Virtual human precise beauty makeup area identification method and system based on image segmentation
By using image segmentation and random forest algorithms in beauty area recognition technology, the problems of inaccurate identification and weak generalization ability in the prior art are solved, and accurate recognition and stability improvement of virtual face beauty areas are achieved.
Patent Information
- Application Number
- CN202411595697.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-08
- Publication Date
- 2025-02-28
AI Technical Summary
The existing beauty area recognition technology is inaccurate in the case of complex backgrounds, changes in lighting or diverse face postures, and fails to effectively utilize high-level semantic features, resulting in weak generalization ability.
An image segmentation-based method is adopted, and the steps of face detection, face alignment, image segmentation and feature extraction are combined with a random forest algorithm to accurately identify and position the beauty area.
It realizes high-accuracy recognition of the beauty areas on virtual faces, can adapt to a variety of beauty areas and lighting conditions, and improves the stability and generalization ability of recognition.
Smart Images
Figure CN119541019A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of face image processing, and more specifically to a method and system for accurately identifying beauty areas of virtual humans based on image segmentation. Background Art
[0002] Existing beauty region recognition technologies often rely on simple image processing methods, such as color threshold segmentation, edge detection, etc. These methods are prone to inaccurate beauty region recognition when dealing with complex backgrounds, lighting changes, or diverse facial postures. For example, fine beauty features such as eyebrows, eye shadow, and lip color may be difficult to accurately segment due to similar lighting or colors.
[0003] The diversity of facial postures (such as rotation, tilt, etc.) poses challenges to the recognition of beauty regions. Existing technologies usually do not have an effective face alignment step, resulting in inaccurate positioning of beauty regions, which in turn affects subsequent beauty processing or analysis.
[0004] The features of beauty makeup areas include not only low-level features such as color and texture, but also high-level semantic features such as shape and structure. Existing technologies often only extract some low-level features and ignore high-level semantic features, resulting in limited recognition effect of beauty makeup areas.
[0005] It can be seen that the existing technology usually uses simple models to identify beauty areas. These models have weak generalization capabilities and are difficult to adapt to beauty area recognition tasks under different lighting, skin color, makeup and other conditions. Therefore, how to improve the accuracy of beauty area recognition results is a problem that technical personnel in this field urgently need to solve. Summary of the invention
[0006] In view of this, the present invention provides a method and system for accurately identifying beauty areas of virtual humans based on image segmentation.
[0007] In order to achieve the above object, the present invention provides the following technical solutions:
[0008] A method for accurately identifying beauty areas of virtual humans based on image segmentation comprises the following steps:
[0009] Step 1: Get the image to be detected and use the face detection algorithm to locate the face area in the image;
[0010] Step 2: After the face is detected, the key points in the face are located using the face key point detection technology, and the face is aligned according to the key points, that is, the face image is transformed into a standardized posture and position by an affine transformation method;
[0011] Step 3: Use an image segmentation algorithm to divide the face image into several beauty areas with specific semantics;
[0012] Step 4: extract features from the beauty makeup area obtained by segmentation, and extract feature information that can characterize the characteristics of the area, including at least shape, texture, and color;
[0013] Step 5: Use the extracted features as input and use the random forest algorithm for learning and training. According to the output results of the random forest algorithm, accurate recognition and positioning of the beauty area can be achieved.
[0014] Furthermore, in step 1, the step of using a face detection algorithm to locate the face area in the image is:
[0015] Step 1.1, image acquisition: obtain the image to be detected from a camera, image file or other image source;
[0016] Step 1.2, image preprocessing: preprocess the acquired image to be detected, including adjusting the image size, normalization or standardization to ensure that the image data is suitable for the input requirements of CNN;
[0017] Step 1.3. Select CNN model: Select Faster R-CNN model.
[0018] Step 1.4: Apply the CNN model for face detection: Input the preprocessed image to be detected into the selected Faster R-CNN model. The Faster R-CNN model traverses the entire image to be detected, extracts image features using convolutional layers, pooling layers, and fully connected layers, and uses a classifier to determine whether each candidate region contains a face. Once a face is detected, the Faster R-CNN model outputs the location of the face in the form of a rectangular box.
[0019] Furthermore, in step 2, the key points include at least eyes, eyebrows, nose, and mouth, and the specific method for locating the key points is:
[0020] First, the acquired face image is preprocessed, including grayscale conversion and histogram equalization, to enhance the image contrast and reduce the impact of illumination changes;
[0021] Use SIFT key point detection model to extract image features;
[0022] Based on the extracted features, the positions of a series of key points are predicted, including but not limited to the four corners of the eyes, the ends of the eyebrows, the tip of the nose, and the corners of the mouth.
[0023] Furthermore, in step 2, the specific method for performing face alignment is:
[0024] Using the detected key points, calculate the affine transformation matrix;
[0025] The face image is transformed using the affine transformation matrix, thereby aligning the face to a standard template and eliminating the effects of head rotation and tilt.
[0026] Furthermore, in step 3, the beauty area includes at least skin, eyebrows, eyes, lips, cheeks, nose, hairline, and jawline.
[0027] Furthermore, in step 4, the specific method for extracting the beauty area features is:
[0028] Step 4.1, shape feature extraction: Use the Canny edge detection algorithm to extract the contour of the beauty area; obtain the precise coordinates of the contour through the contour tracking algorithm, and then analyze the shape characteristics of the area;
[0029] Step 4.2, texture feature extraction: By calculating the co-occurrence matrix GLCM of the grayscale image of the beauty area, the texture features are extracted, including contrast, correlation, and energy, which reflect the spatial relationship and grayscale distribution between pixels within the area. The LBP algorithm is used to encode the texture of the beauty area to extract texture features with rotation invariance and grayscale invariance;
[0030] Step 4.3, color feature extraction: Convert the beauty area from RGB color space to HSV color space, calculate the color histogram of the beauty area, and count the distribution and proportion of each color component; use color moments: first-order moment, second-order moment, and third-order moment to describe the color characteristics of the beauty area.
[0031] Furthermore, the specific method of step 5 is: constructing multiple decision trees based on the random forest algorithm, and when constructing the decision trees, selecting the optimal splitting node to obtain an initial random forest model; using the training data set to train the random forest model, during the training process, each tree independently classifies or regresses the training data and outputs its prediction results; integrating the feature information extracted in step 4 to form a multidimensional feature vector, inputting it into the trained random forest model, the model outputs the prediction results of each beauty area, and according to the prediction results of the model, determines the position and range of each beauty area, and accurately locates it.
[0032] A virtual human precise beauty area recognition system based on image segmentation, comprising:
[0033] The face detection module is used to obtain the image to be detected and locate the face area in the image using the face detection algorithm;
[0034] A face alignment module is used to locate key points in the face using face key point detection technology after detecting the face, and perform face alignment based on the key points, that is, transform the face image into a standardized posture and position by an affine transformation method;
[0035] The preliminary region division module is used to divide the face image into several beauty regions with specific semantics using an image segmentation algorithm;
[0036] A region feature extraction module is used to extract features of the beauty makeup region obtained by segmentation, and extract feature information that can characterize the characteristics of the region, including at least shape, texture, and color;
[0037] The precise positioning module is used to take the extracted features as input, use the random forest algorithm for learning and training, and realize the precise identification and positioning of the beauty area according to the output results of the random forest algorithm.
[0038] Furthermore, in the face detection module, locating the face area in the image using the face detection algorithm specifically includes:
[0039] The image acquisition submodule is used to acquire the image to be detected from a camera, an image file or other image sources;
[0040] The image preprocessing submodule is used to preprocess the acquired image to be detected, including adjusting the image size, normalization or standardization to ensure that the image data is suitable for the input requirements of CNN;
[0041] CNN model selection submodule, used to select the FasterR-CNN model;
[0042] The face detection submodule is used to input the preprocessed image to be detected into the selected Faster R-CNN model. The Faster R-CNN model traverses the entire image to be detected, extracts image features using convolutional layers, pooling layers, and fully connected layers, and uses a classifier to determine whether each candidate area contains a face. Once a face is detected, the Faster R-CNN model outputs the location of the face in the form of a rectangular box.
[0043] Furthermore, in the region feature extraction module, extracting beauty region features specifically includes:
[0044] The shape feature extraction submodule uses the Canny edge detection algorithm to extract the contour of the beauty area; the contour tracking algorithm is used to obtain the precise coordinates of the contour, and then the shape features of the area are analyzed;
[0045] The texture feature extraction submodule calculates the co-occurrence matrix GLCM of the grayscale image of the beauty area to extract texture features, including contrast, correlation, and energy, which reflects the spatial relationship and grayscale distribution between pixels within the area. The LBP algorithm is used to encode the texture of the beauty area to extract texture features with rotation invariance and grayscale invariance.
[0046] The color feature extraction submodule converts the beauty area from the RGB color space to the HSV color space, calculates the color histogram of the beauty area, and counts the distribution and proportion of each color component; it uses color moments: first-order moment, second-order moment, and third-order moment to describe the color features of the beauty area.
[0047] It can be seen from the above technical solutions that the present invention provides a method and system for accurately identifying beauty areas of virtual humans based on image segmentation, which has the following beneficial effects compared with the prior art:
[0048] (1) The present invention combines multiple steps such as face detection, face alignment, image segmentation and feature extraction to achieve accurate recognition of beauty areas on a virtual face. This recognition is not only highly accurate, but also can cover a variety of beauty areas, such as skin, eyebrows, eyes, lips, etc., meeting the sophisticated needs of virtual beauty, face beautification and other fields.
[0049] (2) The present invention utilizes facial key point detection technology and affine transformation method to eliminate the influence of facial posture diversity and illumination changes on beauty area recognition. By aligning the face to a standard template, the subsequent beauty area division and feature extraction are more accurate and stable.
[0050] (3) In the feature extraction stage, the present invention not only considers the low-level features such as shape, texture and color of the beauty area, but also extracts information that can comprehensively describe the characteristics of the beauty area by calculating high-level features such as the co-occurrence matrix GLCM, LBP texture coding and color moment, thus providing strong support for subsequent precise positioning.
[0051] (4) The present invention adopts the random forest algorithm for learning and training, which can fully utilize the parallel processing capability and generalization performance of multiple decision trees to achieve rapid and accurate recognition and positioning of beauty areas. It not only has a fast processing speed but also has good robustness and scalability.
[0052] In summary, the present invention has made significant progress in precise beauty area recognition, which not only improves the accuracy and stability of recognition, but also provides strong support for the development of virtual beauty technology. It has broad application prospects and important practical significance in the fields of face beautification and virtual makeup trial. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0054] Figure 1 It is a schematic diagram of the method steps of the present invention;
[0055] Figure 2 It is a schematic diagram of the system module of the present invention. DETAILED DESCRIPTION
[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0057] The embodiment of the present invention discloses a method for accurately identifying beauty areas of virtual humans based on image segmentation, see Figure 1 , including the following steps:
[0058] Step 1: Get the image to be detected and use the face detection algorithm to locate the face area in the image. Specifically:
[0059] Step 1.1, image acquisition: obtain the image to be detected from a camera, image file or other image source;
[0060] Step 1.2, image preprocessing: preprocess the acquired image to be detected, including adjusting the image size, normalization or standardization to ensure that the image data is suitable for the input requirements of CNN;
[0061] Resize images: According to the input requirements of the Faster R-CNN model, the images are resized to a fixed size (such as 600x600 or 800x800), while using a bilinear interpolation method to maintain the clarity and proportion of the image.
[0062] Normalization: Adjust the pixel value range of an image to [0,1] or [-1,1].
[0063] Standardization: Calculate the mean and standard deviation of the image, subtract the mean from each pixel value and divide it by the standard deviation to make the image data conform to the normal distribution.
[0064] Step 1.3. Select CNN model: Select Faster R-CNN model.
[0065] Select a pre-trained Faster R-CNN model, such as a model based on a backbone network such as VGG16, ResNet, etc., which is not limited by the present invention. Step 1.4, apply the CNN model for face detection: input the preprocessed image to be detected into the selected Faster R-CNN model, the Faster R-CNN model traverses the entire image to be detected, uses the convolution layer to extract image features, the pooling layer is used to reduce the dimension and calculation amount of the feature map, and the fully connected layer is used to further process and classify the features. The region candidate network (RPN) generates a series of candidate regions, and uses the classifier and regressor to determine whether each candidate region contains a face. Once a face is detected, the Faster R-CNN model will output the bounding box (rectangular box) of the face, including the position and size of the face. According to the coordinates of the bounding box, a rectangular box can be drawn on the original image to visualize the position of the face.
[0066] Step 2: After detecting the face, use the facial key point detection technology to locate the key points in the face, and perform face alignment based on the key points, that is, transform the face image into a standardized posture and position through the affine transformation method. The key points include at least eyes, eyebrows, nose, and mouth.
[0067] Among them, the specific process of key point positioning is:
[0068] First, the acquired face image is preprocessed, including grayscale conversion and histogram equalization, to enhance the image contrast and reduce the impact of illumination changes;
[0069] Use SIFT key point detection model to extract image features;
[0070] Based on the extracted features, the positions of a series of key points are predicted, including but not limited to the four corners of the eyes, the ends of the eyebrows, the tip of the nose, and the corners of the mouth.
[0071] After determining the key points, face alignment is performed based on the key points. The specific method is as follows:
[0072] Using the detected key points, calculate the affine transformation matrix;
[0073] The face image is transformed using the affine transformation matrix, thereby aligning the face to a standard template and eliminating the effects of head rotation and tilt.
[0074] Step 3: Use an image segmentation algorithm to divide the face image into several beauty regions with specific semantics, where the beauty regions include at least skin, eyebrows, eyes, lips, cheeks, nose, hairline, jawline, etc. In this implementation, a deep learning algorithm Mask R-CNN can be used, which can simultaneously detect faces and generate accurate segmentation masks.
[0075] Step 4: Perform feature extraction on the segmented beauty area to extract feature information that can characterize the characteristics of the area, including at least shape, texture, and color.
[0076] Among them, the specific method for extracting beauty area features is:
[0077] Step 4.1, shape feature extraction: Use the Canny edge detection algorithm to extract the contour of the beauty area; obtain the precise coordinates of the contour through the contour tracking algorithm, and then analyze the shape characteristics of the area;
[0078] Step 4.2, texture feature extraction: By calculating the co-occurrence matrix GLCM of the grayscale image of the beauty area, the texture features are extracted, including contrast, correlation, and energy, which reflect the spatial relationship and grayscale distribution between pixels within the area. The LBP algorithm is used to encode the texture of the beauty area to extract texture features with rotation invariance and grayscale invariance;
[0079] Step 4.3, color feature extraction: Convert the beauty area from RGB color space to HSV color space, calculate the color histogram of the beauty area, and count the distribution and proportion of each color component; use color moments: first-order moment, second-order moment, and third-order moment to describe the color characteristics of the beauty area.
[0080] Step 5: Use the extracted features as input and use the random forest algorithm for learning and training. According to the output results of the random forest algorithm, accurate recognition and positioning of the beauty area can be achieved.
[0081] Construct multiple decision trees according to the random forest algorithm. When constructing the decision trees, select the optimal splitting node to obtain an initial random forest model. Obtain a training data set, which contains a large amount of image data. The image data is processed according to the process of steps 1-4 to obtain a multidimensional feature vector. At the same time, the beauty area on each image in the data set is annotated in advance. The random forest model is trained using the training data set with the multidimensional feature vector and annotation information. During the training process, each tree independently classifies or regresses the training data.
[0082] The feature information extracted in step 4 is integrated to form a multi-dimensional feature vector, which is input into the trained random forest model. The model outputs the prediction results of each beauty area. According to the prediction results of the model, the position and range of each beauty area are determined and accurately located.
[0083] Another embodiment of the present invention also provides a virtual human accurate beauty area recognition system based on image segmentation, see Figure 2 ,include:
[0084] The face detection module is used to obtain the image to be detected and locate the face area in the image using the face detection algorithm;
[0085] A face alignment module is used to locate key points in the face using face key point detection technology after detecting the face, and perform face alignment based on the key points, that is, transform the face image into a standardized posture and position by an affine transformation method;
[0086] The preliminary region division module is used to divide the face image into several beauty regions with specific semantics using an image segmentation algorithm;
[0087] A region feature extraction module is used to extract features of the beauty makeup region obtained by segmentation, and extract feature information that can characterize the characteristics of the region, including at least shape, texture, and color;
[0088] The precise positioning module is used to take the extracted features as input, use the random forest algorithm for learning and training, and realize the precise identification and positioning of the beauty area according to the output results of the random forest algorithm.
[0089] Furthermore, in the face detection module, locating the face area in the image using the face detection algorithm specifically includes:
[0090] The image acquisition submodule is used to acquire the image to be detected from a camera, an image file or other image sources;
[0091] The image preprocessing submodule is used to preprocess the acquired image to be detected, including adjusting the image size, normalization or standardization to ensure that the image data is suitable for the input requirements of CNN;
[0092] CNN model selection submodule, used to select the FasterR-CNN model;
[0093] The face detection submodule is used to input the preprocessed image to be detected into the selected Faster R-CNN model. The Faster R-CNN model traverses the entire image to be detected, extracts image features using convolutional layers, pooling layers, and fully connected layers, and uses a classifier to determine whether each candidate area contains a face. Once a face is detected, the Faster R-CNN model outputs the location of the face in the form of a rectangular box.
[0094] Furthermore, in the region feature extraction module, extracting beauty region features specifically includes:
[0095] The shape feature extraction submodule uses the Canny edge detection algorithm to extract the contour of the beauty area; the contour tracking algorithm is used to obtain the precise coordinates of the contour, and then the shape features of the area are analyzed;
[0096] The texture feature extraction submodule calculates the co-occurrence matrix GLCM of the grayscale image of the beauty area to extract texture features, including contrast, correlation, and energy, which reflects the spatial relationship and grayscale distribution between pixels within the area. The LBP algorithm is used to encode the texture of the beauty area to extract texture features with rotation invariance and grayscale invariance.
[0097] The color feature extraction submodule converts the beauty area from the RGB color space to the HSV color space, calculates the color histogram of the beauty area, and counts the distribution and proportion of each color component; it uses color moments: first-order moment, second-order moment, and third-order moment to describe the color features of the beauty area.
[0098] As for the system module disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.
[0099] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.
[0100] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for accurately identifying beauty areas of virtual humans based on image segmentation, characterized in that: The following steps are involved: Step 1: Get the image to be detected and use the face detection algorithm to locate the face area in the image; Step 2: After the face is detected, the key points in the face are located using the face key point detection technology, and the face is aligned according to the key points, that is, the face image is transformed into a standardized posture and position by an affine transformation method; Step 3: Use an image segmentation algorithm to divide the face image into several beauty areas with specific semantics; Step 4: extract features from the beauty makeup area obtained by segmentation, and extract feature information that can characterize the characteristics of the area, including at least shape, texture, and color; Step 5: Use the extracted features as input and use the random forest algorithm for learning and training. According to the output results of the random forest algorithm, accurate recognition and positioning of the beauty area can be achieved.
2. According to the method of claim 1, the method is characterized by: In step 1, the steps of using a face detection algorithm to locate the face area in the image are: Step 1.1, image acquisition: obtain the image to be detected from a camera, image file or other image source; Step 1.2, image preprocessing: preprocess the acquired image to be detected, including adjusting the image size, normalization or standardization to ensure that the image data is suitable for the input requirements of CNN; Step 1.
3. Select CNN model: Select Faster R-CNN model. Step 1.4: Apply the CNN model for face detection: Input the preprocessed image to be detected into the selected Faster R-CNN model. The Faster R-CNN model traverses the entire image to be detected, extracts image features using convolutional layers, pooling layers, and fully connected layers, and uses a classifier to determine whether each candidate region contains a face. Once a face is detected, the Faster R-CNN model outputs the location of the face in the form of a rectangular box.
3. The method for accurately identifying beauty areas of virtual humans based on image segmentation according to claim 1, characterized in that: In step 2, the key points include at least eyes, eyebrows, nose, and mouth, and the specific method for locating the key points is: First, the acquired face image is preprocessed, including grayscale conversion and histogram equalization, to enhance the image contrast and reduce the impact of illumination changes; Use SIFT key point detection model to extract image features; Based on the extracted features, the positions of a series of key points are predicted, including but not limited to the four corners of the eyes, the ends of the eyebrows, the tip of the nose, and the corners of the mouth.
4. The method for accurately identifying beauty areas of virtual humans based on image segmentation according to claim 3 is characterized in that: In step 2, the specific method for performing face alignment is: Using the detected key points, calculate the affine transformation matrix; The face image is transformed using the affine transformation matrix, thereby aligning the face to a standard template and eliminating the effects of head rotation and tilt.
5. The method for accurately identifying virtual human beauty makeup areas based on image segmentation according to claim 1, characterized in that: In step 3, the beauty area includes at least skin, eyebrows, eyes, lips, cheeks, nose, hairline, and jawline.
6. The method for accurately identifying beauty areas of virtual humans based on image segmentation according to claim 1, characterized in that: In step 4, the specific method for extracting the beauty area features is: Step 4.1, shape feature extraction: Use the Canny edge detection algorithm to extract the contour of the beauty area; obtain the precise coordinates of the contour through the contour tracking algorithm, and then analyze the shape characteristics of the area; Step 4.2, texture feature extraction: By calculating the co-occurrence matrix GLCM of the grayscale image of the beauty area, the texture features are extracted, including contrast, correlation, and energy, which reflect the spatial relationship and grayscale distribution between pixels within the area. The LBP algorithm is used to encode the texture of the beauty area to extract texture features with rotation invariance and grayscale invariance; Step 4.3, color feature extraction: Convert the beauty area from RGB color space to HSV color space, calculate the color histogram of the beauty area, and count the distribution and proportion of each color component; use color moments: first-order moment, second-order moment, and third-order moment to describe the color characteristics of the beauty area.
7. The method for accurately identifying virtual human beauty makeup areas based on image segmentation according to claim 1, characterized in that: The specific method of step 5 is: construct multiple decision trees according to the random forest algorithm, and when constructing the decision trees, select the optimal split node to obtain an initial random forest model; use the training data set to train the random forest model, and during the training process, each tree independently classifies or regresses the training data and outputs its prediction results; integrate the feature information extracted in step 4 to form a multidimensional feature vector, and input it into the trained random forest model. The model outputs the prediction results of each beauty area, and according to the prediction results of the model, determines the position and range of each beauty area, and accurately locates it.
8. A virtual human precise beauty area recognition system based on image segmentation, characterized in that: include: The face detection module is used to obtain the image to be detected and locate the face area in the image using the face detection algorithm; A face alignment module is used to locate key points in the face using face key point detection technology after detecting the face, and perform face alignment based on the key points, that is, transform the face image into a standardized posture and position by an affine transformation method; The preliminary region division module is used to divide the face image into several beauty regions with specific semantics using an image segmentation algorithm; A region feature extraction module is used to extract features of the beauty makeup region obtained by segmentation, and extract feature information that can characterize the characteristics of the region, including at least shape, texture, and color; The precise positioning module is used to take the extracted features as input, use the random forest algorithm for learning and training, and realize the precise identification and positioning of the beauty area according to the output results of the random forest algorithm.
9. The virtual human precise beauty area recognition system based on image segmentation according to claim 8, characterized in that: In the face detection module, using the face detection algorithm to locate the face area in the image specifically includes: The image acquisition submodule is used to acquire the image to be detected from a camera, an image file or other image sources; The image preprocessing submodule is used to preprocess the acquired image to be detected, including adjusting the image size, normalization or standardization to ensure that the image data is suitable for the input requirements of CNN; CNN model selection submodule, used to select the FasterR-CNN model; The face detection submodule is used to input the preprocessed image to be detected into the selected Faster R-CNN model. The Faster R-CNN model traverses the entire image to be detected, extracts image features using convolutional layers, pooling layers, and fully connected layers, and uses a classifier to determine whether each candidate area contains a face. Once a face is detected, the Faster R-CNN model outputs the location of the face in the form of a rectangular box.
10. The virtual human precise beauty area recognition system based on image segmentation according to claim 8, characterized in that: In the region feature extraction module, extracting beauty region features specifically includes: The shape feature extraction submodule uses the Canny edge detection algorithm to extract the contour of the beauty area; the contour tracking algorithm is used to obtain the precise coordinates of the contour, and then the shape features of the area are analyzed; The texture feature extraction submodule calculates the co-occurrence matrix GLCM of the grayscale image of the beauty area to extract texture features, including contrast, correlation, and energy, which reflects the spatial relationship and grayscale distribution between pixels within the area. The LBP algorithm is used to encode the texture of the beauty area to extract texture features with rotation invariance and grayscale invariance. The color feature extraction submodule converts the beauty area from the RGB color space to the HSV color space, calculates the color histogram of the beauty area, and counts the distribution and proportion of each color component; it uses color moments: first-order moment, second-order moment, and third-order moment to describe the color features of the beauty area.