Guide control method and device for 3D fitting, program product and storage medium

Abstract

The invention discloses a guide control method and device for 3D fitting, a program product and a storage medium, and relates to the technical field of intelligent retail. The method comprises the steps of obtaining member archive data; extracting historical preferences and shopping intentions of the customers entering the store from the member file data, and generating guide path data according to the historical preferences and the shopping intentions; judging whether the customer entering the store arrives at the standing site or not; generating a 3D human body model according to the human body scanning data of the customer entering the store when the customer position data indicates that the customer entering the store arrives at the standing site; selecting a virtual clothing model according to the 3D human body model and the shopping intention, and performing registration rendering on the virtual clothing model and the 3D human body model to generate a virtual fitting image; generating size recommendation information according to the virtual fitting image and the human body scanning data; and sending the size recommendation information and the virtual fitting image to a terminal device of the customer entering the store so as to guide the customer entering the store to carry out shopping. By implementing the technical scheme provided by the invention, the shopping efficiency of customers in a shop can be improved.

Description

Technical Field

[0001] This application relates to the technical field of smart retail, specifically to a 3D virtual fitting guidance and control method, device, program product, and storage medium. Background Technology

[0002] With the deepening of digital transformation in the retail industry, offline brick-and-mortar stores are actively exploring intelligent service models to improve customer shopping experience and operational efficiency. As an important consumer sector, apparel retail's traditional shopping process typically relies on customer self-browsing, personalized recommendations from sales assistants, and in-store fitting.

[0003] To improve the shopping experience, some clothing stores have begun using 3D virtual fitting technology. This technology scans customers' body data to generate a three-dimensional human body model and matches virtual clothing with the model, allowing customers to preview the effect on a screen. This reduces the need for in-person fittings and increases shopping efficiency.

[0004] However, existing 3D virtual fitting technology primarily focuses on the visual presentation of the fitting process itself. Customers typically need to actively find and navigate to the fitting equipment, then manually select desired clothing styles from a large selection of items. This approach lacks a coherent and targeted guidance throughout the customer's experience, preventing the store from proactively providing precise shopping paths and product recommendations based on individual customer needs, resulting in low shopping efficiency. Summary of the Invention

[0005] This application provides a 3D virtual fitting guidance control method, device, program product, and storage medium, which can improve the shopping efficiency of customers in stores.

[0006] The first aspect of this application provides a 3D virtual fitting guidance and control method, specifically including: Obtain biometric data of customers entering the store, and obtain member profile data based on the biometric data; Extract the historical preferences and shopping intentions of customers entering the store from the member profile data, and generate guidance path data based on the historical preferences and shopping intentions; The system acquires the customer location data of the customers entering the store in real time, and compares the customer location data with the guidance path data to determine whether the customers have arrived at the designated station. When the customer location data indicates that a customer has arrived at the designated station, a 3D human body model is generated based on the customer's human body scan data. Based on the 3D human body model and the shopping intention, a virtual clothing model is selected, and the virtual clothing model is registered and rendered with the 3D human body model to generate a virtual fitting image. Generate size recommendation information based on the virtual fitting image and the human body scan data; The size recommendation information and the virtual fitting image are sent to the terminal device of the customer who enters the store to guide the customer in making a purchase.

[0007] By employing the aforementioned technical solutions, customers can be identified and their membership profiles retrieved via biometric data upon entering the store. Personalized guidance paths are then generated based on historical preferences and shopping intentions, enabling precise customer guidance. By tracking customer location data in real time and comparing it with the guidance path data, the system ensures customers accurately reach their designated fitting station, complete a body scan at the optimal location, and generate a 3D human model. Combining the customer's shopping intentions, the system intelligently selects suitable virtual clothing models and generates high-quality virtual fitting images. Simultaneously, it provides accurate size recommendations based on the virtual fitting effect and body scan data. This end-to-end intelligent service, from entry recognition to path guidance, virtual fitting, and size recommendation, allows customers to quickly find and experience the most suitable products, significantly improving shopping efficiency.

[0008] Optionally, generating guidance path data based on the historical preferences and the shopping intent includes: Detect whether there are other customers entering the store. If there are no other customers entering the store, extract the target clothing type and historical browsing path from the historical preferences. A shopping urgency score is generated based on the shopping intention, and a path planning strategy is determined based on the shopping urgency score. The path planning strategy includes a shortest path strategy and a preferred path strategy. Query the corresponding product area set based on the target clothing type; When the path planning strategy is the shortest path strategy, the product area closest to the entrance is selected from the product area set as the target product area; When the path planning strategy is the preferred path strategy, the historical browsing path is matched with the product area set, the product areas in the product area set are sorted according to the access frequency in the historical browsing path, and the product area with the highest access frequency is selected as the target product area. The station location is determined based on the target product area, and guidance path data from the entrance to the station location is generated.

[0009] By adopting the above technical solution, the guidance strategy can be intelligently adjusted based on the presence of other customers in the store. When there are no other customers, the system extracts the target clothing type and historical browsing path from the member's historical preferences, and generates a shopping urgency score based on shopping intent, thereby determining the most suitable path planning strategy. For customers with limited time, the system uses the shortest path strategy, directly guiding them to the relevant product area closest to the entrance; for customers with ample time, the system selects the product area that the customer is most interested in as the target area based on the frequency of visits along their historical browsing path. This dynamic path planning method based on scenarios and customer characteristics can not only meet the shopping needs of different customers, but also optimize customer movement within the store, thereby further improving shopping efficiency.

[0010] Optionally, the method further includes: When other customers enter the store, obtain the other guidance path data generated by those other customers; Extract the location data of occupied stations from the other guidance path data, and construct a station occupancy table; Based on the historical preferences and shopping intentions, the target product area of ​​the in-store customer is determined, and a set of candidate station locations within the target product area is identified; The candidate site location set is compared with the site location occupancy table. When it is detected that the target site location in the candidate site location set has been occupied, the membership level data of the customers who entered the store is extracted from the membership profile data, and the shopping urgency score of the customers who entered the store is extracted. The path priority weight of the customer entering the store is calculated based on the membership level data and the shopping urgency score, and the path priority weight is compared with the path priority weight of other customers entering the store who occupy the target station. When the path priority weight of the customer entering the store is greater than the path priority weight of other customers entering the store at the target station, the other customers occupying the target station are reassigned to select a station and the other guidance path data is updated. The target station is then assigned to the customer entering the store to generate the guidance path data. When the path priority weight of the customer entering the store is less than the path priority weight of other customers entering the store at the target station, an unoccupied station is selected from the candidate station set as a backup station, and guidance path data containing the backup station is generated.

[0011] By adopting the above technical solution, when other customers are present in the store, the system tracks the occupancy status of allocated service stations in real time and identifies available candidate service stations based on the target product area of ​​new customers. When the target service station is already occupied, the system calculates the path priority weight by comprehensively considering the customer's membership level and shopping urgency score. For high-priority customers, the system ensures they receive the best service experience by reallocating service stations from other customers; for low-priority customers, the system automatically allocates unoccupied alternative service stations. This dynamic service station allocation mechanism based on multi-dimensional evaluation not only guarantees the service quality of high-value customers but also ensures that all customers receive appropriate guidance services, thereby achieving efficient resource utilization and overall optimization of the service experience in multi-customer scenarios.

[0012] Optionally, comparing the customer location data with the guidance path data to determine whether the customer has arrived at the designated station includes: Extract station locations and path node sequences from the guidance path data; The customer location data is matched with the path node sequence to determine the path node where the customer is currently located. Based on the positional relationship between the customer's current path node and the station location, determine whether the customer is within a preset first range of the station location; When it is determined that a customer is within a preset first range of the location of the station, the collection frequency of the customer location data is increased according to a preset frequency adjustment rule; The customer's center position is obtained from the improved collection frequency. The customer's center position is compared with the station location data. When the center position falls within the preset second range of the station location, it is determined that the customer has arrived at the station location.

[0013] By adopting the above technical solution, the approximate location of the customer on the guided path is first determined through path node sequence matching. When the customer enters the first preset range of the station location, the system automatically increases the frequency of location data collection to achieve more precise location tracking. Subsequently, by comparing the customer's center position with the second preset range of the station location, it is ensured that the customer is indeed in the optimal scanning position. This progressive position determination method avoids the waste of system resources caused by continuous high-frequency collection and can achieve accurate positioning at key locations, thereby ensuring the accuracy and effectiveness of subsequent 3D scanning and virtual try-on.

[0014] Optionally, after registering and rendering the virtual clothing model with the 3D human body model to generate a virtual fitting image, the method further includes: Obtain customer feedback data on the virtual fitting images; When the evaluation feedback data indicates dissatisfaction, the adjustment requirement type is extracted from the evaluation feedback data; The set of adjustable parameters for the virtual clothing model is determined based on the type of adjustment requirement. Extract historical adjustment preferences corresponding to the adjustment request type from the member profile data; The parameters in the adjustable parameter set are assigned values ​​according to the historical adjustment preferences to generate the adjusted virtual clothing model; The adjusted virtual clothing model and the 3D human body model are re-registered and rendered to generate an updated virtual try-on image.

[0015] By adopting the above technical solution, the virtual try-on effect can be intelligently adjusted based on real-time customer feedback. When a customer is dissatisfied with the current virtual try-on effect, the system analyzes the customer's specific adjustment needs and, combined with their historical adjustment preferences, automatically makes targeted adjustments to the relevant parameters of the virtual clothing model. This intelligent parameter adjustment mechanism based on personalized historical data avoids the hassle of customers manually trying different parameters multiple times, and can quickly generate a virtual try-on effect that matches the customer's personal wearing habits, further improving the accuracy of virtual try-ons and the user experience.

[0016] Optionally, generating size recommendation information based on the virtual fitting image and the human body scan data includes: The human body scan data is divided into regions, and the size data of multiple preset body parts are extracted; Identify clothing sections corresponding to the multiple preset body parts from the virtual clothing model, and extract the section size data of each clothing section; Calculate the size deviation between the size data of each preset body part and the size data of the corresponding clothing section; Extract clothing functional requirements from the shopping intention, and set a fit threshold for each preset body part based on the clothing functional requirements; Each of the size deviation values ​​is compared with the corresponding preset body part fit threshold to generate part size suggestions for each preset body part, and the part size suggestions are summarized to generate size recommendation information.

[0017] By employing the aforementioned technical solution and refining the human body scan data into specific regions, the system can acquire detailed dimensional data for each key body part and accurately compare it with the dimensions of the corresponding sections in the virtual clothing model. Combining this with the functional needs expressed by customers in their shopping intentions, the system sets personalized fit standards for different body parts, thus taking into account the specific wearing requirements of different areas when calculating size deviations. This recommendation mechanism, which integrates body data, clothing dimensions, and functional needs, provides customers with more accurate and personalized size suggestions, effectively reducing returns and exchanges due to inappropriate size selection.

[0018] Optionally, after obtaining the member profile data based on the biometric data, the method further includes: The system detects whether a customer is carrying items. When items are detected, the system obtains the location data of the item placement area and generates an item storage guidance path from the current location to the item placement area. Based on the item storage guidance path, an available locker is identified from the set of smart lockers in the item placement area, and the customer's target locker is determined from the available lockers based on the estimated time of arrival at the item placement area. Obtain the customer's facial feature data or terminal device identifier, generate a dynamic unlock code bound to the target locker, and send the dynamic unlock code to the customer's terminal device; When the door of the target locker is opened and the weight sensor inside the target locker detects a change in weight, it is determined that the item has been stored and the target locker is controlled to output a confirmation message. The occupancy status of the smart locker set is updated based on the confirmation information, and the target locker is associated with and stored in relation to the customer's membership profile data.

[0019] By adopting the above technical solution, when the system detects items carried by a customer, it immediately generates a guidance path to the item placement area and reserves suitable locker resources in advance based on the customer's expected arrival time. The system ensures the security of locker use by generating dynamic unlock codes through facial features or terminal device identification. Simultaneously, the system monitors the storage status of items in real time using weight sensors and automatically updates the locker status and member association information after confirmation of storage, providing customers with a convenient temporary storage service, allowing them to shop light and improving the overall shopping experience.

[0020] In a second aspect, this application provides a guidance and control device for 3D virtual fitting, the guidance and control device for 3D virtual fitting comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors calling the computer instructions to cause the guidance and control device for 3D virtual fitting to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, this application provides a computer program product containing instructions that, when run on a 3D fitting room guidance and control device, cause the 3D fitting room guidance and control device to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, this application provides a computer-readable storage medium including instructions that, when executed on a 3D virtual fitting guide control device, cause the 3D virtual fitting guide control device to perform the method described in the first aspect and any possible implementation thereof. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a 3D virtual fitting guidance and control method provided in an embodiment of this application; Figure 2 This is a schematic diagram illustrating a recommended size provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating a scenario where other customers enter the store, as provided in an embodiment of this application. Figure 4 This is an exemplary hardware structure diagram of a 3D virtual fitting guidance and control device provided in an embodiment of this application. Detailed Implementation

[0024] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0025] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0026] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0027] This application provides a guidance and control method for 3D virtual try-on, referencing... Figure 1 , Figure 1 This is a flowchart illustrating a 3D virtual try-on guidance and control method provided in an embodiment of this application, including steps S101 to S107, as follows: S101: Obtain biometric data of customers entering the store, and obtain member profile data based on the biometric data.

[0028] Biometric data refers to physiological characteristics that uniquely identify a customer after authorization, including facial feature vectors, iris features, and fingerprint features. Facial feature vectors represent the numerical expression of the coordinates and geometric relationships of key facial points acquired through image acquisition devices, such as the positional coordinates and relative distances of facial features like the eyes, nose, and mouth. Member profile data represents comprehensive information records associated with a specific member, including basic member information, historical shopping records, body type data, and preference settings.

[0029] Specifically, after a customer enters the store, biometric data collection devices deployed in the entrance area identify the customer. The devices acquire raw biometric data from sensors, perform preprocessing operations on the raw data, including noise reduction, normalization, and feature enhancement, and then extract feature vectors. These extracted feature vectors are then matched against feature templates stored in the member database. A similarity score is calculated between the feature vectors to determine if the customer is a registered member. The similarity score is calculated using metrics such as Euclidean distance or cosine similarity. When the similarity score exceeds a preset threshold, the member's identity is confirmed, and the corresponding member profile data is retrieved from the database.

[0030] In some embodiments, biometric data collection and member identification can be achieved through various methods. Optionally, facial recognition can be used, where a high-definition camera installed at the entrance captures customer facial images, performs face detection to locate the facial region, extracts key facial feature points to construct a 128-dimensional or 512-dimensional feature vector, and performs batch comparison and matching with facial feature templates in the member database. Member profile data is then returned based on the matching results. Optionally, a multimodal biometric fusion method can be used, simultaneously collecting customer facial and gait features, extracting feature vectors from both features separately, and then performing weighted fusion. The fused comprehensive feature vector is matched with the member database, and member identity is determined and profile data is obtained based on the matching results of the fused features. It is understood that other biometric identification methods can also be used to confirm member identity; this is not limited here.

[0031] S102: Extract the historical preferences and shopping intentions of customers entering the store from the member profile data, and generate guidance path data based on the historical preferences and shopping intentions.

[0032] Among them, historical preferences refer to the consumption tendencies summarized from members' past shopping behavior, including preferred clothing types, colors, styles, price ranges, brands, and historical browsing paths. Shopping intent indicates the customer's purpose for visiting the store this time, including information such as target product type, shopping time budget, and urgency of purchase. Guidance path data represents the spatial movement trajectory from the store entrance to the target station, including the sequence of path nodes, the spatial coordinates of each node, the direction of movement, and the location of the station.

[0033] Specifically, the system reads historical purchase and browsing records from member profiles and performs data mining to extract historical preference features. Historical preference extraction is achieved through statistical analysis, calculating the purchase frequency and browsing duration for each clothing type, and marking the clothing type with the highest frequency as the target clothing type. Shopping intent is inferred by analyzing data such as customer entry time, expected dwell time, and recent member search records. Based on the target clothing type, the system queries the store's product layout database to determine the corresponding target product area location. Combining the time budget information in the shopping intent, a shopping urgency score is calculated. Based on the urgency score, a path planning strategy is selected, and a path planning algorithm is used to calculate the optimal movement path from the entrance to the target product area. Suitable 3D fitting station locations are selected near the target product area, generating guided path data containing path node sequences and station coordinates.

[0034] In some embodiments, the generation of guided paths can be implemented in multiple ways. Optionally, a historical trajectory reproduction-based approach can be used. This involves extracting product areas and movement paths that customers have visited multiple times in past shopping experiences from member profiles, calculating the access frequency of each product area in the historical path, selecting the product area with the highest access frequency as the target area, and generating the current guided path data using the movement trajectory in the historical path. Optionally, a real-time dynamic planning approach can be used. This involves obtaining real-time environmental data based on the current customer distribution and congestion levels in each area, determining the target product area based on customers' historical preferences, using a dynamic path planning algorithm to calculate the optimal path to avoid congested areas, selecting available station locations within the target area based on the real-time occupancy status of station locations, and generating guided path data that includes dynamically adjusted nodes. It is understood that other path planning methods can also be used to generate guided paths, and this is not limited here.

[0035] S103: Real-time acquisition of customer location data of customers entering the store, and comparison of customer location data with guidance path data to determine whether customers have reached the station location.

[0036] Customer location data refers to the real-time spatial location information of customers within the store, including parameters such as planar coordinates, movement speed, and orientation angle. Planar coordinates represent the customer's two-dimensional position in the store's coordinate system, typically established with the store entrance as the origin. For example, a customer located 5 meters horizontally and 3 meters vertically from the entrance can be represented by coordinates (5, 3). The station point represents the fixed location area where the pre-set 3D human body scanning equipment is located. This area has clearly defined spatial boundaries and center coordinates; customers must stand within the station point area to complete an accurate human body scan.

[0037] Specifically, customer location information is continuously collected through positioning devices deployed within the store. These devices acquire customers' real-time coordinates at preset time intervals, forming a continuous stream of location data. The system extracts the path node sequence and the center coordinates of the station locations from the guidance path data. It matches the real-time customer location coordinates with the path nodes, calculating the spatial distance between the customer's current location and each path node to determine the customer's location within that node. The distance calculation uses the Euclidean distance formula; when the distance between the customer's location and a path node is less than a preset node range threshold, the customer is determined to be at that node. The system further calculates the distance between the customer's current location and the center coordinates of the station location. When this distance is less than a first preset range threshold, the customer is determined to be approaching the station location, at which point the location data collection frequency is increased. In this high-frequency collection mode, the customer's precise center location is acquired, and the distance between this center location and the center of the station location is calculated. When this distance is less than a second preset range threshold, the customer is determined to have arrived at the station location.

[0038] In some embodiments, customer location acquisition and arrival determination can be achieved in various ways. Optionally, Bluetooth beacon positioning can be used, deploying multiple Bluetooth beacon devices within the store. The customer's terminal device receives the signal strength of each beacon, calculates the distance from the terminal device to each beacon based on a signal strength attenuation model, calculates the customer's planar coordinates using a triangulation algorithm, and compares the calculated coordinates with the station coordinates to determine arrival. Optionally, visual tracking positioning can be used, capturing customer images through multiple cameras within the store, identifying the customer's position within the camera's field of view using a target detection algorithm, converting the image coordinates into store planar coordinates based on the camera's installation position and viewing angle parameters, continuously tracking the customer's movement trajectory, and determining arrival when the customer's planar coordinates fall within a preset range of the station location. It is understood that other positioning technologies can also be used to achieve customer location tracking and arrival determination; this is not limited here.

[0039] S104: When customer location data indicates that a customer has arrived at the station, a 3D human body model is generated based on the human body scan data of the customer.

[0040] Human body scan data refers to the spatial geometric information of a customer's body surface collected by scanning equipment, including depth data, point cloud data, and mesh data for various body parts. Depth data represents the distance of each point on the body surface from the scanning equipment, while point cloud data represents a set of a large number of discrete spatial points on the body surface, each containing three-dimensional coordinate information. The 3D human body model represents a three-dimensional digital representation of the customer's body generated through computational processing. This model describes the shape and size of the body using a mesh structure, including vertex coordinates, facet topological relationships, and dimensional parameters of various body parts.

[0041] Specifically, once a customer is identified as having arrived at their designated station, the system activates the 3D scanning device deployed at that station. The scanning device issues a scan preparation prompt to the customer, guiding them to adjust their standing posture. The depth sensor of the scanning device scans the customer's body from multiple angles, acquiring a sequence of depth images of the body surface. Coordinate transformation is performed on the depth images, converting the two-dimensional depth images into three-dimensional point cloud data. The three-dimensional coordinates of each point in the point cloud are calculated using the depth value and the camera intrinsic parameter matrix. The acquired multi-viewpoint point cloud data is registered, unifying the point clouds from different perspectives into the same coordinate system. The registration process uses an iterative nearest-point algorithm to calculate the rotation and translation transformation parameters between the point clouds. The registered point clouds are then fused and denoised to remove outliers and duplicate points. Mesh reconstruction is performed based on the fused point cloud, using a Poisson surface reconstruction algorithm or a moving cube algorithm to generate a triangular mesh model. The mesh model is then smoothed and topologically optimized to obtain a complete 3D human body model.

[0042] In some embodiments, 3D human body models can be generated in various ways. Optionally, a depth camera scanning method is used, employing a depth camera equipped with structured light or Time-of-Flight (TOF) technology to acquire depth images of the customer from the front, side, and back. These three depth images are converted into point cloud data, and the three sets of point clouds are registered and fused. Based on the fused point cloud, a surface reconstruction algorithm is used to generate a complete 3D mesh model, which is then refined to obtain the final human body model. Optionally, a parametric model fitting method is used, employing a single depth camera to acquire a frontal depth image of the customer, extracting key body size parameters, selecting a base template matching the customer's gender and body type from a pre-set parametric human body template library, deforming and adjusting the template according to the extracted size parameters, and fitting the template to the actual scan data by adjusting the template's shape parameters to obtain a personalized 3D human body model. It is understood that other 3D reconstruction techniques can also be used to generate human body models; this is not limited here.

[0043] S105: Select a virtual clothing model based on the 3D human body model and shopping intention, and register and render the virtual clothing model with the 3D human body model to generate a virtual fitting image.

[0044] A virtual clothing model is a three-dimensional digital representation of clothing, including its geometric shape, material properties, and texture mapping. The geometric shape represents the clothing's three-dimensional mesh structure, material properties represent the physical properties of the fabric, and texture mapping represents the color and pattern information of the clothing surface. Registration and rendering refers to the process of spatially aligning and visually presenting the virtual clothing model with a human body model, including adjusting the clothing model's position, adapting deformations, and calculating lighting and material rendering. Virtual fitting images represent the two-dimensional or three-dimensional visualization results generated after registration and rendering, showcasing the visual effect of a customer wearing the clothing.

[0045] Specifically, the system filters candidate clothing models from a clothing model database based on the target product type in the shopping intent. The filtering process is achieved by matching product type tags; for example, if the shopping intent is to buy a dress, all virtual models of dresses in the database are filtered out. Further, the system evaluates the suitability of the candidate models based on the size parameters of the 3D human body model, calculating the matching degree between the size of each candidate clothing model and the human body model size, and selecting the clothing model with the highest matching degree as the target virtual clothing model. The selected virtual clothing model is then registered with the 3D human body model. The registration process first extracts the skeletal keypoints of the human body model, aligns the corresponding keypoints of the clothing model with the human body keypoints, and calculates the spatial transformation parameters required for the clothing model based on the positional relationship of the keypoints. The clothing mesh is then deformed according to the transformation parameters to make the clothing model fit the surface shape of the human body model. The deformation calculation uses a skinning algorithm, applying weighted deformation to the vertices of the clothing mesh based on the transformation of the skeletal keypoints. After registration, rendering calculations are performed, setting virtual light sources and camera views, calculating the lighting effects at various points on the clothing surface, and generating the final virtual try-on image based on material properties and texture mapping.

[0046] In some embodiments, the registration and rendering of virtual clothing and human body models can be achieved in various ways. Optionally, a skeletal-driven deformation method can be used to establish skeletal rigging for the 3D human body model, annotate key skeletal nodes such as shoulders, elbows, and waists on the human body model, establish corresponding skeletal structures for the virtual clothing model, associate and bind the clothing skeleton with the human skeleton, drive the human skeleton to drive the clothing model to deform synchronously, and perform ray tracing rendering based on the deformed model to generate a virtual try-on image. Optionally, a physical simulation method can be used, setting the virtual clothing model as a flexible fabric material, assigning physical property parameters to the fabric, setting the human body model as a collider, running a physical simulation engine to calculate the natural drooping of the fabric under gravity and its collision contact with the human body, iteratively calculating to stabilize the fabric, and then rendering the stabilized clothing model in real time to generate a virtual try-on image. It is understood that other registration and rendering techniques can also be used to generate virtual try-on effects, and this is not limited here.

[0047] S106: Generate size recommendation information based on virtual fitting images and human body scan data.

[0048] Size recommendations refer to clothing size suggestions generated based on a comparative analysis of human body dimensions and clothing dimensions. This includes the recommended size, fit assessments for various body parts, and explanations of the reasons for choosing the size. The fit assessment indicates the tightness or looseness of the clothing at specific body parts, described numerically or in grades, such as a loose, moderate, or tight fit at the chest. The reasons for choosing the size explain the basis for recommending that size, including the body size parameters and functional requirements of the clothing used to make the judgment.

[0049] Specifically, the system extracts precise body size parameters from human body scan data, including key dimensions such as chest circumference, waist circumference, hip circumference, shoulder width, and sleeve length. It segments the human body scan data into predefined regions such as shoulders, chest, waist, and hips, calculating the dimensions for each region. It extracts clothing size data for the corresponding regions from a virtual clothing model, obtaining clothing dimensions by measuring the perimeter or length of the corresponding parts in the clothing mesh model. It calculates the deviation between the dimensions of each body part and the corresponding clothing part, with the deviation equal to the clothing size minus the body size. It identifies the functional requirements of the clothing from the purchase intent; for example, when buying sportswear, the functional requirement is loose and comfortable, while when buying formal wear, the functional requirement is a fitted and tailored fit. Based on these functional requirements, it sets a fit threshold range for each body part, representing a reasonable range for dimensional deviation. The calculated dimensional deviation values ​​for each part are compared with the corresponding fit thresholds. When the deviation value falls within the threshold range, the fit is considered good; when the deviation value exceeds the threshold, adjustment suggestions are generated. Based on the combined fit assessment results of each part, the overall recommended clothing size is determined, and size recommendation information including size, fit assessment and reasons for selection is generated.

[0050] In some embodiments, size recommendation information can be generated in multiple ways. Optionally, a multi-size comparison method can be used. Virtual models of multiple adjacent sizes of the same garment are obtained from a clothing database. The clothing models for sizes S, M, and L are registered with the human body model. The size deviations of each size garment from the human body at key locations such as chest, waist, and hip circumference are calculated. Based on the fit preference in the shopping intention, the deviation values ​​of each size are scored, and the size with the highest score is selected as the recommended size, generating recommendation information containing the multi-size comparison results. Optionally, a historical data-assisted method can be used. Customer's historical purchase and return records are extracted from member profiles. The size distribution of past purchases and satisfaction feedback for each size are statistically analyzed. The current body scan data is compared with the body shape data at the time of historical purchase to calculate the change in body shape. The current size recommendation is adjusted based on historical preferred sizes and the change in body shape, generating size recommendation information that incorporates historical preferences. It is understood that other analytical methods can also be used to achieve size recommendations; this is not limited here.

[0051] like Figure 2 As shown, Figure 2This is a schematic diagram illustrating a size recommendation provided in an embodiment of this application. The left side of the diagram shows a human body model generated through 3D scanning. After dividing the human body scan data into regions, the system extracts the size data of preset body parts such as the shoulder, chest, and waist. The shoulder width is 45 cm, the chest circumference is 90 cm, and the waist circumference is 75 cm. The middle part of the diagram shows a virtual clothing model matching the shopping intention. This clothing is size M. The system identifies the clothing sections corresponding to different parts of the human body from the virtual clothing model and extracts the section size data for each section. The shoulder section size is 46 cm, the chest section size is 95 cm, and the waist section size is 78 cm. The right side of the diagram illustrates the size recommendation calculation process. The system first extracts the clothing functional requirement from the shopping intention as a formal, fitted style. Based on this functional requirement, it sets corresponding fit thresholds for each body part. For example, the shoulder fit threshold is set to 0.5 to 1.5 cm, the chest fit threshold is set to 2 to 4 cm, and the waist fit threshold is set to 2 to 4 cm. The system calculates the size deviation value for each preset body part. A deviation of ±1 cm for the shoulder falls within the fit threshold and is considered a good fit. A deviation of ±3 cm for the waist also falls within the threshold and is considered a good fit. However, a deviation of ±5 cm for the chest exceeds the upper limit of the threshold and is considered too large, generating a size suggestion to reduce the size by one size. After summarizing the size suggestions for each preset body part, the system uses a majority vote to determine the overall recommended size as size S. This generates complete size recommendation information that includes the recommended size, fit assessment for each body part, and the reasons for selection, ensuring that customers can choose the most suitable clothing size based on their body shape and wearing needs.

[0052] S107: Send size recommendations and virtual fitting images to the terminal devices of customers entering the store to guide them in making purchases.

[0053] Terminal devices refer to electronic devices carried by customers that can receive and display information, including smartphones, tablets, smartwatches, and other devices with network communication capabilities and display screens. These terminal devices establish a communication connection with the store's system via a wireless network and can receive shopping guidance information pushed by the system in real time. The information sent includes image files of virtual try-on images, text data of size recommendations, and relevant product details and purchase links.

[0054] Specifically, the system retrieves the customer's bound terminal device identifier from the member profile data. This identifier includes information such as the device's network address, application identifier, or communication account. Based on the device identifier, the system determines the target terminal and communication method for information transmission. The generated virtual fitting images are converted and compressed to a format suitable for mobile device display. Size recommendation information is organized into structured data, including recommended size, fit measurements for various body parts, and garment number. The data packet to be sent is encapsulated according to the communication protocol and transmitted to the customer's terminal device via the store's wireless or mobile network. Upon receiving the data, the shopping application installed on the terminal device parses the data content and displays the virtual fitting image and size recommendation information in the application interface. The interface also displays detailed information about the garment, its inventory status, and its shelf location, providing options to add it to the cart or purchase directly. Customers make purchasing decisions by viewing the recommendations and virtual fitting effects, and then proceed to the corresponding shelf to obtain the physical product based on the interface prompts.

[0055] In some embodiments, information can be sent to terminal devices in multiple ways. Optionally, a mobile application push method can be used. The system calls a message push service interface to construct a push message containing virtual fitting images and size recommendations, sends the message to the push service platform, and the push platform pushes the message to the shopping application installed on the customer's mobile phone based on the device identifier. After receiving the push, the application displays a message notification, and the customer clicks the notification to open the application to view detailed fitting images and recommendations. Optionally, a QR code scanning method can be used. The system uploads the virtual fitting images and size recommendations to a cloud server and generates a unique access link. A QR code image is generated based on the access link and displayed on the display screen at the station. The customer scans the QR code with their mobile phone, the mobile browser opens the access link, and the fitting images and recommendations are loaded from the cloud. The customer then views the complete shopping guide on their mobile phone. It is understood that other communication methods can also be used to transmit information, and this is not limited here.

[0056] Based on the above embodiments, as an optional embodiment, S102: the step of generating guidance path data based on historical preferences and shopping intentions may specifically include the following steps: S201: Detect whether there are other customers entering the store. If there are no other customers entering the store, extract the target clothing type and historical browsing path from historical preferences.

[0057] Other customers refer to customers who are currently active in the store, excluding the currently identified customer. Historical browsing path represents the actual movement trajectory of a customer within the store during their historical shopping process, recording the sequence of product areas visited by the customer. Target clothing type refers to the clothing categories that the customer tends to purchase, identified from historical preferences, such as dresses, shirts, and coats.

[0058] The system uses in-store personnel detection equipment to obtain the current number of customers and determines if the number is greater than one. If the number of customers is one, it is determined that no other customers have entered the store. Under this condition, historical purchase records are read from the member profile's historical preference data to count the number of purchases for each clothing type, and the clothing type with the most purchases is extracted as the target clothing type. Simultaneously, the system reads historical shopping movement records from the historical preferences to extract the product area sequences visited in each purchase, and the set of these area sequences is used as the historical browsing path.

[0059] S202: Generate a shopping urgency score based on shopping intention, and determine a route planning strategy based on the shopping urgency score. The route planning strategy includes the shortest path strategy and the preferred path strategy.

[0060] Shopping urgency score indicates how urgently a customer needs to complete their shopping, quantified numerically with a score range of 0 to 100. Shortest path strategy refers to selecting the path with the shortest spatial distance from the entrance to the target area. Preference path strategy refers to selecting familiar or preferred paths based on the customer's historical browsing habits.

[0061] The system extracts customers' expected dwell time and entry time from their shopping intentions to calculate a shopping urgency score. The calculation method is as follows: the time difference between the current time and the customer's expected departure time is taken as the available time. This available time is compared with a preset standard shopping time. When the available time is less than 50% of the standard shopping time, the urgency score is calculated as 100 minus the ratio of available time to standard time multiplied by 100. An urgency threshold of 60 is set. When the score is higher than 60, the path planning strategy is set to the shortest path strategy, guiding customers to the target area quickly; when the score is lower than or equal to 60, the path planning strategy is set to the preferred path strategy, allowing customers to browse along familiar routes.

[0062] S203: Query the corresponding product area set based on the target clothing type; when the path planning strategy is the shortest path strategy, select the product area closest to the entrance from the product area set as the target product area; when the path planning strategy is the preferred path strategy, match the historical browsing path with the product area set, sort the product areas in the product area set according to the access frequency in the historical browsing path, and select the product area with the highest access frequency as the target product area.

[0063] A product area set refers to the collection of all areas within a store that display a specific type of clothing. Visit frequency indicates the cumulative number of times a particular product area has been visited throughout the browsing history. A target product area refers to the specific product area selected based on a strategy to guide customers to.

[0064] The system queries the store layout database to obtain the product area corresponding to the target clothing type, thus obtaining a set of product areas. It then determines the current path planning strategy: If the shortest path strategy is used, it calculates the Euclidean distance between the center coordinates of each area in the product area set and the entrance coordinates. The distance is calculated as the square root of the sum of the squares of the differences between the coordinate components. The area with the smallest distance is selected as the target product area. If the preferred path strategy is used, it iterates through the recorded access sequences in the historical browsing path, counts the number of times each area in the product area set appears in the historical path as the access frequency, sorts the product area set in descending order of access frequency, and selects the area with the highest frequency as the target product area.

[0065] S204: Determine the station location based on the target product area and generate guidance path data from the entrance to the station location.

[0066] A station location refers to a fixed location equipped with 3D scanning equipment, pre-designated near the target product area, with clearly defined spatial coordinates. The guidance path data includes a sequence of path nodes from the entrance to the station location, with each node possessing spatial coordinates and direction of movement information.

[0067] The system queries the store equipment deployment database to retrieve all station locations within a preset distance range of the target product area, obtaining the coordinates and current occupancy status of each station. It then filters out currently unoccupied available station locations and selects the available station closest to the center of the target product area as the target station location. A path planning algorithm is used to calculate the movement path from the entrance to the target station location: a grid map of the store is created, dividing the store floor into grid cells. Grid cells occupied by shelves and obstacles are marked as impassable. The optimal path from the entrance grid cell to the station location grid cell is calculated using either A* or Dijkstra's algorithm. The path consists of a series of continuous traversable grid cells. The grid sequence is converted into a path node sequence, and the coordinates of each node and the movement direction between adjacent nodes are extracted to generate complete guidance path data.

[0068] Based on the above embodiments, as an optional embodiment, S102: the step of generating guidance path data based on historical preferences and shopping intentions also includes the case where there are other customers entering the store, which may specifically include the following steps: S301: When there are other customers entering the store, obtain the other guidance path data generated by the other customers entering the store; extract the occupied station location data from the other guidance path data, and construct a station location occupancy table.

[0069] Other guided path data refers to the path planning information already generated for other customers in the store, including each customer's target station location and path node sequence. The station location occupancy table represents a data structure that records the current occupancy status of each station location, including fields such as station location number, location coordinates, occupancy status, and occupant customer identifier.

[0070] When the number of customers in the store is greater than 1, the system queries the path management database for all active guided paths within the current time period to obtain guided path data for other customers entering the store. It iterates through each guided path, extracting the target station information for each path, reading the station number and location coordinates. A data structure for a station occupancy table is created, and each field is initialized. The extracted station information is written into the occupancy table one by one: the station number field is filled with the corresponding number, the location coordinate field is filled with the 3D coordinate value, the occupancy status field is marked as occupied, and the customer membership number corresponding to the station is recorded in the occupying customer identifier field. After traversing all guided path data, a station occupancy table containing information on all occupied station locations is generated.

[0071] S302: Determine the target product area for customers entering the store based on historical preferences and shopping intentions, and identify the candidate station set within the target product area; compare the candidate station set with the station occupancy table; when it is detected that the target station in the candidate station set has been occupied, extract the membership level data of the customers entering the store from the membership profile data, and extract the shopping urgency score of the customers entering the store.

[0072] The candidate site set refers to the collection of all available sites near the target product area that can be used for human body scanning. Membership level data represents the customer's level classification within the membership system, such as regular member, silver member, gold member, diamond member, etc., with different levels corresponding to different numerical weights.

[0073] Extracting target clothing types from historical preferences and combining them with product needs from shopping intent, the corresponding target product area is queried in the store layout database. All station locations within a preset range of the target product area are retrieved from the station location deployment database, and their numbers and coordinates are combined to form a candidate station location set. Each station in the candidate station location set is traversed, and its number is extracted. The occupancy status of that number is then checked in the station location occupancy table. Based on the distance between the candidate station location and the center of the target product area, the closest station location is selected as the target station location. The occupancy status field of the target station location in the occupancy table is then checked: if the status is "occupied," the membership level field of the currently entering customer is read from the member profile data to obtain the membership level data, and the calculated shopping urgency score is read from the shopping intent.

[0074] S303: Calculate the path priority weight of customers entering the store based on membership level data and shopping urgency score, and compare the path priority weight with the path priority weight of other customers entering the store occupying the target station location.

[0075] Path priority weight refers to the priority value calculated by combining membership level and shopping urgency. It is used to determine the priority order of customers in the allocation of site resources. The larger the value, the higher the priority.

[0076] Query the membership level weight configuration table based on membership level data to obtain the base weight value for the corresponding level. For example, the base weight for a regular member is 1, for a silver member it is 1.2, for a gold member it is 1.5, and for a diamond member it is 2. Normalize the shopping urgency score, mapping scores from 0 to 100 to the range of 0 to 1. The normalized value is equal to the score divided by 100. Calculate the path priority weight using a weighted summation method. The weight calculation formula is: path priority weight equals the membership level base weight multiplied by 0.6 plus the shopping urgency normalized value multiplied by 0.4. Query the membership ID of other customers occupying the target location from the location occupancy table. Based on this ID, retrieve their membership level and shopping urgency score from their member profile, and calculate their path priority weight using the same formula. Compare the values ​​of the two path priority weights and record the comparison results.

[0077] S304: When the path priority weight of a customer entering the store is greater than the path priority weight of other customers entering the store at the target station, reallocate alternative station locations to other customers occupying the target station location and update other guidance path data, and assign the target station location to the customer entering the store to generate guidance path data; when the path priority weight of a customer entering the store is less than the path priority weight of other customers entering the store at the target station location, select unoccupied station locations from the candidate station location set as alternative station locations, and generate guidance path data containing alternative station locations.

[0078] Alternate site locations refer to alternative site locations selected from the set of candidate site locations when the target site location is unavailable.

[0079] The comparison results of path priority weights are as follows: When the weight value of the customer entering the store is greater than the weight value of the customer occupying the store, other stations besides the target station are selected from the candidate station set. These stations are compared with the occupancy table, and the station that is not occupied and is closest to the target product area of ​​the occupied customer is selected as the new candidate station for the occupied customer. The path planning algorithm is re-executed to calculate the path from the current location of the occupied customer to the new candidate station. The guidance path data for this customer in the path management database is updated. In the occupancy table, the occupied customer identifier of the target station is modified to the membership number of the current customer entering the store. Guidance path data from the entrance to the target station is planned for the current customer entering the store. When the weight value of the customer entering the store is less than or equal to the weight value of the occupied customer, the candidate station set is traversed, and the status of each station in the occupancy table is queried. Stations with an unoccupied status are selected. The station closest to the target product area is selected from the unoccupied stations as the candidate station. Guidance path data from the entrance to the candidate station is planned. The occupancy record of the candidate station is added to the occupancy table.

[0080] like Figure 3 As shown, Figure 3 This is a schematic diagram illustrating a scenario where other customers are present in the store, as provided in an embodiment of this application. Multiple stations are deployed within the target product area (dress area), including station S1, station S2, and station S3. S1 is the target station closest to the center of the target product area, while S2 and S3 are alternative stations. When customer A has already occupied target station S1, a newly arrived customer B also needs to go to the target product area for virtual try-on. At this time, the system detects that target station S1 in the candidate station set is already occupied. Therefore, it extracts membership level data and shopping urgency scores from the membership profiles of customers A and B, respectively, calculating customer A's path priority weight as 1.2 and customer B's path priority weight as 1.8. Since customer B's priority weight is greater than customer A's, the system reallocates alternative station S2 to customer A who has occupied S1. Simultaneously, it updates customer A's guidance path data, adjusting the endpoint of their guidance path from S1 to S2, and assigns target station S1 to the higher-priority customer B, generating guidance path data for customer B from their current location to S1. This priority-based dynamic station allocation mechanism ensures that high-value members or customers with more urgent shopping needs can obtain the best service resources. At the same time, by adjusting the station allocation of other customers in real time, it achieves efficient utilization of store resources and optimization of the overall service experience in multi-customer scenarios.

[0081] Based on the above embodiments, as an optional embodiment, S301: the step of comparing customer location data with guidance path data to determine whether the customer has arrived at the station point may specifically include the following steps: S401: Extract station location and path node sequence from the guidance path data; match customer location data with path node sequence to determine the path node where the customer is currently located.

[0082] The location of a station refers to its three-dimensional coordinates in the store's coordinate system. The path node sequence represents an ordered set of discrete nodes arranged in the order of movement along a guide path; each node contains coordinates and a sequence number. Customer location data refers to the real-time coordinates of the customer's current position in the store's coordinate system, collected by the positioning device. Matching refers to the process of determining the corresponding path node for the customer by calculating the spatial distance between the customer's location and each path node.

[0083] The station location field is read from the data structure of the guidance path data, and the coordinate values ​​of the station locations are extracted. The path node sequence field is read to obtain an ordered list containing the coordinates and sequence numbers of all nodes. The real-time location coordinates of the customer are obtained from the positioning device. Each node in the path node sequence is traversed, and the Euclidean distance between the customer's location coordinates and the coordinates of each node is calculated. The distance calculation formula is the square root of the sum of the squares of the differences of the coordinate axis components. The calculated distance value is compared with a preset node matching threshold, and the node with the smallest distance value that is less than the matching threshold is selected as the path node where the customer is currently located.

[0084] S402: Based on the positional relationship between the customer's current path node and the station location, determine whether the customer is within the preset first range of the station location; when it is determined that the customer is within the preset first range of the station location, increase the collection frequency of customer location data according to the preset frequency adjustment rules.

[0085] The preset first range refers to a spherical spatial region centered on the station location with a radius equal to a first threshold distance, used to detect the customer's approach to the station location. The frequency adjustment rule defines the rules for adjusting the acquisition frequency, such as increasing it from the initial frequency to 3 or 5 times the initial frequency. The acquisition frequency refers to the number of times the positioning device acquires customer location data per unit time, measured in Hertz (Hz).

[0086] The system reads the coordinates of the customer's current path node and station location, and calculates the Euclidean distance between the two coordinates. It reads a preset threshold distance from the configuration parameters, for example, a threshold distance set to 2 meters. It then determines if the calculated distance is less than the first-range threshold distance: if the distance is less than the threshold, the customer is confirmed to be within the preset first range of the station location. The system reads the frequency boost factor set in the frequency adjustment rules, for example, a rule set to boost the frequency to 5 times the original frequency. It obtains the current sampling frequency reference value of the positioning device, multiplies the reference value by the boost factor to calculate the new sampling frequency, and sends a frequency adjustment command to the positioning device, setting the sampling frequency to the calculated new frequency value, so that the positioning device collects location data at the boosted frequency.

[0087] S403: Obtain the customer center position collected by the improved collection frequency, compare the customer center position with the station position data, and determine that the customer has arrived at the station position when the center position falls within the preset second range of the station position.

[0088] The customer center position refers to the spatial coordinates of the geometric center point of the customer's body, calculated by taking the center point from the boundary coordinates of the customer's body detection area. The preset second range represents a spherical spatial region centered on the station location with a radius equal to a second threshold distance. This second threshold distance is less than the first threshold distance and is used to accurately determine arrival status. Comparison refers to the process of calculating the distance between the customer center position and the station location and comparing it to the threshold.

[0089] The positioning device continuously acquires customer location data at an increased acquisition frequency. It detects the customer's body contour boundaries through image recognition or depth sensing, extracts the minimum bounding rectangle or bounding box of the body contour, calculates the maximum and minimum values ​​of the bounding box along each coordinate axis, and uses the average of the maximum and minimum values ​​for each axis as the center coordinate component of that axis. Combining these center coordinate components yields the three-dimensional coordinates of the customer's center position. The device reads the coordinates of the station location and a preset second-range threshold distance (e.g., 0.3 meters). It calculates the Euclidean distance between the customer's center position coordinates and the station location coordinates. It then determines if the distance is less than the second-range threshold: if the distance is less than the threshold, the customer's center position is considered to fall within the preset second-range, and the customer is determined to have arrived at the station location.

[0090] Based on the above embodiments, as an optional embodiment, S105: after the step of registering and rendering the virtual clothing model with the 3D human body model to generate a virtual fitting image, the method further includes an adjustment step based on user feedback, which may specifically include the following steps: S501: Obtain customer feedback data on virtual fitting images; when the feedback data indicates dissatisfaction, extract the adjustment request type from the feedback data; determine the set of adjustable parameters for the virtual clothing model based on the adjustment request type.

[0091] Evaluation feedback data refers to customer satisfaction ratings for virtual fitting images, including satisfaction scores and specific feedback. Adjustment request type indicates the category of clothing attributes the customer wishes to modify, such as color adjustment, size adjustment, style adjustment, etc. Adjustable parameter set refers to the set of modifiable parameters for the clothing model corresponding to a specific adjustment request type; for example, parameters for color adjustment include hue, saturation, and brightness.

[0092] Customer feedback data is collected through the interactive interface of the terminal device, and the value of the satisfaction rating field is read. It is determined whether the rating is lower than the satisfaction threshold; if the rating is lower than the threshold, the evaluation is considered unsatisfactory. The text of the customer's feedback is read from the feedback data, and a text classification algorithm is used to perform semantic analysis on the feedback, identifying keywords and matching them to a pre-set adjustment request type library to determine the adjustment request type. The parameter items associated with this adjustment request type are queried from the parameter configuration table of the virtual clothing model. For example, when the adjustment request type is color adjustment, the hue parameter, saturation parameter, and brightness parameter are retrieved; when the adjustment request type is size adjustment, the overall scaling parameter, local size parameter, etc., are retrieved. The retrieved parameter items are compiled into an adjustable parameter set, and the name, value range, and current value of each parameter are recorded.

[0093] S502: Extract historical adjustment preferences corresponding to the adjustment requirement type from member profile data; assign values ​​to parameters in the adjustable parameter set according to historical adjustment preferences to generate the adjusted virtual clothing model; re-register and render the adjusted virtual clothing model with the 3D human body model to generate the updated virtual fitting image.

[0094] Historical adjustment preferences refer to the parameter adjustment tendencies summarized from customers' past virtual fitting adjustment records, recording the commonly used adjustment values ​​for parameters under each adjustment need type. Assignment refers to the process of applying the parameter values ​​from historical adjustment preferences to the corresponding parameters in the adjustable parameter set.

[0095] The system reads historical virtual try-on records from member profiles, filters out records containing adjustment operations, and extracts the adjustment request type and parameter adjustment values ​​for each adjustment. It then calculates the historical adjustment values ​​for each parameter under the current adjustment request type, selects the most frequent adjustment value as the historical adjustment preference value for that parameter, iterates through the adjustable parameter set, finds the corresponding preference value from the historical adjustment preferences, and modifies the current value of the parameter to the preference value. Next, it performs parameter update operations on the virtual clothing model, regenerating the clothing's geometry, color, texture, and other attributes based on the modified parameter values ​​to obtain the adjusted virtual clothing model. Finally, it registers the adjusted virtual clothing model with a 3D human body model, performs skeletal binding and mesh deformation operations, sets rendering parameters, calculates lighting effects, and generates an updated virtual try-on image.

[0096] Based on the above embodiments, as an optional embodiment, S106: the step of generating size recommendation information based on virtual fitting images and human body scan data may specifically include the following steps: S601: Divide the human body scan data into regions and extract the part size data of multiple preset body parts; identify the clothing partitions corresponding to the multiple preset body parts from the virtual clothing model and extract the partition size data of each clothing partition.

[0097] Region segmentation refers to the process of dividing human body scan data into different body parts according to anatomical structure. Preset body parts represent predefined key body areas, such as shoulders, chest, waist, hips, and thighs. Part size data refers to the circumference, width, and other dimensional measurements of each preset body part. Clothing zones represent areas in a virtual clothing model corresponding to body parts, such as the shoulder zone or chest zone of a top. Zone size data refers to the spatial dimensional measurements of each clothing zone.

[0098] Skeletal keypoint detection is performed on the point cloud or mesh model of human body scan data to identify the locations of key points such as the shoulder joint, chest center, waist, and hip joint. Based on the vertical coordinates of the keypoints, the human body model is horizontally divided into multiple regions, for example, the chest region is divided using shoulder joint height and waist height as boundaries. The perimeter of the cross-sectional contour of each region is calculated, and the perimeter value is extracted as the part size data. Mesh regions that overlap with the spatial positions of each body part are identified from the mesh structure of the virtual clothing model and marked as the corresponding clothing zones. The cross-sectional contour of each clothing zone is measured, and the perimeter or width is calculated to obtain the zone size data. A correspondence table between body parts and clothing zones is established, recording the part size data of each body part and the zone size data of the corresponding zone.

[0099] S602: Calculate the size deviation between the part size data of each preset body part and the part size data of the corresponding clothing section.

[0100] Size deviation values ​​represent the difference between the size of a garment section and the size of a body part. They are used to quantify the tightness or looseness of the garment in that part. A positive value indicates that the garment size is larger than the body size, while a negative value indicates that the garment size is smaller than the body size.

[0101] The system iterates through the mapping table between body parts and clothing sections, reading the part size data for each preset body part and the corresponding section size data for the clothing section. For each pair of corresponding parts and sections, the system calculates the size deviation value using the formula: the size deviation value equals the section size data minus the part size data. For example, when the chest part size is 90 cm and the corresponding clothing section size is 95 cm, the size deviation value is calculated as 95 minus 90, which equals 5 cm. The calculated size deviation value is then associated with and stored with the corresponding body part, forming a data record containing the body part name, part size, section size, and size deviation value. The size deviation value calculation is completed for all preset body parts, generating a complete list of size deviation values.

[0102] S603: Extract clothing functional requirements from shopping intentions, set fit thresholds for each preset body part based on clothing functional requirements; compare each size deviation value with the corresponding preset body part fit threshold, generate part size suggestions for each preset body part, and summarize part size suggestions to generate size recommendation information.

[0103] Clothing functional requirements refer to customers' demands for the wearing scenarios and effects of clothing. For example, sportswear requires a loose and comfortable fit, formal wear requires a slim and well-fitting look, and casual wear requires a moderate fit. Fit thresholds represent the reasonable range of size deviations for various body parts under specific functional requirements, defined by minimum and maximum values. Body part size recommendations refer to size adjustment suggestions for specific body parts, including maintaining the current size, increasing by one size, or decreasing by one size.

[0104] The system reads the clothing functional requirement field from the shopping intent to identify the functional requirement type. It then queries the fit threshold configuration library for the corresponding body part for that functional requirement type; for example, the chest fit threshold for sportswear is set to 5-10 cm, and for formal wear, it's set to 2-4 cm. Next, it iterates through the size deviation value list for each body part, reading its size deviation value and fit threshold range. It determines whether the size deviation value is within the threshold range: if it is, it generates a size suggestion to maintain the current size; if the deviation value is less than the minimum threshold, it generates a suggestion to decrease by one size; if the deviation value is greater than the maximum threshold, it generates a suggestion to increase by one size. Finally, it compiles all size suggestions for all body parts, uses a majority vote to determine the overall recommended size, and summarizes the overall size, fit assessments for each body part, and adjustment reasons to generate complete size recommendation information.

[0105] Based on the above embodiments, as an optional embodiment, S101: after the step of obtaining member profile data based on biometric data, the method further includes a step of guiding the user to store items, which may specifically include the following steps: S701: Detects whether a customer is carrying items. When a customer is detected carrying items, it obtains the location data of the item placement area and generates an item storage guidance path from the current location to the item placement area. Based on the item storage guidance path, it identifies available lockers from the set of smart lockers in the item placement area and determines the target locker for the customer from the available lockers based on the estimated time of the customer's arrival at the item placement area.

[0106] The item storage area refers to a designated area within the store for customers to temporarily store their personal belongings. The smart locker collection represents all lockers within the item storage area that are equipped with electronic locks and sensors. An available locker is one that is currently unoccupied and in a usable state. Estimated time refers to the time required for a customer to reach their destination, calculated based on the guide path length and average movement speed.

[0107] Image recognition devices deployed at the entrance detect whether customers are carrying backpacks, handbags, or other items, using object detection algorithms to identify the type and quantity of these items. When a customer is detected carrying items, the coordinates of the item placement area are retrieved from the store layout database. The customer's current location coordinates are obtained, and a path planning algorithm is used to calculate the shortest path from the current location to the item placement area, generating a guide path for item storage. The occupancy status of all lockers within the item placement area is retrieved from the smart locker management system, and lockers with an vacant status are selected to form a list of vacant lockers. The total length of the guide path is calculated, and the estimated time for the customer to reach the item placement area is calculated based on a preset average walking speed; the estimated time is equal to the path length divided by the walking speed. From the list of vacant lockers, the locker closest to the item placement area entrance and not expected to be reserved by other customers within the estimated time is selected as the target locker.

[0108] S702: Obtain the customer's facial feature data or terminal device identifier, generate a dynamic unlock code bound to the target locker, and send the dynamic unlock code to the customer's terminal device; when the locker door is opened and the weight sensor inside the locker detects a weight change, determine that the item has been stored and control the locker to output confirmation information; update the occupancy status of the smart locker collection based on the confirmation information, and associate the target locker with the customer's membership profile data for storage.

[0109] A dynamic unlock code is a one-time password or QR code dynamically generated based on a timestamp and customer identification information, used to unlock the target locker. A weight sensor is a sensing device installed at the bottom of the locker to detect changes in the weight of items inside. Confirmation information indicates that the locker system has successfully stored the item. Associated storage refers to establishing a mapping between locker numbers and customer membership numbers in the database and recording the storage time.

[0110] The system reads the customer's facial features from the biometric data collection system or the customer's terminal device identifier from the membership file. It obtains the current system timestamp, combines the timestamp, customer membership number, and target locker number, and generates a dynamic unlock code using an encryption algorithm, setting the unlock code's validity period to 10 minutes. The dynamic unlock code is sent to the customer's terminal device via push service, displaying the unlock code and usage instructions on the device interface. The system monitors the locker door sensor status; when the door changes from closed to open, it initiates weight sensor data collection. It records the baseline weight value before the door opens and continuously collects weight data during the door's open state. When the weight value increases relative to the baseline value and the increment exceeds a preset threshold, the item is considered successfully stored. The system controls the target locker's display screen or buzzer to output a confirmation message, notifying the customer of successful storage. The system updates the target locker's occupancy status field to "occupied" in the smart locker database and creates a record in the associated storage table, recording the locker number, customer membership number, and storage timestamp.

[0111] The following describes an exemplary 3D virtual fitting guidance and control device provided in an embodiment of this application. Figure 4 This is an exemplary hardware structure diagram of a 3D virtual fitting guidance and control device provided in an embodiment of this application.

[0112] In some embodiments, the 3D virtual fitting guidance and control device is a computer device or includes a computer device. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores data. The network interface of the computer device is used to communicate with other external terminals or servers via a network connection. In some embodiments, the network interface can be a wired network interface; in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, it implements the methods in the embodiments of this application.

[0113] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0114] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0115] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0116] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0117] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for guiding and controlling 3D virtual try-on, characterized in that, The method includes: Obtain biometric data of customers entering the store, and obtain member profile data based on the biometric data; Extract the historical preferences and shopping intentions of customers entering the store from the member profile data, and generate guidance path data based on the historical preferences and shopping intentions; The system acquires the customer location data of the customers entering the store in real time, and compares the customer location data with the guidance path data to determine whether the customers have arrived at the designated station. When the customer location data indicates that a customer has arrived at the designated station, a 3D human body model is generated based on the customer's human body scan data. Based on the 3D human body model and the shopping intention, a virtual clothing model is selected, and the virtual clothing model is registered and rendered with the 3D human body model to generate a virtual fitting image. Generate size recommendation information based on the virtual fitting image and the human body scan data; The size recommendation information and the virtual fitting image are sent to the terminal device of the customer who enters the store to guide the customer in making a purchase.

2. The 3D virtual fitting guidance and control method according to claim 1, characterized in that, The step of generating guided path data based on the historical preferences and the shopping intent includes: Detect whether there are other customers entering the store. If there are no other customers entering the store, extract the target clothing type and historical browsing path from the historical preferences. A shopping urgency score is generated based on the shopping intention, and a path planning strategy is determined based on the shopping urgency score. The path planning strategy includes a shortest path strategy and a preferred path strategy. Query the corresponding product area set based on the target clothing type; When the path planning strategy is the shortest path strategy, the product area closest to the entrance is selected from the product area set as the target product area; When the path planning strategy is the preferred path strategy, the historical browsing path is matched with the product area set, the product areas in the product area set are sorted according to the access frequency in the historical browsing path, and the product area with the highest access frequency is selected as the target product area. The station location is determined based on the target product area, and guidance path data from the entrance to the station location is generated.

3. The 3D fitting guidance and control method according to claim 2, characterized in that, The method further includes: When other customers enter the store, obtain the other guidance path data generated by those other customers; Extract the location data of occupied stations from the other guidance path data, and construct a station occupancy table; Based on the historical preferences and shopping intentions, the target product area of ​​the in-store customer is determined, and a set of candidate station locations within the target product area is identified; The candidate site location set is compared with the site location occupancy table. When it is detected that the target site location in the candidate site location set has been occupied, the membership level data of the customers who entered the store is extracted from the membership profile data, and the shopping urgency score of the customers who entered the store is extracted. The path priority weight of the customer entering the store is calculated based on the membership level data and the shopping urgency score, and the path priority weight is compared with the path priority weight of other customers entering the store who occupy the target station. When the path priority weight of the customer entering the store is greater than the path priority weight of other customers entering the store at the target station, the other customers occupying the target station are reassigned to select a station and the other guidance path data is updated. The target station is then assigned to the customer entering the store to generate the guidance path data. When the path priority weight of the customer entering the store is less than the path priority weight of other customers entering the store at the target station, an unoccupied station is selected from the candidate station set as a backup station, and guidance path data containing the backup station is generated.

4. The 3D virtual fitting guidance and control method according to claim 1, characterized in that, The step of comparing the customer location data with the guidance path data to determine whether the customer has arrived at the designated station includes: Extract station locations and path node sequences from the guidance path data; The customer location data is matched with the path node sequence to determine the path node where the customer is currently located. Based on the positional relationship between the customer's current path node and the station location, determine whether the customer is within a preset first range of the station location; When it is determined that a customer is within a preset first range of the location of the station, the collection frequency of the customer location data is increased according to a preset frequency adjustment rule; The customer's center position is obtained from the improved collection frequency. The customer's center position is compared with the station location data. When the center position falls within the preset second range of the station location, it is determined that the customer has arrived at the station location.

5. The 3D virtual fitting guidance and control method according to claim 1, characterized in that, After registering and rendering the virtual clothing model with the 3D human body model to generate a virtual fitting image, the method further includes: Obtain customer feedback data on the virtual fitting images; When the evaluation feedback data indicates dissatisfaction, the adjustment requirement type is extracted from the evaluation feedback data; The set of adjustable parameters for the virtual clothing model is determined based on the type of adjustment requirement. Extract historical adjustment preferences corresponding to the adjustment request type from the member profile data; The parameters in the adjustable parameter set are assigned values ​​according to the historical adjustment preferences to generate the adjusted virtual clothing model; The adjusted virtual clothing model and the 3D human body model are re-registered and rendered to generate an updated virtual try-on image.

6. The 3D virtual fitting guidance and control method according to claim 1, characterized in that, The step of generating size recommendation information based on the virtual fitting image and the human body scan data includes: The human body scan data is divided into regions, and the size data of multiple preset body parts are extracted; Identify clothing sections corresponding to the multiple preset body parts from the virtual clothing model, and extract the section size data of each clothing section; Calculate the size deviation between the size data of each preset body part and the size data of the corresponding clothing section; Extract clothing functional requirements from the shopping intention, and set a fit threshold for each preset body part based on the clothing functional requirements; Each of the size deviation values ​​is compared with the corresponding preset body part fit threshold to generate part size suggestions for each preset body part, and the part size suggestions are summarized to generate size recommendation information.

7. The 3D virtual fitting guidance and control method according to claim 1, characterized in that, After obtaining member profile data based on the biometric data, the process further includes: The system detects whether a customer is carrying items. When items are detected, the system obtains the location data of the item placement area and generates an item storage guidance path from the current location to the item placement area. Based on the item storage guidance path, an available locker is identified from the set of smart lockers in the item placement area, and the customer's target locker is determined from the available lockers based on the estimated time of arrival at the item placement area. Obtain the customer's facial feature data or terminal device identifier, generate a dynamic unlock code bound to the target locker, and send the dynamic unlock code to the customer's terminal device; When the door of the target locker is opened and the weight sensor inside the target locker detects a change in weight, it is determined that the item has been stored and the target locker is controlled to output a confirmation message. The occupancy status of the smart locker set is updated based on the confirmation information, and the target locker is associated with and stored in relation to the customer's membership profile data.

8. A 3D virtual fitting guidance and control device, characterized in that, The 3D virtual fitting guidance and control device includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the 3D virtual fitting guidance and control device to perform the method as described in any one of claims 1-7.

9. A computer program product containing instructions, characterized in that, When the computer program product is run on the 3D fitting room guidance and control device, the 3D fitting room guidance and control device performs the method as described in any one of claims 1-7.

10. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the guidance and control device of the 3D fitting room, the guidance and control device of the 3D fitting room performs the method as described in any one of claims 1-7.

Images