An intelligent try-on system and an intelligent try-on processing method thereof

Abstract

An intelligent virtual fitting system and processing method are disclosed. The system includes a fitting mirror main body equipped with a camera, display screen, and communication device, and an intelligent analysis server connected to it via the communication device. The server includes a body shape recognition module, a skin tone analysis module, a clothing analysis module, a clothing database, a face recognition module, and a data storage module. The processing method involves: capturing user images and transmitting them to the server; extracting human contour, height, and posture information through body shape recognition; generating skin tone codes through skin tone analysis; combining these two factors to select clothing from the clothing database and complete body shape matching; and transmitting the matched outfit image back to the fitting mirror for real-time display. This invention improves the accuracy of body shape recognition and skin tone analysis, achieves precise deformation-fitting of clothing, can reuse user historical data, provides realistic virtual fitting effects, and is easy to operate, making it suitable for scenarios such as clothing retail and smart homes.

Description

1. Description of Intelligent Fitting System and Intelligent Fitting Processing Method Technical Field This invention relates to the field of intelligent fitting technology, specifically, to an intelligent fitting system and intelligent fitting processing method designed using artificial intelligence technology. Background Technology With the rapid development of computer vision technology, artificial intelligence technology, and image processing technology, virtual fitting technology has gradually become a research hotspot in the fields of clothing retail and personalized dressing. Traditional offline fitting methods require users to try on clothes in person, which has drawbacks such as low fitting efficiency, inconvenience of frequent putting on and taking off, and prominent hygiene issues. Especially in retail places with high traffic, the user fitting experience is greatly limited. To solve the above problems, various virtual fitting systems have emerged in the prior art. Some systems use cameras to capture user images and display a composite image of the user and virtual clothing on a display device, allowing users to preview the wearing effect without actually trying on the clothes. However, existing virtual fitting systems still have shortcomings in terms of the accuracy of body shape recognition, the adaptability of skin tone and clothing color, and the personalization of the overall dressing effect. Specifically, existing technologies for body shape recognition mostly employ simple human contour detection or single height estimation algorithms, making it difficult to accurately obtain refined body parameters such as shoulder width, waist circumference, and body shape classification. This results in poor size and fit when matching clothing. In skin color analysis, most systems only perform skin color recognition through simple color sampling or RGB threshold segmentation, lacking precise extraction and quantification of skin color values, thus failing to effectively guide personalized clothing color recommendations. Furthermore, existing systems often use simple image overlay methods during clothing matching, failing to perform refined alignment and deformation fitting based on key points of the human body and clothing, leading to distorted and unnatural fitting effects. To address these technical problems, existing technologies have not yet provided a comprehensive intelligent fitting solution that integrates refined body shape recognition, accurate skin color analysis, and highly realistic clothing fit. Therefore, developing an intelligent fitting system and processing method that can improve the accuracy of body shape recognition, the accuracy of skin color analysis, and the realism of fitting images has significant technical and application value. Summary of the Invention This invention provides an intelligent fitting system and processing method that improves the accuracy of body shape recognition, the accuracy of skin color analysis, and the realism of fitting images.To address the related technical problems, this invention provides the following technical solution: An intelligent fitting room system, comprising: a fitting room mirror body equipped with a camera, a display device, and a communication device; and an intelligent analysis server, which connects to the fitting room mirror body via the communication device and acquires the camera's captured content, including a skin tone analysis module for color analysis of the user's skin tone, a clothing database, and a clothing analysis module; the clothing analysis module selects suitable clothing from the clothing database based on the output of the skin tone analysis module and transmits the selection to the fitting room mirror body via the communication device, displaying a real-time image of the user and the suitable clothing on the display device. Preferably, the intelligent server also includes a body shape recognition module for recognizing the user's body contour, height, and posture; the clothing analysis module selects suitable clothing from the clothing database based on the output of the body shape recognition module and the skin tone analysis module, matches the body shape, transmits the selection to the fitting room mirror body via the communication device, and displays a real-time image of the user and the suitable clothing on the display device. Preferably, the intelligent analysis server also includes a module for facial recognition by extracting facial features from content captured by a camera, and a module for storing user data. The facial recognition module identifies the user and reads the user's skin color, height, weight, body shape, and clothing history from the data storage module, and inputs the above data into the clothing analysis module for clothing matching. The intelligent fitting room processing method includes the following steps: Step S1: Acquire user image data through the camera of the fitting room mirror body; Step S2: Transmit the acquired user image data to the intelligent analysis server via a communication device; Step S3: The intelligent analysis server performs body shape recognition processing on the user image data, extracting the user's body contour, height, and posture information; Step S4: The intelligent analysis server performs skin color analysis processing on the user image data, generating the user's skin color code; Step S5: Based on the body shape recognition results and skin color analysis results, the intelligent analysis server selects suitable clothing from the clothing database and performs body shape matching; Step S6: Transmit the matched clothing combination image to the fitting room mirror body via the communication device, and display the image of the user and the suitable clothing combination in real time through a display device.Preferably, the body shape recognition processing in step S3 includes the following: Step S31: Preprocessing the input human body image, including size normalization, color space conversion, and noise filtering; Step S32: Performing pixel-level semantic segmentation using a pre-trained human body segmentation model to generate a human body region mask, and extracting the external contour of the human body based on the mask; Step S33: Inferring the human body image using a pre-trained human body key point detection model to obtain the coordinate set of human skeletal key points, including the top of the head, neck, shoulders, waist, hips, knees, and soles of the feet; Step S34: Calculating the human height based on the vertical height of the human body contour and the pixel distance between the key points, combined with the mapping relationship between image proportion and physical size; Step S35: Calculating body shape parameters such as shoulder width, waist circumference, hip circumference, waist-to-hip ratio, and shoulder-to-waist ratio based on the key point spacing and the geometric features of the human body contour, and completing body shape classification according to preset rules; The execution steps of steps S34 and S35 are not performed in sequence. Preferably, the skin color analysis processing in step S4 includes the following steps: Step S41: Load the face detection model and segment the face region to extract the skin region; Step S42: Extract the color values ​​of all pixels in the skin region and process them to obtain the skin color code. Preferably, the clothing and body shape matching steps in step S5 include the following: Step S51: Select clothing that matches the skin color code from the clothing database according to the user's skin color code; Step S52: Obtain the image of the clothing to be matched and the clothing key points, clothing size, and clothing pattern information; Step S53: Scale the clothing according to the overall proportion of the human height to match the length of the clothing with the human height; establish a mapping relationship between the human skeletal key points and the clothing key points to align the position of the clothing; according to the human contour and body shape feature parameters, use a deformation transformation algorithm to fit the clothing to the human contour shape, and use the matching clothing layer to occlude and merge the human image to generate a matched clothing matching image; Step S54: Transmit the matched clothing matching image to the main body of the dressing mirror through a communication device, and display the matching image of the user and the suitable clothing in real time through a display device. The beneficial effects of this invention: Compared with the prior art, the intelligent fitting system and intelligent fitting processing method provided by this invention have the following beneficial effects: This invention achieves pixel-level semantic segmentation and precise localization of skeletal key points in human images through the collaborative processing of a pre-trained human segmentation model and a human key point detection model. Based on the pixel distance between the vertical height of the human contour and the key points, combined with the mapping relationship between image proportion and physical size, the height of the human body is accurately calculated; at the same time, based on the key point spacing and the geometric features of the human contour, multiple body posture parameters such as shoulder width, waist circumference, hip circumference, waist-to-hip ratio, and shoulder-to-waist ratio are calculated, and body type classification is completed according to preset rules.Compared to existing technologies that rely solely on simple contour detection or single height estimation, this invention can acquire more comprehensive and accurate body features, providing reliable data support for subsequent clothing matching. This invention segments the face region using a face detection model, accurately extracts skin region pixels, and generates standardized skin color codes after processing. This skin color analysis method, compared to simple color sampling or RGB threshold segmentation in existing technologies, can more accurately quantify user skin color features and avoid interference from factors such as background lighting and clothing occlusion. Based on the accurate skin color code, the system can select clothing colors that match the user's skin tone from the clothing database, achieving personalized color recommendations and improving the aesthetics and harmony of the outfit. In the clothing and body shape matching process, this invention employs a multi-dimensional deformation fitting algorithm. Specifically, the clothing is scaled proportionally according to the user's height to match the clothing length; a mapping relationship is established between key points of the human skeleton and key points of the clothing for precise positional alignment; based on the human contour and body feature parameters, a deformation transformation algorithm is used to fit the clothing to the human contour shape; finally, the matched clothing layer is used to occlude and blend the human image. Compared to the simple image overlay method in existing technologies, the fitting images generated by this invention are more realistic and natural in terms of clothing size, style, position, and fit, significantly improving the user experience. This invention identifies the user through a facial recognition module and reads the user's skin color, height, weight, body shape, and clothing history data from the user data storage module, inputting this data into the clothing analysis module. This design allows the system to identify returning users and reuse their historical data, reducing the time spent on repeated collection and processing, improving the efficiency of clothing matching and the continuity of recommendation results. Simultaneously, the accumulation of user historical data provides a data foundation for the continuous optimization of personalized recommendation algorithms. This invention integrates a body shape recognition module, a skin color analysis module, a facial recognition module, and a clothing analysis module on an intelligent analysis server, forming a complete automated processing flow. The user only needs to stand in front of the fitting room mirror, and the system can automatically complete the entire process of image acquisition, body shape recognition, skin tone analysis, clothing selection, body shape matching, and fitting room image generation and display. No manual operation or parameter input is required from the user. The system responds quickly and is suitable for various application scenarios such as retail stores and smart homes. The accompanying drawings illustrate the invention in detail below. Figure 1 is a system architecture diagram of the invention. Figure 2 is the overall flow chart of the intelligent fitting room processing method. Figure 3 is the body shape recognition processing flow chart. Figure 4 is the clothing and body shape matching flow chart. Detailed Description of the Embodiments To more clearly illustrate the technical solution of the invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the following embodiments are only for illustrating the invention and do not constitute a limitation of the invention.HK 30135831 A 7 I. System overall structure, as shown in Figure 1. This embodiment provides an intelligent fitting system, the overall architecture of which is shown in Figure 1, including: Fitting mirror main body The fitting mirror main body is an integrated intelligent terminal device, integrating: High-definition camera: used to collect the user's full-body frontal image, face image and human posture image in real time; Display device: adopts a high-resolution touch screen, used to display the user image, the matching clothing effect, and the clothing selection interface in real time; Communication device: supports Wi-Fi, Ethernet or 5G communication, used to upload the user image data to the server and receive the clothing matching image returned by the server. Intelligent analysis server The server is a cloud or local high-performance computing device, which establishes a data connection with the fitting mirror main body through the communication device, including the following functional modules: (1) Body shape recognition module is used to perform image processing on the user image, extract the human body contour, calculate the actual height, detect the human body key points, calculate the body shape parameters such as shoulder width / waist circumference / hip circumference, and complete the body shape classification. (2) The skin color analysis module is used to extract skin color features from the user's face or exposed skin area and generate standardized skin color codes for clothing color matching. (3) The clothing database stores massive amounts of clothing data. Each garment includes: clothing image, clothing key points, pattern information, size parameters, main color code, body type label, style label, etc. (4) The core control module of the clothing analysis module is used to select matching clothing from the clothing database based on the body type recognition results and skin color analysis results, and to scale, align, deform, and fit the clothing to generate the final outfit fusion image. (5) The face recognition module is used to extract facial features from the user's image to complete user identity recognition and verification. (6) The data storage module is used to store the user's historical data, including: user skin color, height, weight, body shape parameters, historical fitting records, preferred clothing styles, etc., to realize personalized fitting services. HK 30135831 A 8 II. System Workflow The intelligent fitting room processing method of this embodiment, as shown in Figure 2, has the following overall flow: Step S1: Acquiring User Image The user stands in front of the fitting room mirror body. The camera automatically starts and continuously acquires full-body image data of the user's front, including the complete outline of the human body, the face area, shoulders, waist, hips, and the complete posture of the lower limbs. Step S2: Uploading Data The fitting room mirror body transmits real-time image data to the intelligent analysis server through a communication device. The server enters real-time processing mode.Step S3: Body Shape Recognition Processing. Referring to Figure 3, the server performs the following sub-steps through the body shape recognition module: Image Preprocessing: The input human image is normalized to a standard resolution; color space conversion is performed, converting the RGB image to LAB or YCbCr space to reduce the influence of lighting; Gaussian filtering is used to remove image noise and improve the accuracy of subsequent processing. Human Contour Extraction: A pre-trained human segmentation model (such as MediaPipe Selfie Segmentation or SAM Segmentation Model) is used to perform pixel-level semantic segmentation of the human image, generating a binary mask of the human region; the maximum connected region is extracted using the OpenCV contour detection algorithm to obtain the coordinates of the user's external human contour. Human Keypoint Detection: A pre-trained human pose model (such as MediaPipe Pose or OpenPose) is used to perform keypoint inference on the human image, outputting a set of human skeletal keypoint coordinates, including: head vertex, neck point, left and right shoulder points, left and right waist points, hip point, knee point, and foot point. HK 30135831 A 9 Calculate actual height: Based on the vertical pixel distance between the top of the head and the bottom of the feet, combined with the preset image ratio and physical size mapping relationship, the user's actual physical height (unit: cm) is calculated. Calculate body posture parameters and classify body types: Calculate shoulder width based on the distance between shoulder key points; calculate waist circumference based on waist key points and contour width; calculate hip circumference based on hip key points and contour width; further calculate waist-to-hip ratio and shoulder-to-waist ratio; classify body types into: pear, apple, hourglass, rectangular, and standard according to preset rules. Height calculation and body posture parameter calculation can be performed in parallel to improve processing efficiency. Step S4: Skin color analysis processing: The server performs the following processing through the skin color analysis module: Face and skin region segmentation: Load the face detection model, locate the user's face region, remove non-skin regions such as eyes, lips, and hair from the face region, and retain stable skin regions. The skin tone code generation process extracts the color values ​​of all valid pixels in the skin area, calculates the mean, and standardizes the values ​​to generate a unique skin tone code (such as a LAB value, hexadecimal color code, or Fitzpatrick skin tone type). Step S5: Clothing matching and adjustment, as shown in Figure 4. The clothing analysis module performs intelligent clothing matching based on body shape and skin tone results: Color filtering: Based on the user's skin tone code, clothing with coordinated colors and suitable skin tone is selected from the clothing database, excluding clothing with conflicting colors. HK 30135831 A 10 Body shape and size matching: Based on the user's height, shoulder width, waist circumference, hip circumference, and body shape type, clothing of corresponding size and suitable fit is selected from the clothing database to generate a candidate clothing list.The automatic adjustment and fitting of clothing involves loading images, key points, and size information of candidate clothing: Clothing is scaled proportionally to the user's height, ensuring the garment length, sleeve length, and pant length match the user's height; a one-to-one mapping relationship is established between the user's shoulder, waist, and hip key points and the clothing's standard key points; Thin-plate spline transformation (TPS) or perspective transformation algorithms are used to deform the clothing, ensuring the clothing outline perfectly matches the human body outline; layer fusion is performed based on the human body outline mask, naturally overlaying the clothing onto the user's human body image to generate a matching effect image. Step S6: The real-time display server transmits the fused clothing matching image back to the main body of the fitting mirror via a communication device. The display device shows the effect of the user wearing the adapted clothing in real time and smoothly, supporting rapid switching between multiple garments. III. Preferred Embodiment In a preferred embodiment: When a user uses the system for the first time, the system automatically extracts and stores the user's body contour, height, body shape, and skin tone. Upon subsequent use, the facial recognition module quickly identifies the user and directly reads historical features from the data storage module, eliminating the need for recalculation and significantly improving response speed. The clothing analysis module (HK 30135831 A 11) supports recommending preferred styles based on the user's fitting history and allows users to manually switch clothing styles, colors, and sizes, updating the display effect in real time. This embodiment automatically acquires the user's body shape and skin tone through non-contact visual recognition technology, enabling intelligent clothing selection and precise fit adjustment. Real-time outfit display is achieved on the fitting mirror terminal, featuring ease of use, accurate matching, and smooth interaction. It is suitable for various application scenarios such as clothing stores, online malls, and outfit recommendations. HK 30135831 A 1 Claims 1. An intelligent fitting room system, characterized in that: it includes a fitting room mirror body equipped with a camera, a display device, and a communication device; an intelligent analysis server, which connects to the fitting room mirror body via the communication device and acquires the content captured by the camera, including a skin tone analysis module for color analysis of the user's skin tone, a clothing database, and a clothing analysis module; the clothing analysis module selects suitable clothing from the clothing database based on the output results of the skin tone analysis module and transmits it to the fitting room mirror body via the communication device, and displays the matching image of the user and the suitable clothing in real time via the display device. 2. The intelligent fitting room system according to claim 1, characterized in that: the intelligent server further includes a body shape recognition module for recognizing the user's human body contour, height, and posture; the clothing analysis module selects suitable clothing from the clothing database and matches the body shape based on the output results of the body shape recognition module and the skin tone analysis module, and transmits it to the fitting room mirror body via the communication device, and displays the matching image of the user and the suitable clothing in real time via the display device.3. The intelligent fitting system according to claim 2, characterized in that: the intelligent analysis server further includes a module for facial recognition by extracting facial features from the content captured by the camera, and a module for storing user data. The facial recognition module identifies the user and reads the user's skin color, height, weight, body shape and clothing history from the data storage module, and inputs the above data into the clothing analysis module for clothing matching. 4. A method for intelligent fitting process using the intelligent fitting system of claim 3, characterized by comprising the following steps: Step S1: Acquiring user image data through the camera of the fitting mirror body; Step S2: Transmitting the acquired user image data to the intelligent analysis server via a communication device; Step S3: The intelligent analysis server performs body shape recognition processing on the user image data, extracting the user's human body contour, height, and posture information; Step S4: The intelligent analysis server performs skin color analysis processing on the user image data, generating the user's skin color code; Step S5: Based on the body shape recognition result and skin color analysis result, the intelligent analysis server selects suitable clothing from the clothing database and performs body shape matching; Step S6: Transmitting the matched clothing combination image to the fitting mirror body via a communication device, and displaying the combination image of the user and the suitable clothing in real time via a display device. 5. The intelligent fitting room processing method according to claim 4, characterized in that the body shape recognition processing in step S3 includes the following: Step S31: Preprocessing the input human body image, including size normalization, color space conversion, and noise filtering; Step S32: Performing pixel-level semantic segmentation using a pre-trained human body segmentation model to generate a human body region mask, and extracting the external contour of the human body based on the mask; Step S33: Inferring the human body image using a pre-trained human body key point detection model to obtain the coordinate set of human skeletal key points, including the top of the head, neck, shoulder, waist, hip, knee, and foot points; Step S34: Calculating the human height based on the vertical height of the human body contour and the pixel distance between the key points, combined with the mapping relationship between image proportion and physical size; Step S35: Calculating body shape parameters such as shoulder width, waist circumference, hip circumference, waist-to-hip ratio, and shoulder-to-waist ratio based on the key point spacing and the geometric features of the human body contour, and completing body shape classification according to preset rules; wherein steps S34 and S35 The execution steps are not executed in sequence.6. The intelligent fitting processing method according to claim 4, characterized in that the skin color analysis processing in step HK 30135831 A 3 S4 includes the following processing steps: Step S41: Load the face detection model and segment the face region to extract the skin region Step S42: Extract the color values ​​of all pixels in the skin region, and process them to obtain the skin color code. 7. The intelligent fitting method according to claim 5, characterized in that the clothing and body shape matching step in step S5 includes the following: Step S51: Select clothing that matches the skin color code from the clothing database according to the user's skin color code; Step S52: Obtain the image of the clothing to be matched and the clothing key points, clothing size and clothing pattern information; Step S53: Scale the clothing according to the overall proportion of the human body height so that the clothing length matches the human body height; establish a mapping relationship between the human body skeletal key points and the clothing key points to align the clothing position; according to the human body contour and body shape feature parameters, use a deformation transformation algorithm to fit the clothing to the human body contour shape, and use the matching clothing layer to occlude and merge the human body image to generate a matched clothing matching image; Step S54: Transmit the matched clothing matching image to the fitting mirror body through a communication device, and display the matching image of the user and the suitable clothing in real time through a display device. HK 30135831 A 1 Instruction Manual Attachments Figure 1 HK 30135831 A 2 Figure 2 HK 30135831 A 3 Figure 3 HK 30135831 A 4 Figure 4 HK 30135831 A.

Images