ].

[0107] Among them, B2 sep This represents the spectral image data of the second preset band B2.

[0108] For the third preset band B3 of each effective region, processor 21 can utilize the weighted w of melanin. B3,Mel Weight of hemoglobin w B3,Hem The weight of collagen w B3,Col The weight of oils w B3,Oil and the weight of water content w B3,MoiThe spectral image data of the third preset band B3 is weighted and fused to form the fused feature data FF3 of the third preset band B3 for each effective region.

[0109] FF3=[w B3,Mel *B3 sep , w B3,Hem *B3 sep , w B3,Col *B3 sep , w B3,Oil *B3 sep , w B3,Moi *B3 sep ].

[0110] Among them, B3 sep This represents the spectral image data of the third preset band B3.

[0111] For the fourth preset band B4 in each effective region, processor 21 can utilize the weighted w of melanin. B4,Mel Weight of hemoglobin w B4,Hem The weight of collagen w B4,Col The weight of oils w B4,Oil and the weight of water content w B4,Moi The spectral image data of the fourth preset band B4 are weighted and fused to form the fused feature data FF4 of the fourth preset band B4 for each effective region.

[0112] FF4=[w B4,Mel *B4 sep , w B4,Hem *B4 sep , w B4,Col *B4 sep , w B4,Oil *B4 sep , w B4,Moi *B4 sep ].

[0113] Among them, B4 sep This represents the spectral image data of the fourth preset band B4.

[0114] For the fifth preset band B5 in each effective region, processor 21 can utilize the weighted w of melanin. B5,Mel Weight of hemoglobin w B5,Hem The weight of collagen w B5,Col The weight of oils w B5,Oil and the weight of water content w B5,Moi The spectral image data of the fifth preset band B5 is weighted and fused to form the fused feature data FF5 of the fifth preset band B5 for each effective region.

[0115] FF5=[w B5,Mel *B5 sep , w B5,Hem *B5 sep , w B5,Col *B5 sep , w B5,Oil *B5 sep , w B5,Moi *B5 sep ].

[0116] Among them, B5 sep This represents the spectral image data of the fifth preset band B5.

[0117] By operating OP4, processor 21 can train a machine learning model based on standard instrument detection data and fused feature data to obtain a skin texture detection model. The standard instrument detection data may include data on at least one preset component detected by the standard instrument, such as melanin, hemoglobin, collagen, oil, or moisture. Processor 21 can input the standard instrument detection data for each effective region and the fused feature data FF1, FF2, FF3, FF4, and FF5 of each preset band in each effective region into the machine learning model for training to obtain the skin texture detection model. For example, processor 21 can input the standard instrument detection data for the forehead, corners of the eyes, cheeks, chin, and nose, as well as the fused feature data FF1, FF2, FF3, FF4, and FF5 of the first preset band B1 to the fifth preset band B5, into the machine learning model for training to obtain the skin texture detection model. It is understood that the skin texture detection model can be iteratively optimized multiple times during training. In some embodiments, the machine learning model may include at least one of the following: support vector machine, decision tree, random forest, or neural network model.

[0118] The following example uses a neural network model to illustrate how processor 21 trains a machine learning model based on standard instrument detection data and fused feature data to obtain a skin texture detection model.

[0119] Step 1: Data Preparation and Dataset Construction. The dataset contains multiple samples. Each sample corresponds to a face image, including the fused feature data (FF1~FF5) of the five extracted effective regions (forehead, corner of the eye, cheek, chin, and nose) in five preset bands (B1~B5), as well as the standard instrument detection data for each region (i.e., the true values ​​of the five preset components: melanin, hemoglobin, collagen, oil, and water in each region).

[0120] The second step is data standardization. Processor 21 can perform normalization processing on each feature dimension.

[0121] The third step is to construct a neural network model. Processor 21 can construct a standard multilayer perceptron model. The specific structure of a multilayer perceptron model may include an input and normalization layer, a feature abstraction hidden layer, and a multi-task output layer.

[0122] For example, the input and normalization layers can include an input layer and a batch normalization layer. The input layer can receive a 25-dimensional feature vector, corresponding to all fused features of 5 valid regions and 5 preset bands. The batch normalization layer can normalize this 25-dimensional input feature vector to accelerate subsequent training and improve model stability.

[0123] For example, the feature abstraction hidden layer can contain one or more sequentially connected fully connected hidden layers to progressively learn non-linear relationships in the features. For instance, the feature abstraction hidden layer includes a first hidden layer that maps the 25-dimensional input to a higher dimension (e.g., 64 or 128 dimensions). The output of the first hidden layer is connected to an activation function layer (e.g., ReLU activation function and Dropout operation (with a dropout rate of approximately 20%) to prevent overfitting). Optionally, the feature abstraction hidden layer may also include a second hidden layer that maps the output of the first hidden layer to a similar or slightly lower dimension (e.g., 32 or 64 dimensions). The output of the second hidden layer is connected to an activation function layer (e.g., ReLU activation function and Dropout operation). Optionally, the feature abstraction hidden layer may also include more hidden layers. These hidden layers can act as feature extractors, effectively extracting skin texture-related abstract patterns from the original fused features without complex attention or branching mechanisms, thus extracting skin texture features.

[0124] For example, the structure of a multi-task output layer can include five independent neurons. Each neuron in the multi-task output layer can be connected to all neurons in the last hidden layer of the feature abstraction hidden layer. The outputs of these five independent neurons are connected to an activation function layer, such as the sigmoid function, which restricts the output of each neuron to the (0,1) interval, making it easier to directly correspond to the normalized predicted values ​​of each preset component (melanin, hemoglobin, collagen, oil, water).

[0125] Step 4: Configure training parameters and loss function. The processor 21 can configure optimization objectives and hyperparameters for model training. The skin texture detection model of this invention can simultaneously predict five preset component data, essentially belonging to a multi-task regression problem. In some embodiments, the processor 21 can define a composite loss function, such as a weighted mean squared error loss, which is the sum of the prediction losses for the five preset components. The loss weight of each preset component can be adjusted according to its measurement accuracy or importance. Training hyperparameters may include learning rate, batch size, number of iterations, etc. Based on the training hyperparameters, the processor 21 can employ learning rate decay or early stopping strategies to optimize the training process and prevent overfitting. In practical applications, it can be flexibly configured according to requirements.

[0126] Step 5: Perform model training and iterative optimization. Processor 21 can divide the prepared dataset into training, validation, and test sets. Then, processor 21 can use the training set data to iteratively update the parameters of the neural network model using the backpropagation algorithm. After each iteration or training cycle, processor 21 can use the validation set to evaluate the model's performance and adjust the hyperparameters based on the validation loss, or decide whether to terminate training early. This iterative process continues until the model's performance on the validation set reaches a preset standard or converges.

[0127] Step 6: Model Validation and Deployment. After training, the processor 21 can use an independent test set to comprehensively evaluate the final skin texture detection model. Evaluation metrics may include, but are not limited to, root mean square error, mean absolute error, and coefficient of determination, used to quantify the model's prediction accuracy for each skin texture component. After validation and meeting application requirements, the model is saved and deployed to the skin texture detection system. In practical applications, the processor 21 only needs to input the spectral image data of the skin to be detected or the fusion feature data extracted based on the spectral image of the skin to be detected into the skin texture detection model. Through this model, the estimated values ​​of the five preset components corresponding to each region can be quickly inferred, thus completing comprehensive skin texture analysis without relying on standard instruments.

[0128] In some embodiments, sub-step S135, establishing the association between the spectral image of each preset band and multiple preset components, further includes: calculating the gender correlation coefficient and age correlation coefficient between the gender information and age information of the spectral image data of each preset band and the data of multiple preset components. The processor 21 can calculate the correlation coefficient between the gender information and age information of the spectral image data of each preset band and multiple preset components in multiple preset bands of each effective region.

[0129] For example, for the first preset band B1 of each effective region, during operation OP1, processor 21 can also calculate the sex correlation coefficient r between the sex information of the spectral image data of the first preset band B1 and the sex correlation coefficients r between melanin data, hemoglobin data, collagen data, oil data, and moisture data. B1Sex,Mel r B1Sex,Hem r B1Sex,Col r B1Sex,Oil r B1Sex,Moi and the age correlation coefficient r B1Age,Mel r B1Age,Hem r B1Age,Col r B1Age,Oil r B1Age,Moi .

[0130] Similarly, for the second preset band B2 of each effective region, during operation OP1, the processor 21 can also calculate the sex correlation coefficient r between the sex information of the spectral image data of the second preset band B2 and the sex correlation coefficients r between the melanin data, hemoglobin data, collagen data, oil data, and moisture data. B2Sex,Mel r B2Sex,Hem r B2Sex,Col r B2Sex,Oil r B2Sex,Moi and the age correlation coefficient r B2Age,Mel r B2Age,Hem r B2Age,Col r B2Age,Oil r B2Age,Moi .

[0131] Similarly, for the third preset band B3 of each effective region, during operation OP1, processor 21 can also calculate the sex correlation coefficient r between the sex information of the spectral image data of the third preset band B3 and the sex correlation coefficients r between melanin data, hemoglobin data, collagen data, oil data, and moisture data. B3Sex,Mel r B3Sex,Hem r B3Sex,Col r B3Sex,Oil r B3Sex,Moi and the age correlation coefficient r B3Age,Mel r B3Age,Hem r B3Age,Col r B3Age,Oil r B3Age,Moi .

[0132] Similarly, for the fourth preset band B4 of each effective region, during operation OP1, processor 21 can also calculate the sex correlation coefficient r between the sex information of the spectral image data of the fourth preset band B4 and the sex correlation coefficients r between melanin data, hemoglobin data, collagen data, oil data, and moisture data. B4Sex,Mel r B4Sex,Hem r B4Sex,Col r B4Sex,Oil rB4Sex,Moi and the age correlation coefficient r B4Age,Mel r B4Age,Hem r B4Age,Col r B4Age,Oil r B4Age,Moi .

[0133] Similarly, for the fifth preset band B5 of each effective region, during operation OP1, processor 21 can also calculate the sex correlation coefficient r between the sex information of the spectral image data of the fifth preset band B5 and the sex correlation coefficients r between melanin data, hemoglobin data, collagen data, oil data, and moisture data. B5Sex,Mel r B5Sex,Hem r B5Sex,Col r B5Sex,Oil r B5Sex,Moi and the age correlation coefficient r B5Age,Mel r B5Age,Hem r B5Age,Col r B5Age,Oil r B5Age,Moi .

[0134] In some embodiments, sub-step S135, establishing the association between the spectral image of each preset band and multiple preset components, further includes: determining the gender weight and age weight of each preset component of the spectral image data of each preset band based on the gender correlation coefficient and the age correlation coefficient. For example, for the first preset band B1 of each effective region, in operation OP2, the processor 21 can also determine the gender weight and age weight of each preset component based on the gender correlation coefficient r. B1Sex,Mel r B1Sex,Hem r B1Sex,Col r B1Sex,Oil r B1Sex,Moi The gender weights w for melanin, hemoglobin, collagen, lipids, and water in the spectral image data of the first preset band B1 are determined respectively. B1Sex,Mel w B1Sex,Hem w B1Sex,Col w B1Sex,Oil w B1Sex,Moi Processor 21 can adjust the age-related coefficient r. B1Age,Mel r B1Age,Hem r B1Age,Col r B1Age,Oil r B1Age,Moi The age weights w for melanin, hemoglobin, collagen, lipids, and water in the spectral image data of the first preset band B1 are determined respectively. B1Age,Mel w B1Age,Hem w B1Age,Col w B1Age,Oil w B1Age,Moi .

[0135] Similarly, for the second preset band B2 of each effective region, during operation OP2, processor 21 can also adjust the frequency band based on the gender correlation coefficient r. B2Sex,Mel r B2Sex,Hem r B2Sex,Col r B2Sex,Oil r B2Sex,Moi The gender weights w for melanin, hemoglobin, collagen, lipids, and water in the spectral image data of the second preset band B2 are determined respectively. B2Sex,Mel w B2Sex,Hem w B2Sex,Col w B2Sex,Oil w B2Sex,Moi Processor 21 can adjust the age-related coefficient r. B2Age,Mel r B2Age,Hem r B2Age,Col r B2Age,Oil r B2Age,Moi The age weights w for melanin, hemoglobin, collagen, lipids, and water in the spectral image data of the second preset band B2 are determined respectively. B2Age,Mel w B2Age,Hem w B2Age,Col w B2Age,Oil w B2Age,Moi .

[0136] Similarly, for the third preset band B3 of each effective region, during operation OP2, processor 21 can also adjust the frequency band based on the gender correlation coefficient r. B3Sex,Mel r B3Sex,Hem r B3Sex,Col r B3Sex,Oil r B3Sex,Moi The gender weights w for melanin, hemoglobin, collagen, lipids, and water in the spectral image data of the third preset band B3 were determined respectively. B3Sex,Mel w B3Sex,Hem w B3Sex,Col w B3Sex,Oil w B3Sex,Moi Processor 21 can adjust the age-related coefficient r. B3Age,Mel r B3Age,Hem r B3Age,Col r B3Age,Oil r B3Age,Moi The age weights w for melanin, hemoglobin, collagen, lipids, and water in the spectral image data of the third preset band B3 were determined respectively. B3Age,Mel w B3Age,Hem w B3Age,Col w B3Age,Oil w B3Age,Moi .

[0137] The examples for the fourth preset band B4 and the fifth preset band B5 in each effective zone are similar and will not be repeated here. It should be noted that the gender weight can be the same as or different from the gender correlation coefficient. The age weight can also be the same as or different from the age correlation coefficient. In practical applications, these settings can be configured according to requirements.

[0138] In some embodiments, sub-step S135, establishing the association between the spectral image of each preset band and multiple preset components, further includes: weighting the spectral image data of each preset band using gender weights and age weights to form fused feature data. For multiple preset bands in each effective region, the processor 21 can use the weights of various preset components, gender weights, and age weights in the spectral image data of each preset band to perform weighted fusion of the spectral image data of each preset band to form fused feature data for each preset band.

[0139] For example, for the first preset band B1 of each effective region, during operation OP3, processor 21 can also utilize the weights of melanin, hemoglobin, collagen, oil, water, and gender weight w. B1Sex,Mel w B1Sex,Hem w B1Sex,Col w B1Sex,Oil w B1Sex,Moi and age weight w B1Age,Mel w B1Age,Hem w B1Age,Col w B1Age,Oil w B1Age,Moi The spectral image data of the first preset band B1 are weighted and fused to form the fused feature data FF1' of the first preset band B1 for each effective region.

[0140] Example,

[0141] FF1'

[0142] =[w B1Sex,Mel *w B1Age,Mel *w B1,Mel *B1 sep ,w B1Sex,Hem *w B1Age,Hem *w B1,Hem *B1 sep , w B1Sex,Col *w B1Age,Col *w B1,Col *B1 sep ,w B1Sex,Oil *w B1Age,Oil *w B1,Oil *B1 sep , w B1Sex,Moi *w B1Age,Moi *w B1,Moi *B1sep ].

[0143] Among them, B1 sep This represents the spectral image data of the first preset band B1.

[0144] Similarly, for the second preset band B2 of each effective region, during operation OP3, processor 21 can also utilize the weights of melanin, hemoglobin, collagen, lipids, water, and gender weights w. B2Sex,Mel w B2Sex,Hem w B2Sex,Col w B2Sex,Oil w B2Sex,Moi and age weight w B2Age,Mel w B2Age,Hem w B2Age,Col w B2Age,Oil w B2Age,Moi The spectral image data of the second preset band B2 are weighted and fused to form the fused feature data FF2' of the second preset band B2 for each effective region.

[0145] Similarly, for the third preset band B3 of each effective region, the processor 21 can also use the weights of melanin, hemoglobin, collagen, oil, water, gender, and age to perform weighted fusion of the spectral image data of the third preset band B3 to form the fused feature data FF3' of the third preset band B3 of each effective region.

[0146] Similarly, for the fourth preset band B4 of each effective region, the processor 21 can also use the weights of melanin, hemoglobin, collagen, oil, water, gender, and age to perform weighted fusion of the spectral image data of the fourth preset band B4 to form the fused feature data FF4' of the fourth preset band B4 of each effective region.

[0147] Similarly, for the fifth preset band B5 of each effective region, the processor 21 can also use the weights of melanin, hemoglobin, collagen, oil, water, gender, and age to perform weighted fusion of the spectral image data of the fifth preset band B5 to form the fused feature data FF5' of the fifth preset band B5 of each effective region.

[0148] It is understood that the present invention is not limited to the examples provided here, which are the first preset band B1 to the fifth preset band B5.

[0149] In some embodiments, sub-step S135, establishing the association between the spectral image of each preset band and multiple preset components, further includes: training a machine learning model based on standard instrument detection data and fused feature data to obtain a skin texture detection model. For example, in operation OP4, processor 21 can also input standard instrument detection data for each effective region and fused feature data FF1', FF2', FF3', FF4', and FF5' of each preset band in each effective region into the machine learning model for training to obtain a skin texture detection model. In some embodiments, the machine learning model may include at least one of the following: support vector machine, decision tree, random forest, or neural network model.

[0150] In step S14, the processor 21 can acquire spectral image data of the skin to be detected. In some embodiments, the skin texture detection device 20 can acquire spectral image data of the skin to be detected. Step S14 is the same as or similar to the operation of acquiring spectral image data of the skin sample in step S11, and will not be described again here.

[0151] In step S15, the processor 21 can input the spectral image data of the skin to be detected into the skin texture detection model and output multi-dimensional skin texture detection results. Figure 6 A schematic diagram of step S15 according to some embodiments of the present invention is shown. Figure 6 As shown, the processor 21 can input the skin spectral image data IMGtest to be detected into the skin texture detection model M, and use the input skin texture detection model M to process the skin spectral image data IMGtest to be detected, and output the detection results of various preset components, such as the detection result R of melanin data. Mel Detection results of hemoglobin data R Hem Collagen data detection results R Col Detection results of oil and fat data R Oil , Moisture data detection results R Moi The detection results for each preset component may include content, density, etc. In some embodiments, the processor 21 may base its detection results R on melanin data. Mel Detection results of hemoglobin data R Hem Collagen data detection results R Col Detection results of oil and fat data R Oil , Moisture data detection results R MoiThis involves performing quantitative or qualitative analysis of multiple skin quality indicators to obtain multi-dimensional skin quality detection results. For example, multi-dimensional skin quality detection results may include quantitative or qualitative assessments of at least two skin quality indicators, such as skin tone, pigmentation, erythema, sebum secretion level, hydration status, sensitive skin, skin texture, elasticity, or pore characteristics. In this way, skin quality detection results for multiple preset components of the skin being tested, as well as multi-dimensional skin quality detection results, can be obtained.

[0152] In some embodiments, the skin texture detection method 10 further includes preprocessing the spectral image data of the skin to be detected. This step can be performed before or after the spectral image data of the skin to be detected is input into the skin texture detection model. This step is the same as or similar to the operation of processor 21 preprocessing the spectral image data of the skin sample in step S12, and will not be described again here.

[0153] In some embodiments, the skin texture detection method 10 further includes: determining a target effective region. This step can be performed before or after the spectral image data of the skin to be detected is input into the skin texture detection model. The processor 21 can first determine the target effective region, and then input the spectral image data of the target effective region of the spectral image data of the skin to be detected into the skin texture detection model. Alternatively, the processor 21 can first input the spectral image data of the skin to be detected into the skin texture detection model, and then determine the target effective region. This step is the same as or similar to sub-step S131. The target effective region is the target detection area of ​​the skin to be detected.

[0154] In some embodiments, the skin texture detection method 10 further includes: performing spectral inversion on the spectral image data of the target effective area of ​​the skin to be detected to obtain a spectral image of the target effective area. This step is the same as or similar to sub-step S132.

[0155] In some embodiments, the skin texture detection method 10 further includes: obtaining spectral images of multiple preset bands of the target effective region based on the spectral image of the target effective region. This step is the same as or similar to sub-step S133.

[0156] In some embodiments, the skin texture detection method 10 further includes: establishing a correlation between the spectral image of each preset band of the target effective region and multiple preset components. This step is the same as or similar to sub-step S135.

[0157] In some embodiments, establishing the correlation between the spectral image of each preset band of the target effective region and multiple preset components includes calculating the correlation coefficient between the spectral image data of each preset band of the target effective region and the data of multiple preset components. This operation is the same as or similar to operation OP1.

[0158] In some embodiments, establishing the association between the spectral image of each preset band of the target effective region and multiple preset components includes: determining the weight of each preset component in the spectral image data of each preset band based on the correlation coefficient. This operation is the same as or similar to operation OP2.

[0159] In some embodiments, establishing the association between the spectral image of each preset band of the target effective region and multiple preset components includes: weighted fusion of the spectral image data of each preset band using weights to form the fused feature data to be detected for each preset band of the target effective region, namely FF1test, FF2test, FF3test, FF4test, and FF5test. This operation is the same as or similar to operation OP3.

[0160] Figure 7 A schematic diagram of step S15 according to some embodiments of the present invention is shown. Figure 7 As shown, the processor 21 can input the fused feature data to be detected in each preset band of the target effective region, FF1test, FF2test, FF3test, FF4test, and FF5test, into the skin texture detection model M. The skin texture detection model M then uses this data to detect the fused feature data and outputs detection results for various preset components, such as the detection result R for melanin data. Mel Detection results of hemoglobin data R Hem Collagen data detection results R Col Detection results of oil and fat data R Oil , Moisture data detection results R Moi The test results for each preset component can include content, density, etc.

[0161] In some embodiments, the processor 21 can base its detection results R on the melanin data. Mel Detection results of hemoglobin data R Hem Collagen data detection results R Col Detection results of oil and fat data R Oil , Moisture data detection results R Moi This involves performing quantitative or qualitative analysis on multiple skin quality indicators to obtain multidimensional skin quality assessment results. For example, multidimensional skin quality assessment results may include quantitative or qualitative assessments of at least two skin quality indicators among skin color, pigmentation, erythema, sebum secretion level, hydration status, sensitive skin, skin texture, elasticity, or pore characteristics.

[0162] In this way, the processor 21 can obtain skin texture detection results and multi-dimensional skin texture detection results of multiple preset components on the skin to be detected.

[0163] In some embodiments, processor 21 can perform quantitative or qualitative assessments of multiple skin quality indicators based on a preset ingredient. For example, processor 21 can perform assessments based on the detection results R of oil data. Oil The processor 21 performs quantitative or qualitative analysis on sebum secretion levels, hydration status, and skin texture. For example, if the sebum data detection result is greater than the sebum threshold, the processor 21 can determine that the skin is oily, poorly hydrated, has a rough skin texture, large pores, and is prone to blackheads and acne. If the sebum data detection result is not greater than the sebum threshold, the processor 21 can determine that the skin is not oily, well-hydrated, has a fine and smooth skin texture, and inconspicuous pores. If the sebum data detection result is in the first interval, the processor 21 can determine that the sebum secretion level is at level one, the hydration status is at level one, and the skin texture is at level one. If the sebum data detection result is in the second interval, the processor 21 can determine that the sebum secretion level is at level two, the hydration status is at level two, and the skin texture is at level two. If the sebum data detection result is in the third interval, the processor 21 can determine that the sebum secretion level is at level three, the hydration status is at level three, and the skin texture is at level three. For example, the first interval is smaller than the second interval, and the second interval is smaller than the third interval. For example, the first state is not moist, the second state is moderately moist, and the third state is moist. For example, the first level is rough with visible pores. The second level is slightly rough with slightly visible pores. The third level is not rough (fine) with invisible pores.

[0164] In some embodiments, processor 21 can perform quantitative or qualitative assessments of various skin quality indicators based on a variety of preset ingredients. For example, processor 21 can perform assessments based on the detection results R of melanin data. Mel Detection results of hemoglobin data R Hem The processor 21 performs quantitative or qualitative analysis on skin color, pigmentation, and erythema. For example, if the melanin data detection result is greater than the melanin threshold, and the hemoglobin data detection result is greater than the hemoglobin threshold, then the processor 21 can determine that the skin color is dark, the pigmentation is severe, and the erythema is severe. If the melanin data detection result is not greater than the melanin threshold, and the hemoglobin data detection result is not greater than the hemoglobin threshold, then the processor 21 can determine that the skin color is light, the pigmentation is not severe, and the erythema is not severe or there is no erythema.

[0165] For example, processor 21 can base its detection results R on collagen data. Col , Moisture data detection results R MoiThe processor 21 performs quantitative or qualitative analysis on skin elasticity and hydration status. For example, if the collagen data detection result is greater than the collagen threshold and the moisture data detection result is greater than the moisture threshold, the processor 21 can determine that the skin has good elasticity and good hydration status (e.g., the hydration status is in the second state (moderately moist) or the third state (moist)). Conversely, if the collagen data detection result is not greater than the collagen threshold and the moisture data detection result is not greater than the moisture threshold, the processor 21 can determine that the skin has poor elasticity and poor hydration status (e.g., the hydration status is in the first state (not moist)).

[0166] For example, processor 21 can base its detection results R on oil data. Oil , Moisture data detection results R Moi The processor 21 performs quantitative or qualitative analysis on sebum secretion levels, hydration status, sensitive skin, and skin texture. For example, if the sebum data exceeds the sebum threshold but the moisture data does not exceed the moisture threshold, the processor 21 can determine that the skin is oily, poorly hydrated, sensitive, has a rough texture, large pores, and is prone to blackheads and acne. Conversely, if the sebum data does not exceed the sebum threshold but the moisture data exceeds the moisture threshold, the processor 21 can determine that the skin is not oily, well-hydrated, not sensitive, has a fine and smooth texture, inconspicuous pores, and is not prone to blackheads and acne.

[0167] It should be noted that the above embodiments are merely illustrative examples, and the present invention is not limited thereto. The processor 21 can perform quantitative or qualitative evaluations of one or more skin quality indicators based on one or more preset ingredients. In practical applications, the preset ingredients, skin quality indicators, and corresponding ranges or thresholds can be arbitrarily combined and set as needed.

[0168] In some embodiments, the skin texture detection device 20 can communicate with the user's mobile terminal. The skin texture detection device 20 can synchronize the detection results of multiple preset ingredients and multi-dimensional skin texture detection results to the user's mobile terminal through the processor 21, so that the user can easily and promptly grasp the detection results.

[0169] In some embodiments, the processor 21 can output skin conditioning suggestions for users based on the detection results of multiple preset ingredients and multi-dimensional skin texture detection results.

[0170] In some embodiments, the processor 21 can visualize the detection results of various preset ingredients, multi-dimensional skin texture detection results, and conditioning suggestions through the display module and output them to the user for reference, so that the user can intuitively grasp the skin detection information and improve the user experience.

[0171] The skin texture detection method of the present invention can accurately, efficiently and reliably perform in-depth and multi-dimensional detection of the skin texture of real skin, supports non-contact detection, supports overall or partial detection, and can improve user experience.

[0172] The skin texture detection device of the present invention covers the visible light band and the near-infrared band, covers the broadband band, and covers the characteristic bands of real skin and various preset components. It can simultaneously acquire spectral image data of multiple spectral channels in a single shot, and can realize deep, multi-dimensional, high-precision, non-contact skin texture detection of the whole or local area of ​​real skin. Moreover, it has fast detection speed, high efficiency, and small size, which helps to improve the user experience.

[0173] The present invention also provides a computer-readable storage medium. The computer-readable storage medium includes computer-executable instructions stored thereon, which, when executed by a processor, implement the skin texture detection method 10 as described above.

[0174] In some embodiments, the present invention may take the form of a computer program product implemented on one or more storage media containing program code. Computer-usable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: PRAM, SRAM, DRAM, other types of RAM, ROM, EEPROM, flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital video disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0175] It should be noted that this specification provides method operation steps as shown in the embodiments or diagrams, but based on conventional or non-inventive labor, more or fewer operation steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many steps and does not represent the only execution order. In actual system or device products, the methods shown in the embodiments or flowcharts can be executed sequentially or in parallel.

[0176] It should be noted that although several modules of the skin texture detection device have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more modules described above can be implemented in one module. Conversely, the features and functions of one module described above can be further divided and embodied by multiple modules. Furthermore, the various modules mentioned in this invention can be implemented in hardware, software, or a combination of both.

[0177] It should be noted that the present invention may include Figure 1-7 Any one or more features of any one or more embodiments. In other words, not all features shown in the figures need to be implemented simultaneously in the skin texture detection method / skin texture detection device / computer-readable storage medium of the present invention. Features of the skin texture detection method / skin texture detection device / computer-readable storage medium of the present invention can be used interchangeably.

[0178] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A skin texture detection method, characterized in that, include: Acquire spectral image data and standard instrument test data from skin samples; The spectral image data of the skin sample is preprocessed; A skin texture detection model is constructed based on the standard instrument detection data and the preprocessed skin sample spectral image data. Acquire spectral image data of the skin to be tested; The skin spectral image data to be detected is input into the skin texture detection model, which outputs multi-dimensional skin texture detection results. The construction of the skin texture detection model includes: Determine the valid area; Spectral inversion is performed on the spectral image data of the skin sample in the effective region to obtain the spectral image of the effective region; Based on the spectral image of the effective region, spectral images of multiple preset bands are obtained; Obtain standard instrument detection data of the effective region to obtain data on various preset components; Establishing the correlation between the spectral image of each preset band and the multiple preset components includes: Calculate the correlation coefficient between the spectral image data of each preset band and the data of the multiple preset components; Based on the correlation coefficient, the weight of each preset component in the spectral image data of each preset band is determined; The spectral image data of each preset band are weighted and fused using the weights to form fused feature data; The skin texture detection model is obtained by training a machine learning model based on the standard instrument detection data and the fused feature data.

2. The skin texture detection method according to claim 1, characterized in that, The spectral image data includes data from multiple spectral channels, which cover the visible light band and the near-infrared band.

3. The skin texture detection method according to claim 1, characterized in that, Preprocessing of the spectral image data of the skin sample includes: Perform one or more of the following operations on the spectral image data of the skin sample: background subtraction, shadow removal, interpolation, or noise reduction.

4. The skin texture detection method according to claim 1, characterized in that, The skin samples include a preset number of skin samples of different genders, ages, and skin types; the multi-dimensional skin quality detection results include quantitative or qualitative assessment results of at least two skin quality indicators among skin color, pigmentation, erythema, sebum secretion level, hydration status, sensitive skin, skin texture, elasticity, or pore characteristics.

5. The skin texture detection method according to claim 1, characterized in that, The machine learning model includes at least one of the following: support vector machine, decision tree, random forest, or neural network model.

6. The skin texture detection method according to claim 1, characterized in that, Establishing the correlation between the spectral image of each preset band and the multiple preset components also includes: Calculate the gender and age correlation coefficients between the gender and age information of the spectral image data for each preset band and the data of the multiple preset components; Based on the gender correlation coefficient and age correlation coefficient, determine the gender weight and age weight of each preset component in the spectral image data of each preset band; The spectral image data of each preset band is weighted using the gender weight and the age weight to form the fused feature data; The machine learning model is trained based on the standard instrument detection data and the fused feature data to obtain the skin texture detection model.

7. The skin texture detection method according to claim 1, characterized in that, The correlation coefficients are calculated based on the Pearson algorithm.

8. The skin texture detection method according to claim 1, characterized in that, The preset ingredients include at least one of melanin, hemoglobin, collagen, oil, or water.

9. The skin texture detection method according to any one of claims 1-8, characterized in that, Also includes: The spectral image data of the skin to be detected is preprocessed; Determine the effective area of ​​the target; Spectral inversion is performed on the spectral image data of the skin to be detected in the target effective region to obtain the spectral image of the target effective region; Based on the spectral image of the target effective region, spectral images of multiple preset bands of the target effective region are obtained; Establish the correlation between the spectral image of each preset band of the target effective region and the multiple preset components.

10. The skin texture detection method according to claim 9, characterized in that, Establishing the correlation between the spectral image of each preset band of the target effective region and the multiple preset components includes: Based on the characteristic bands of the multiple preset components in the spectral image of the skin to be detected, the data of the multiple preset components are obtained; Calculate the correlation coefficient between the spectral image data of each preset band of the target effective region and the data of the multiple preset components; Based on the correlation coefficient, the weight of each preset component in the spectral image data of each preset band is determined; The spectral image data of each preset band are weighted and fused using the weights to form the fused feature data to be detected.

11. The skin texture detection method according to claim 10, characterized in that, The skin spectral image data to be detected is input into the skin texture detection model, and the multi-dimensional skin texture detection results are output, including: The fused feature data to be detected is input into the skin texture detection model, and the multi-dimensional skin texture detection result is output.

12. A skin texture detection device, characterized in that, include: A processor configured to perform the skin texture detection method as described in any one of claims 1-11.

13. The skin texture detection device according to claim 12, characterized in that, Also includes: The system comprises a microlens array, a filter unit array, and a photoelectric sensor array coupled to the processor, arranged along an optical path. Each filter unit includes a filter sub-unit array configured to form multiple spectral channels. The photoelectric sensor is configured to convert incident light signals into electrical signals. The processor is configured to generate spectral image data of the skin sample or the skin to be tested based on the electrical signals. The spectral image data includes data from the multiple spectral channels, which cover the visible light band and the near-infrared band.

14. A computer-readable storage medium, characterized in that, It includes computer-executable instructions stored thereon, which, when executed by a processor, implement the skin texture detection method as described in any one of claims 1-11.

Images