Tongue coating separation method, system, terminal, and medium

Abstract

This invention provides a method, system, terminal, and medium for separating tongue coating and tongue texture. It involves fusing multiple acquired RGB tongue images using color space channels, then performing Gaussian Mixture Model (GMM) clustering on the resulting fused images to construct a training dataset. This training dataset is then used to train a deep segmentation network to obtain a corresponding tongue coating and tongue texture separation model, enabling the generation of tongue coating and tongue texture separation results based on the input RGB tongue image. This invention achieves tongue coating and tongue texture separation in the objectification of traditional Chinese medicine tongue imagery through a tongue coating and tongue texture separation method based on color multi-channel GMM clustering and deep learning, exhibiting good prediction accuracy and performance.

Description

Technical Field

[0001] This invention relates to the field of tongue image segmentation, and in particular to a method, system, terminal, and medium for separating tongue coating. Background Technology

[0002] Tongue diagnosis primarily involves observing changes in two aspects: the tongue body and the tongue coating. The tongue body, also known as the tongue's muscular and vascular tissue, is the tongue's surface. The tongue coating is a layer of moss-like substance adhering to the tongue body, produced by the stomach's qi. During the pathological changes of disease, the balance of yin and yang, the harmony of qi and blood, and the presence or absence of body fluids are all directly reflected in changes in the tongue coating. Observing changes in the tongue coating allows one to understand the state of the internal organs, the pathological changes in internal organs, and the severity of the disease.

[0003] The objective study of tongue diagnosis in Traditional Chinese Medicine (TCM) involves qualitative, quantitative, and locational analysis of tongue appearance using objective observation indicators. Traditional qualitative and quantitative standards for tongue diagnosis are influenced by the physician's academic level and clinical experience, resulting in a degree of uncertainty. The focus of objective tongue diagnosis research is primarily on exploring specific indicators that directly reflect various tongue appearances in TCM. With the development of information technology, objective TCM tongue diagnosis research has gradually begun to utilize computer equipment to analyze images of human tongues, automatically calculating detailed information on various tongue characteristics. The separation of tongue coating and tongue body is a crucial step in the objective tongue diagnosis process; the accuracy of this segmentation directly affects the feasibility and accuracy of subsequent objective tongue diagnosis.

[0004] Early methods of separating tongue coating and tongue body typically involved manual observation and identification. This method had several drawbacks: manual selection was affected by the human eye and display devices, making pixel-level separation difficult; it was time-consuming and required significant manpower and resources, hindering scalability; and it was susceptible to the operator's experience and subjective biases, making it difficult to establish a unified standard. In recent years, with the development of objectification in Traditional Chinese Medicine (TCM), tongue coating and tongue body separation has received widespread attention and research. The main research methods include threshold-based segmentation and cluster-based segmentation. Threshold-based segmentation requires TCM experts and researchers to determine the final threshold experimentally. However, due to the diverse colors of tongue coating and tongue body, it is difficult to achieve universal segmentation using a single threshold, especially at the boundary between the coating and body. Cluster-based segmentation primarily uses the k-means clustering algorithm to separate the coating and body. This method uses continuous iterative clustering by a computer to automatically cluster the tongue image into two categories: tongue coating and tongue body, achieving separation and avoiding the complex threshold selection process. However, the k-means algorithm is affected by initial values ​​and outliers, resulting in unstable results each time. It is also prone to local optima during convergence iterations, leading to poor or even failed tongue coating and tongue body separation. Furthermore, the iterative clustering process involves repeated calculations, which is time-consuming. In Traditional Chinese Medicine (TCM) tongue diagnosis, tongue body colors include pale red, red, crimson, white, and purple, while tongue coating colors include white, yellowish-white, yellow, grayish-black, and green. Currently, single-channel images commonly used in TCM imaging typically only provide good separation results for tongue coating and tongue body of a specific color, even when using k-means clustering. In tongue coating and tongue body separation, the shape, size, and distribution of the tongue coating are varied, and some coatings are even discretely distributed, making manual annotation extremely difficult. Summary of the Invention

[0005] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a method, system, terminal and medium for separating tongue coating to solve the problems of the prior art.

[0006] To achieve the above and other related objectives, this invention provides a method for separating tongue coating and tongue texture. The method includes: performing color space channel fusion processing on multiple acquired RGB tongue images to obtain corresponding fused images; performing GMM clustering on each fused image to construct a training dataset; wherein, the training dataset includes: each RGB tongue image and the corresponding tongue coating and tongue texture separation results; training a deep segmentation network using the training dataset to obtain a corresponding tongue coating and tongue texture separation model; and obtaining the corresponding tongue coating and tongue texture separation results based on the tongue coating and tongue texture separation model and the input RGB tongue image to be tested.

[0007] In one embodiment of the present invention, color space channel fusion processing is performed on multiple acquired RGB tongue images to obtain corresponding fused images, including: converting each RGB tongue image from the RGB color space to the YUV color space and the CIELAB color space respectively to obtain the YUV tongue image and CIELAB tongue image corresponding to each RGB tongue image; extracting the G channel data from each RGB tongue image, the V channel data from each YUV tongue image, and the a* channel data from each CIELAB tongue image to obtain the G channel data, V channel data, and a* channel data corresponding to each RGB tongue image; and fusing the G channel data, V channel data, and a* channel data corresponding to the same RGB tongue image to obtain the fused image corresponding to each RGB tongue image.

[0008] In one embodiment of the present invention, fusing the G channel data, V channel data, and a* channel data in the same RGB tongue image to obtain a fused image corresponding to each RGB tongue image includes: fusing the G channel data, V channel data, and a* channel data in the same RGB tongue image based on the determined weight coefficients corresponding to the G channel data, V channel data, and a* channel data respectively to obtain a fused image corresponding to each RGB tongue image.

[0009] In one embodiment of the present invention, the step of constructing a training dataset based on GMM clustering of each fused image includes: preprocessing each fused image to obtain a corresponding fused processed image; clustering each fused processed image based on the GMM clustering algorithm to obtain a corresponding GMM clustering result; and constructing the training dataset based on the RGB tongue image and the tongue coating and tongue body separation results of each RGB tongue image determined by correcting each GMM clustering result.

[0010] In one embodiment of the present invention, the preprocessing of each fused image includes: normalizing, histogram equalization and gamma correction of each fused image.

[0011] In one embodiment of the present invention, the deep segmentation network is a UNet network.

[0012] In one embodiment of the present invention, the downsampling structure of the UNet network used in the depth segmentation network is to use a dilated convolution structure instead of a max pooling layer.

[0013] To achieve the above and other related objectives, this invention provides a tongue image and tongue coating separation system. The system includes: a color space channel fusion module, used to perform color space channel fusion processing on multiple acquired RGB tongue images to obtain corresponding fused images; a training dataset construction module, connected to the color space channel fusion module, used to perform GMM clustering based on each fused image to construct a training dataset; wherein the training dataset includes: each RGB tongue image and the corresponding tongue coating and tongue coating separation results; a model training module, connected to the training dataset construction module, used to train a deep segmentation network using the training dataset to obtain a corresponding tongue image and tongue coating separation model; and a tongue coating and tongue coating separation module, connected to the model training module, used to obtain the corresponding tongue coating and tongue coating separation results based on the tongue image and tongue coating separation model and the input RGB tongue image to be tested.

[0014] To achieve the above and other related objectives, the present invention provides a tongue image coating separation terminal, comprising: one or more memory and one or more processors; the one or more memory is used to store a computer program; the one or more processors are connected to the memory and are used to run the computer program to perform the tongue image coating separation method.

[0015] To achieve the above and other related objectives, the present invention provides a computer-readable storage medium storing a computer program that is executed by one or more processors to perform the tongue coating separation method.

[0016] To achieve the above and other related objectives, the present invention provides a computer-readable storage medium, characterized in that it stores a computer program, which is executed by one or more processors to perform the method described thereon.

[0017] As described above, this invention provides a method, system, terminal, and medium for separating tongue coating and tongue texture, offering the following advantages: This invention performs color space channel fusion processing on multiple acquired RGB tongue images, then performs GMM clustering on the resulting fused images to construct a training dataset. Using this training dataset, a deep segmentation network is trained to obtain the corresponding tongue coating and tongue texture separation model, enabling the acquisition of corresponding tongue coating and tongue texture separation results based on the input RGB tongue image to be tested. This invention achieves tongue coating and tongue texture separation in the objectification of traditional Chinese medicine tongue images through a tongue coating and tongue texture separation method based on color multi-channel GMM clustering and deep learning, exhibiting good prediction accuracy and performance. Attached Figure Description

[0018] Figure 1 The diagram shown is a flowchart of a tongue coating separation method according to an embodiment of the present invention.

[0019] Figure 2The diagram shown is a structural schematic of the UNet network in one embodiment of the present invention.

[0020] Figure 3 The diagram shown is a structural schematic of a tongue coating separation system according to an embodiment of the present invention.

[0021] Figure 4 The diagram shown is a structural schematic of a tongue-shaped coating separation terminal according to an embodiment of the present invention. Detailed Implementation

[0022] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0023] It should be noted that in the following description, reference is made to the accompanying drawings, which illustrate several embodiments of the invention. It should be understood that other embodiments may also be used, and changes in mechanical composition, structure, electrical system, and operation may be made without departing from the spirit and scope of the invention. The following detailed description should not be considered limiting, and the scope of the embodiments of the invention is defined only by the claims of the published patents. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. Spatially related terms, such as “upper,” “lower,” “left,” “right,” “below,” “below,” “lower part,” “above,” “upper part,” etc., may be used herein to illustrate the relationship between one element or feature shown in the figures and another element or feature.

[0024] Throughout this specification, when it is said that a part is "connected" to another part, this includes not only "direct connection" but also "indirect connection" by placing other elements in between. Furthermore, when it is said that a part "includes" a certain constituent element, unless otherwise stated otherwise, this does not exclude other constituent elements, but rather means that other constituent elements may also be included.

[0025] The terms "first," "second," and "third," etc., used herein are for the purpose of describing various parts, components, regions, layers, and / or segments, but are not limiting. These terms are used only to distinguish one part, component, region, layer, or segment from others. Therefore, the "first part," "component," "region," "layer," or "segment" described below may refer to a "second part," "component," "region," "layer," or "segment" without departing from the scope of this invention.

[0026] Furthermore, as used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context indicates otherwise. It should be further understood that the terms “comprising,” “including,” indicate the presence of the stated feature, operation, element, component, item, kind, and / or group, but do not preclude the presence, occurrence, or addition of one or more other features, operations, elements, components, items, kinds, and / or groups. The terms “or” and “and / or” as used herein are interpreted as inclusive, or mean any one or any combination thereof. Thus, “A, B, or C” or “A, B, and / or C” means “any one of: A; B; C; A and B; A and C; B and C; A, B, and C.” Exceptions to this definition arise only when combinations of elements, functions, or operations are inherently mutually exclusive in some manner.

[0027] This invention discloses a method for separating tongue coating and tongue texture in traditional Chinese medicine. This method involves fusing multiple RGB tongue images in color space, then performing Gaussian Mixture Model (GMM) clustering on the fused images to construct a training dataset. A deep segmentation network is trained using this training dataset to obtain a corresponding tongue coating and tongue texture separation model. This allows for the generation of tongue coating and tongue texture separation results based on the input RGB tongue image. This invention achieves tongue coating and tongue texture separation in the objectification of traditional Chinese medicine tongue images through a tongue coating and tongue texture separation method based on color multi-channel GMM clustering and deep learning, exhibiting good prediction accuracy and performance.

[0028] The present invention will now be described in detail with reference to the accompanying drawings, so that those skilled in the art can readily implement it. The present invention can be embodied in many different forms and is not limited to the embodiments described herein.

[0029] like Figure 1 A flowchart illustrating a method for separating tongue coating texture according to an embodiment of the present invention is shown.

[0030] The method includes:

[0031] Step S1: Perform color space channel fusion processing on the acquired multiple RGB tongue images to obtain the corresponding fused image.

[0032] Specifically, the device for acquiring various RGB tongue images can be a tongue image analyzer, collecting a sufficient number of images covering three age groups: old, middle-aged, and young, including male and female tongue images, as well as tongue images with different coating and body colors. For example, acquiring 3000 RGB tongue images. In traditional Chinese medicine tongue diagnosis, tongue body colors include light red, red, crimson, white, and purple, while tongue coating colors include white, yellowish-white, yellow, grayish-black, and green.

[0033] In one embodiment, step S1 includes:

[0034] Each RGB tongue image is converted from the RGB color space to the YUV color space and the CIELAB color space respectively to obtain the YUV tongue image and CIELAB tongue image corresponding to each RGB tongue image.

[0035] Specifically, this step involves color space transformation, converting each RGB tongue image from the RGB color space to the YUV color space and the CIELAB color space, respectively.

[0036] The RGB color space of the RGB tongue image, where R, G, and B are the three channels of the RGB color space, and their values ​​range from [0, 255].

[0037] In the context of YUV color space conversion, YUV is an image color encoding method used to compile a true color space, commonly used in various image processing components. The terms Y'UV, YUV, YCbCr, and YPbPr can all be referred to as YUV, and there is some overlap between them. Y represents luminance (Luminance, Luma) information, while U and V represent chrominance (Chrominance, Chroma) information, respectively.

[0038] Equation (1-5) is the specific formula for converting RGB to YUV color space.

[0039] Y = 16 + 219 × (W) r ×R+W g ×G+W b ×B) / 255; (1)

[0040] Cb=128+224×(U max ×(BY) / (1-W b )) / 255; (2)

[0041] Cr = 128 + 224 × (V) max ×(RY) / (1-W r )) / 255; (3)

[0042] W r +W g +W b =1; (4)

[0043] Umax = Vmax = 0.5; (5)

[0044] Where Wr, Wg, and Wb are the weighting coefficients of each channel in the RGB color space; in the BT601 international standard, Wr = 0.299, Wg = 0.587, and Wb = 0.114, thus deriving:

[0045] Y=16+0.257×R+0.504×G+0.098×B; (6)

[0046] Cb=128-0.148×R-0.291×G+0.439×B; (7)

[0047] Cr=128+0.439×R-0.368×G-0.071×B; (8)

[0048] Therefore, each RGB tongue image can be converted from the RGB color space to the YUV color space to obtain a YUV tongue image.

[0049] The CIELAB color space (also known as L*a*b*) was defined by the International Commission on Illumination (ICI) in 1976. CIELAB uses three variables to represent color: L* represents perceived brightness, and a* and b* represent red, green, blue, and yellow, respectively. RGB color space cannot be directly converted to CIELAB; it requires CIEXYZ as an intermediate variable before conversion to L*a*b*.

[0050] The RGB colors are subjected to a non-linear gamma hue transformation, as shown in formula (9-12):

[0051]

[0052]

[0053]

[0054]

[0055] The RGB values ​​after gamma transformation are converted to CIEXYZ, and equation (13) is the conversion formula:

[0056]

[0057] Where M is typically selected as:

[0058]

[0059] Substituting (14) into (13) gives:

[0060]

[0061] Therefore, each RGB tongue image is converted from the RGB color space to the CIEXYZ color space:

[0062] Next, the image needs to be converted from the CIEXYZ color space to the CIELAB color space, using equation (16-19):

[0063]

[0064]

[0065]

[0066]

[0067] In equation (16-19): X n Y n Z n The values ​​are 95.047, 100.0, and 108.883 respectively.

[0068] Therefore, the image has been converted from the CIEXYZ color space to the CIELAB color space, thereby converting each RGB tongue image from the RGB color space to the CIEXYZ color space and obtaining each CIELAB tongue image.

[0069] After conversion, the G channel data, V channel data, and a* channel data of each RGB tongue image, are extracted sequentially from each YUV tongue image and each CIELAB tongue image to obtain the corresponding G channel data, V channel data, and a* channel data of each RGB tongue image.

[0070] The G channel data, V channel data, and a* channel data of the same RGB tongue image are fused to obtain the fused image of each RGB tongue image.

[0071] In one embodiment, the G channel data, V channel data, and a* channel data corresponding to the same RGB tongue image are fused to obtain a fused image corresponding to each RGB tongue image, including:

[0072] Based on the determined weight coefficients corresponding to the G channel data, V channel data, and a* channel data, the G channel data, V channel data, and a* channel data in the same RGB tongue image are fused to obtain the fused image of each RGB tongue image.

[0073] Specifically, the G channel in the RGB color space, the V channel in YUV, and the a* channel in CIELAB are fused. The fusion method is to multiply G, V, and a* by their respective weights and then add them together. The fused mixed channel is then normalized and restored to the [0-255] color range. Equation (20) is the fusion formula:

[0074] Mix = Wg G+W v V+W a a * (20)

[0075] In the formula W g W V W a G, V, a * The weighting coefficient of the channel.

[0076] Preferably, the weighting coefficients W corresponding to the G channel data, V channel data, and a* channel data are respectively... g W V W a The determination method is to set random initial weights, update the weight coefficients by controlling variables, compare the updated results with the labeled grayscale image, and repeat this step until the weight coefficients with the best fitting effect are obtained.

[0077] Step S2: Perform GMM clustering based on each fused image to construct a training dataset.

[0078] In detail, the training dataset includes: each RGB tongue image and the corresponding tongue coating and tongue body separation results.

[0079] In one embodiment, step S2 includes:

[0080] Each fused image is preprocessed to obtain the corresponding fused image;

[0081] The GMM clustering algorithm is used to cluster the fused images to obtain the corresponding GMM clustering results;

[0082] Among them, the GMM clustering algorithm, namely Gaussian mixture clustering algorithm, uses a probability model to express the cluster of each sample. When the sample data X is one-dimensional data, the probability density function of the Gaussian distribution is given by equation (21):

[0083]

[0084] Where μ is the data mean and σ is the data standard deviation.

[0085] A Gaussian mixture model can be viewed as a model composed of K individual Gaussian models. The probability distribution of a Gaussian mixture model can be represented by equation (22):

[0086]

[0087] In the formula, K represents the number of sub-Gaussian models in the mixture model, k = 1, 2, 3, ..., K; α k Let α be the probability that the observed data belongs to the k-th sub-model.k ≥0, parameter The expected value, variance, and probability of occurrence in the GMM model for each sub-model.

[0088] The training dataset is constructed based on the RGB tongue images and the tongue coating and tongue body separation results of each RGB tongue image determined by correcting the GMM clustering results.

[0089] Specifically, the GMM clustering results corresponding to each RGB tongue image are manually reviewed and corrected, and the corrected tongue coating and tongue body separation results are used to train the deep segmentation network.

[0090] In one embodiment, the preprocessing of each fused image includes: normalizing, histogram equalization, and gamma correction of each fused image.

[0091] For each fused image, normalization is performed separately, and equation (23) is used to normalize and restore it to the [0-255] color range:

[0092]

[0093] In the formula, Mix is ​​the mixed channel obtained by formula (20), and Mmax is the maximum value in Mix.

[0094] Histogram equalization is a method in image processing that adjusts contrast using an image histogram. This method allows brightness to be better distributed across the histogram. This enables the enhancement of local contrast without affecting overall contrast; histogram equalization achieves this by effectively expanding commonly used brightness levels.

[0095] Gamma correction, also known as gamma nonlinearity or gamma encoding, is used to perform nonlinear calculations or inverse calculations on the luminance or tristimulus values ​​of light in a film or imaging system.

[0096] Step S3: Train a deep segmentation network using the training dataset to obtain the corresponding tongue image and coating separation model.

[0097] In one embodiment, the deep segmentation network employs a UNet network.

[0098] The UNet network is mainly used in medical image segmentation tasks. Its network structure is simple, its performance is efficient, and it can effectively utilize the spatial structure of image features and the contextual information between pixels. It has good segmentation results even when there is a shortage of high-quality medical labeled image data.

[0099] Figure 2 The diagram shows the UNet network architecture. The UNet network mainly consists of encoder downsampling, decoder upsampling, and skip connections. The encoder part is composed of four MPCR (MaxPool, Conv3×3, ReLU) structures. In each MPCR structure, there is a 2×2 Max Pool layer, followed by a Conv3×3 convolutional layer. After the first convolutional layer is a ReLU activation function layer, followed by a second Conv3×3 convolutional layer, and then another ReLU activation function layer. The decoder part is composed of four UPCR (UP-conv, Conv3×3, ReLU) structures, where UP-conv uses a 2×2 transposed convolution kernel, and Conv3×3 uses a 3×3 ordinary convolution kernel. The final output layer is a 1×1 ordinary convolutional layer.

[0100] In a preferred embodiment, the downsampling structure of the UNet network used in the deep segmentation network uses a dilated convolutional structure instead of a max pooling layer to reduce the loss of feature information caused by downsampling.

[0101] Step S4: Based on the tongue image and tongue texture separation model, obtain the corresponding tongue coating and tongue texture separation results according to the input RGB tongue image to be tested.

[0102] In detail, based on the constructed tongue image coating and tongue body separation model, the corresponding tongue coating and tongue body separation results are output according to the input RGB tongue image to be tested, and the tongue coating and tongue body separation results of the RGB tongue image to be tested can be obtained directly.

[0103] To better illustrate the above-mentioned method for separating tongue coating, the present invention provides the following specific embodiments.

[0104] Example 1: A method for separating tongue coating based on color multi-channel GMM clustering and deep learning.

[0105] The method includes:

[0106] The experiment used 3000 tongue images of patients acquired by a tongue imaging system (model GMSX001). This image data covered three age groups: elderly, middle-aged, and young, and included both male and female tongue images. The original RGB tongue images, the converted YUV images, and the CIELAB images were separated into channels. The G, V, and a* channels were then merged and gamma correction was applied. The merged channel image shows that the tongue coating and tongue texture are significantly more distinct compared to other single-channel images.

[0107] To verify the practical effectiveness of the proposed method, three tongue images—one with cracks, one with yellow coating, and one with a thin white coating—were randomly selected from a validation set of 3000 images for preliminary testing. GMM clustering was used to predict the separation of tongue coating and texture from 10000 images. Of the predicted results, 77.69% required no manual correction, 18.01% required manual correction of the predicted edges, and 4.3% were poor and required manual correction. The corrected images were then used to train a deep segmentation network.

[0108] Tongue images were randomly selected from a set of 3000 tongue images for comparison of the clustering effects of UNet network and GMM. The results showed that the tongue coating separation boundary was more obvious and the separation result was more accurate after using UNet network.

[0109] To further verify the accuracy and stability of the method, a validation set of 3000 tongue images was tested in the same manner, and the tongue coating separation results were evaluated by five professional TCM researchers. The evaluation method involved the five researchers voting on the satisfaction level of the separation effect without communicating with each other. Table 1 shows the comparative evaluation performance of this scheme. The test results on the 3000 independent test sets show that the UNet method proposed in this paper performs better in terms of prediction speed, Dice Score, and satisfaction.

[0110] Table 1: Comparison and Evaluation Table of this Plan

[0111]

[0112] This embodiment proposes a tongue coating and tongue texture separation method based on color multi-channel GMM clustering and deep learning to achieve tongue coating and tongue texture separation in the objectification of tongue images in traditional Chinese medicine. The feasibility of this method was tested using 3000 tongue images, and its prediction accuracy and performance showed good results, indicating that the method has certain feasibility and practical value. As the number of collected tongue images increases, further research will be conducted on tongue images. This will involve combining pre-trained deep learning segmentation networks such as UNet and U2Net as pre-labeling models to pre-label existing data and incorporating reinforcement learning mechanisms to continuously enhance the model's prediction capabilities.

[0113] Similar in principle to the above embodiments, the present invention provides a tongue coating separation system.

[0114] The following specific embodiments are provided in conjunction with the accompanying drawings:

[0115] like Figure 3 A schematic diagram of a tongue coating separation system according to an embodiment of the present invention is shown.

[0116] The system includes:

[0117] The color space channel fusion module 31 is used to perform color space channel fusion processing on multiple acquired RGB tongue images to obtain the corresponding fused image;

[0118] The training dataset construction module 32 is connected to the color space channel fusion module 31 and is used to perform GMM clustering based on each fused image to construct a training dataset; wherein, the training dataset includes: each RGB tongue image and the corresponding tongue coating and tongue body separation results;

[0119] The model training module 33 is connected to the training dataset construction module 32 and is used to train a deep segmentation network using the training dataset to obtain the corresponding tongue image coating separation model.

[0120] The tongue coating and tongue body separation module 34 is connected to the model training module 33 and is used to obtain the corresponding tongue coating and tongue body separation results based on the input RGB tongue image to be tested, according to the tongue image and tongue body separation model.

[0121] It should be noted that, as should be understood Figure 3 The division of modules in the system embodiment is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software through processing element calls; they can be implemented entirely in hardware; or some modules can be implemented through processing element calls in software, while others are implemented in hardware.

[0122] For example, each module can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more digital signal processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together to form a system-on-a-chip (SOC).

[0123] Since the implementation principle of the tongue coating separation system has been described in the foregoing embodiments, it will not be repeated here.

[0124] In one embodiment, the color space channel fusion module 31 is used to convert each RGB tongue image from the RGB color space to the YUV color space and the CIELAB color space respectively, to obtain the YUV tongue image and CIELAB tongue image corresponding to each RGB tongue image; extract the G channel data from each RGB tongue image, the V channel data from each YUV tongue image, and the a* channel data from each CIELAB tongue image to obtain the G channel data, V channel data, and a* channel data corresponding to each RGB tongue image; and fuse the G channel data, V channel data, and a* channel data corresponding to the same RGB tongue image to obtain the fused image corresponding to each RGB tongue image.

[0125] In one embodiment, fusing the G channel data, V channel data, and a* channel data in the same RGB tongue image to obtain a fused image for each RGB tongue image includes: fusing the G channel data, V channel data, and a* channel data in the same RGB tongue image based on the determined weight coefficients corresponding to the G channel data, V channel data, and a* channel data respectively to obtain a fused image for each RGB tongue image.

[0126] In one embodiment, the training dataset construction module 32 is used to preprocess each fused image to obtain the corresponding fused processed image; to cluster each fused processed image based on the GMM clustering algorithm to obtain the corresponding GMM clustering result; and to construct the training dataset based on the RGB tongue image and the tongue coating and tongue body separation results of each RGB tongue image determined by correcting each GMM clustering result.

[0127] In one embodiment, the preprocessing of each fused image includes: normalizing, histogram equalization, and gamma correction of each fused image.

[0128] In one embodiment, the deep segmentation network employs a UNet network.

[0129] In one embodiment, the downsampling structure of the UNet network used in the deep segmentation network employs a dilated convolutional structure instead of a max pooling layer.

[0130] like Figure 4 A schematic diagram of the structure of the tongue image texture separation terminal 40 in an embodiment of the present invention is shown.

[0131] The tongue coating separation terminal 40 includes a memory 41 and a processor 42. The memory 41 stores computer programs; the processor 42 runs the computer programs to implement, for example... Figure 1 The method for separating tongue coating texture.

[0132] Optionally, the number of memories 41 can be one or more, and the number of processors 42 can be one or more. Figure 1 Each example is taken as an instance.

[0133] Optionally, the processor 42 in the tongue coating separation terminal 40 will perform the following... Figure 1 The steps described involve loading one or more instructions corresponding to the process of an application into memory 41, and having the processor 42 run the application stored in the first memory 41, thereby achieving the following: Figure 1 The various functions of the tongue coating separation method.

[0134] Optionally, the memory 41 may include, but is not limited to, high-speed random access memory and non-volatile memory. For example, one or more disk storage devices, flash memory devices, or other non-volatile solid-state storage devices; the processor 42 may include, but is not limited to, a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0135] Optionally, the processor 42 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0136] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed, implements as follows: Figure 1 The method for separating tongue coating is shown. The computer-readable storage medium may include, but is not limited to, floppy disks, optical disks, CD-ROMs (Read-Only Optical Disk Memory), magneto-optical disks, ROMs (Read-Only Memory), RAMs (Random Access Memory), EPROMs (Erasable Programmable Read-Only Memory), EEPROMs (Electrically Erasable Programmable Read-Only Memory), magnetic cards or optical cards, flash memory, or other types of media / machine-readable media suitable for storing machine-executable instructions. The computer-readable storage medium may be a product not connected to a computer device or a component used with a computer device.

[0137] In summary, the tongue image and coating separation method, system, terminal, and medium of this invention perform color space channel fusion processing on multiple acquired RGB tongue image images, then perform GMM clustering on the obtained fused images to construct a training dataset. Using this training dataset, a deep segmentation network is trained to obtain the corresponding tongue image and coating separation model, thereby achieving the corresponding tongue coating and coating separation result based on the input RGB tongue image to be tested. This invention achieves tongue coating and coating separation in the objectification of traditional Chinese medicine tongue images through a tongue image and coating separation method based on color multi-channel GMM clustering and deep learning, exhibiting good prediction accuracy and performance. Therefore, this invention effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0138] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for separating tongue coating, characterized in that, The method includes: Multiple RGB tongue images were subjected to color space channel fusion processing to obtain corresponding fused images. This process included: converting each RGB tongue image from the RGB color space to the YUV and CIELAB color spaces respectively, obtaining the corresponding YUV and CIELAB tongue images; extracting the G channel data from each RGB tongue image, the V channel data from each YUV tongue image, and the a* channel data from each CIELAB tongue image to obtain the corresponding G channel, V channel, and a* channel data for each RGB tongue image; and fusing the G channel, V channel, and a* channel data from the same RGB tongue image to obtain the corresponding fused image for each RGB tongue image. A training dataset is constructed by performing GMM clustering on each fused image. The training dataset includes: each RGB tongue image and the corresponding tongue coating and tongue body separation results. The construction of the training dataset by performing GMM clustering on each fused image includes: preprocessing each fused image to obtain the corresponding fused processed image; clustering each fused processed image using the GMM clustering algorithm to obtain the corresponding GMM clustering results; and constructing the training dataset based on the RGB tongue images and the tongue coating and tongue body separation results of each RGB tongue image determined by correcting the GMM clustering results. A deep segmentation network was trained using the training dataset to obtain the corresponding tongue image coating separation model. Based on the tongue image and tongue texture separation model, the corresponding tongue coating and tongue texture separation results are obtained according to the input RGB tongue image to be tested.

2. The method for separating tongue coating as described in claim 1, characterized in that, The G channel data, V channel data, and a* channel data from the same RGB tongue image are fused to obtain a fused image of each RGB tongue image, including: Based on the determined weight coefficients corresponding to the G channel data, V channel data, and a* channel data, the G channel data, V channel data, and a* channel data in the same RGB tongue image are fused to obtain the fused image of each RGB tongue image.

3. The method for separating tongue coating as described in claim 1, characterized in that, The preprocessing methods for each fused image include: normalization, histogram equalization, and gamma correction for each fused image.

4. The method for separating tongue coating as described in claim 1, characterized in that, The deep segmentation network used is the UNet network.

5. The method for separating tongue coating as described in claim 4, characterized in that, The deep segmentation network uses a UNet network with a dilated convolutional structure instead of a max pooling layer for downsampling.

6. A tongue coating separation system, characterized in that, The system includes: The color space channel fusion module is used to perform color space channel fusion processing on multiple acquired RGB tongue images to obtain corresponding fused images. This process includes: converting each RGB tongue image from the RGB color space to the YUV and CIELAB color spaces respectively, obtaining the corresponding YUV and CIELAB tongue images; extracting the G channel data from each RGB tongue image, the V channel data from each YUV tongue image, and the a* channel data from each CIELAB tongue image to obtain the corresponding G channel, V channel, and a* channel data for each RGB tongue image; and fusing the G channel, V channel, and a* channel data from the same RGB tongue image to obtain the corresponding fused image for each RGB tongue image. A training dataset construction module, connected to the color space channel fusion module, is used to construct a training dataset by performing GMM clustering on each fused image. The training dataset includes: each RGB tongue image and the corresponding tongue coating and tongue body separation results. Constructing the training dataset by performing GMM clustering on each fused image includes: preprocessing each fused image to obtain a corresponding fused image; clustering each fused image using the GMM clustering algorithm to obtain corresponding GMM clustering results; and constructing the training dataset based on the RGB tongue images and the tongue coating and tongue body separation results of each RGB tongue image determined by correcting the GMM clustering results. The model training module is connected to the training dataset construction module and is used to train a deep segmentation network using the training dataset to obtain the corresponding tongue image coating separation model. The tongue coating and tongue body separation module is connected to the model training module and is used to obtain the corresponding tongue coating and tongue body separation results based on the input RGB tongue image to be tested, according to the tongue image and tongue body separation model.

7. A tongue coating separation terminal, characterized in that, include: One or more memories and one or more processors; The one or more memories are used to store computer programs; The one or more processors are connected to the memory and are used to run the computer program to perform the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The device contains a computer program that is executed by one or more processors to perform the method as described in any one of claims 1 to 5.

Images