E-commerce virtual scene generation method and system based on digital twinning and visual interaction

Abstract

The invention discloses an e-commerce virtual scene generation method and system based on digital twinning and visual interaction, and belongs to the technical field of computer interaction. According to the method, the basic data set of the e-commerce scene is acquired, and then scene modeling is performed by using digital twinning, so that the initial virtual scene model containing the three-dimensional space layout of the commodity and the visual features of the environmental elements can be generated; performing visual interaction feature analysis on the initial model, and innovatively mining user interaction attention region features and interaction operation preference features as key bases for optimizing the virtual scene; and on the basis of the characteristics, interaction optimization adjustment is performed on the initial model, and finally the e-commerce virtual scene supporting dynamic visual interaction is generated, so that the user interaction experience of the e-commerce scene is comprehensively improved, the crossing from the traditional static display to the dynamic and intelligent interaction experience is realized, and the immersive interactive e-commerce virtual scene construction is realized.

Description

Technical Field

[0001] The present invention belongs to the field of computer interaction technology, and specifically relates to a method and system for generating e-commerce virtual scenes based on digital twins and visual interaction. Background Art

[0002] In the realm of digital e-commerce, traditional e-commerce platforms primarily display products using two-dimensional images and text descriptions, making it difficult to provide users with an immersive shopping experience. Some technologies that attempt to create virtual scenes simply present a fixed virtual space, lacking precise simulation of product layout and environmental elements; or they fail to deeply analyze user interaction behavior and optimize scenes based on user preferences.

[0003] In view of this, how to combine digital twin modeling, visual interaction feature analysis and targeted optimization and adjustment to generate a virtual scene that supports dynamic visual interaction e-commerce is a technical challenge that needs to be overcome. Summary of the Invention

[0004] The present invention provides a method and system for generating e-commerce virtual scenes based on digital twins and visual interaction, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0005] An embodiment of the present invention provides an e-commerce virtual scene generation method based on digital twin and visual interaction, which is applied to an e-commerce virtual scene generation system. The method includes: Obtain the basic data set corresponding to the e-commerce scenario; Performing digital twin scene modeling processing based on the basic data set to generate an initial virtual scene model that includes product three-dimensional spatial layout information and visual features of environmental elements; Performing visual interaction feature analysis on the initial virtual scene model to generate user interaction focus area features and interaction operation preference features; The initial virtual scene model is interactively optimized and adjusted according to the user interaction focus area characteristics and the interactive operation preference characteristics to generate an e-commerce virtual scene that supports dynamic visual interaction.

[0006] An embodiment of the present invention provides an e-commerce virtual scene generation system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the above-mentioned method.

[0007] An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0008] In an embodiment of the present invention, by obtaining a basic data set of e-commerce scenarios and then using digital twins for scene modeling, an initial virtual scene model containing the three-dimensional spatial layout of goods and visual features of environmental elements can be generated; the initial model is subjected to visual interaction feature analysis, and the user interaction focus area features and interactive operation preference features that serve as the key basis for optimizing the virtual scene are innovatively mined; based on these features, the initial model is interactively optimized and adjusted, and finally an e-commerce virtual scene that supports dynamic visual interaction is generated, which comprehensively improves the user interaction experience of the e-commerce scenario, realizes the leap from traditional static display to dynamic and intelligent interactive experience, and realizes the construction of an immersive and interactive e-commerce virtual scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a flowchart of a method for generating an e-commerce virtual scene based on digital twin and visual interaction provided by an embodiment of the present invention.

[0010] Figure 2 A schematic diagram of the structure of an e-commerce virtual scene generation system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0011] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0012] The terms "first," "second," and the like are used in the present invention to distinguish similar objects, and are not intended to describe a specific order or precedence. It should be understood that such terms are interchangeable where appropriate, so that embodiments of the present invention can be implemented in an order other than that illustrated or described herein. Furthermore, the objects distinguished by "first," "second," and the like generally refer to a class of objects and do not limit the number of objects. For example, the first object can be one or more. Furthermore, the term "and / or" in the present invention refers to at least one of the connected objects, and the character " / " generally indicates an "or" relationship between the connected objects.

[0013] Figure 1 A method for generating an e-commerce virtual scene based on digital twins and visual interaction is shown, which is applied to an e-commerce virtual scene generation system. The method includes the following steps 110 to 140.

[0014] It should be noted that, in the entire process of acquiring and processing user historical interaction behavior data in the embodiment of the present invention, ensuring the user's right to know and authorization is the core premise. All collection operations are performed on the basis of the user's explicit knowledge and authorization consent.

[0015] When a user visits an e-commerce platform for the first time, the system will clearly explain through prominent prompts (such as overlay pop-ups or strong guidance pages) that behavioral data such as browsing paths, click operations and dwell time will be collected, and the use of the data will be fully informed (including for building virtual scenes, optimizing interactive experience, etc.). Users must actively check the "Agree to Agreement" or click the "Authorize" button before using the service. The behavioral data of unauthorized users will not be tracked.

[0016] Furthermore, the data collection process adheres to the principle of minimum necessary. For example, the user behavior tracking system is only activated when a user visits a product detail page. The click event monitoring module only records the coordinates of actions related to product interactions. The page dwell statistics module only calculates the duration of time spent on the product display area to avoid collecting irrelevant behavioral data. Before performing corresponding operations (such as mapping screen click coordinates to real-world coordinates), the system will provide a second confirmation of authorization status through an interactive notification. Users can review the current authorization scope at any time in the privacy settings, disable specific data collection functions in real time, or request the deletion of historical data.

[0017] During the data processing phase, strict data isolation and anonymization mechanisms are implemented. All behavioral data is stripped of direct identifiers before timestamp alignment and spatial coordinate conversion, retaining only the impersonal features required to connect to the virtual scene. Data security is ensured through encrypted storage and access controls. The platform regularly sends users data usage reports, including specific data processing scenarios (e.g., "Your click location data is used to recreate the physical store's merchandise display layout"), demonstrations of anonymization effectiveness, and examples of actual data contributions to feature optimization. This ensures users consistently understand data usage and maintains transparency regarding dynamic authorization.

[0018] Step 110: Obtain a basic data set corresponding to the e-commerce scenario.

[0019] It can be understood that this step is intended to collect the necessary data for generating e-commerce virtual scenes. In one example, the basic data set includes user historical interaction behavior data and physical scene visual data. The user historical interaction behavior data consists of the user's browsing path data, click operation data, and dwell time data on the e-commerce platform. The physical scene visual data consists of multi-angle image records of the real product display space and ambient lighting condition records. Furthermore, the basic data set corresponding to the e-commerce scene is obtained, including: Step 111: Collect the browsing path data of users on the product details page through the user behavior tracking system of the e-commerce platform, and the browsing path data includes the order of product pages visited by users and the page jump timestamp; record the user's click operation data on the product thumbnail through the click event monitoring module, and the click operation data includes the click location coordinates and the click duration; collect the user's residence time data in different product display areas through the page residence time statistics module, and the residence time data includes area identification information and corresponding residence time records.

[0020] For example, on an electronics e-commerce platform, when user A enters the platform and begins browsing the mobile phone product details page, the system begins recording their browsing path. User A first visits the page for mobile phone model M1, then spends some time there before jumping to the page for mobile phone model M2. The system accurately records the order in which these two pages were visited, as well as the timestamps of the page jumps. The click event monitoring module operates synchronously. When user A clicks the thumbnail of mobile phone M1, the module records the click location coordinates, for example, at the horizontal coordinate X1 and the vertical coordinate Y1 on the screen, and also records the click duration as T1. The page dwell time statistics module also collects data. If user A spends S1 in the display area of ​​mobile phone M1, the system records the area as "Mobile phone M1 display area" and the corresponding dwell time S1. Similarly, similar data from numerous users is continuously collected, forming a rich data set of historical user interaction behaviors.

[0021] Step 112: Use a visual acquisition device to capture multi-angle images of the real product display space to generate a multi-angle image record containing the product position relationship and environmental decoration details; use a light sensor to monitor the ambient light conditions of the real product display space in real time to generate an ambient light condition record containing light intensity values ​​and light source direction information for different time periods.

[0022] Next, physical scene visual data was collected for the offline physical display area of ​​the electronic product e-commerce platform. The visual acquisition equipment captured the display area from multiple angles, such as shooting the mobile phone display rack from different angles such as the front, side, and diagonally above. The multi-angle image records obtained not only show the positional relationship between the mobile phones, such as mobile phone M1 on the left and mobile phone M2 on the right and slightly behind, but also record the details of the environmental decoration, such as the color and pattern of the background wall. At the same time, the light sensor monitors the ambient light conditions in real time. During the period from 10 am to 11 am, the light intensity value L1 was detected, and the light source direction was from a 45-degree angle on the upper left. Such data from different time periods were collected to form a record of the ambient light conditions.

[0023] Step 113: The browsing path data, the click operation data, the dwell time data, the multi-angle image records and the ambient lighting condition records are uniformly processed in data format to generate a basic data set with timestamp alignment and spatial coordinate alignment.

[0024] It is understandable that in order for these different types of data to work together, data formats need to be unified. As a preferred embodiment, the browsing path data, the click operation data, the dwell time data, the multi-angle image records, and the ambient lighting condition records are unified in data format to generate a basic data set with timestamp alignment and spatial coordinate alignment, including: Step 1131: Perform timestamp standardization processing on the browsing path data, click operation data, and dwell time data, and convert timestamps from different sources into a unified platform time format.

[0025] In the application scenario of this electronics e-commerce platform, the timestamp formats recorded by different systems may vary. For example, the timestamps recorded by the user behavior tracking system may be based on the system's internal time code, while the timestamps recorded by the click event monitoring module may have a different format. Through pre-set timestamp standardization (converting timestamps of different formats to a unified format specified by the platform, such as year-month-day hour:minute:second format), the timestamps of browsing path data, click operation data, and dwell time data are all converted to the platform time format, facilitating subsequent time-based data correlation and analysis.

[0026] Step 1132: performing timestamp matching processing on the multi-angle image records and the ambient light condition records, and correlating the image capture time with the light monitoring time.

[0027] In this embodiment of the present invention, multi-angle image records and ambient lighting condition records each have their own time stamp. To match the lighting information in the image with the actual lighting monitoring data, a timestamp matching process is used (by comparing the time stamps in the image records with the time stamps in the lighting condition records, finding the closest data pair in time for association). This allows the multi-angle image captured at a specific moment to be associated with the ambient lighting conditions monitored at or near the same moment. This allows subsequent scene modeling to accurately reproduce the product display scenes under different lighting conditions at different times.

[0028] Step 1133: Perform spatial coordinate calibration processing on the multi-angle image records, by setting reference marking points in the real commodity display space to convert the image pixel coordinates into real space coordinates.

[0029] Alternatively, some reference markers can be set in the actual mobile phone display space, such as at the four corners of the display shelf. Then, through spatial coordinate calibration (based on image formation principles and the known real-world coordinates of the markers, the image pixel coordinates are converted to real-world coordinates by calculating the relative relationship between the product position in the image and the markers). This allows the pixel coordinates of the mobile phone or other product in the multi-angle image to be converted to real-world coordinates, thereby locating the product's position in real space.

[0030] Step 1134: Perform spatial coordinate conversion processing on the click position coordinates in the user's historical interaction behavior data, and map the screen coordinates of the e-commerce page to real space coordinates.

[0031] It can be understood that the click location of the user on the e-commerce page is expressed in screen coordinates. In order to establish a connection with the real space coordinates, spatial coordinate conversion processing is used (based on the mapping relationship between the e-commerce page and the actual product display space, the screen coordinates are converted into real space coordinates by calculating the correspondence between the screen coordinates and the reference points in the real space). The click location coordinates are converted from screen coordinates to real space coordinates, so that the user's interactive behavior can be located in the real space.

[0032] Step 1135: Perform correlation verification on the standardized timestamp data and the converted spatial coordinate data to eliminate abnormal data with missing timestamps or spatial coordinate deviations exceeding a preset range.

[0033] Specifically, timestamp data and spatial coordinate data can be cross-checked to verify the consistency of time and space information. For example, the spatial coordinates corresponding to a timestamp can be checked to see if they are reasonable. This allows for verification of the processed timestamp and spatial coordinate data. If a data record is found to have a missing timestamp or a large deviation in spatial coordinates that exceeds a preset reasonable range, the abnormal data is removed to ensure data quality.

[0034] Step 1136: Bind and store the verified timestamp data and spatial coordinate data with the original data record to generate a basic data set with timestamp alignment and spatial coordinate alignment.

[0035] Furthermore, the verified timestamp data and spatial coordinate data are bound to their corresponding original browsing path data, click operation data, dwell time data, multi-angle image records, and ambient lighting condition records, and stored in the database, thereby forming a basic data set with aligned timestamps and spatial coordinates.

[0036] Step 120: Perform digital twin scene modeling processing based on the basic data set to generate an initial virtual scene model that includes the three-dimensional spatial layout information of the product and the visual features of the environmental elements.

[0037] The embodiment of the present invention constructs an initial model of the e-commerce virtual scene based on the previously acquired and processed basic data set. In another example, the three-dimensional spatial layout information of the goods includes the position coordinate relationship and display hierarchy relationship of the goods in the virtual scene, and the visual characteristics of the environmental elements include the material texture information and light intensity distribution information of the background decoration in the virtual scene. In detail, the digital twin scene modeling processing based on the basic data set to generate an initial virtual scene model containing the three-dimensional spatial layout information of the goods and the visual characteristics of the environmental elements includes: Step 121: extracting multi-angle image records in the physical scene visual data from the basic data set, performing feature point matching processing on the multi-angle image records, and generating three-dimensional point cloud data of the real product display space.

[0038] As an implementation method, performing feature point matching processing on the multi-angle image records to generate three-dimensional point cloud data of the real product display space includes: Step 1211: grayscale preprocessing is performed on the multi-angle image record; feature points with scale invariance are extracted from the grayscale image using a feature point detection algorithm; descriptor generation processing is performed on the feature points to generate a feature description vector containing feature point position, scale and direction information.

[0039] When processing multi-angle image recordings of mobile phone displays, the images are first pre-processed by grayscale conversion, converting color images into grayscale images to simplify subsequent processing. Next, feature point detection (analyzing the grayscale changes in the image to identify points with distinct features that maintain stable characteristics at different scales) is used to extract scale-invariant feature points from the grayscaled images, such as corner points on the phone's edges and special marking points on the screen. These feature points are then subjected to descriptor generation processing (based on the grayscale distribution around the feature point, a feature description vector containing the feature point's position, scale, and orientation information is calculated to facilitate subsequent feature point matching).

[0040] Step 1212: Establish a correspondence between feature points in different images through similarity matching processing of feature description vectors; and calculate the three-dimensional spatial coordinates of the feature points using triangulation based on the correspondence and the internal and external parameters of the image capture device.

[0041] The feature description vectors of images taken at different angles are similarly matched, and the similarity between the different feature description vectors is calculated. Feature point pairs with similarities exceeding a certain threshold are found, and the corresponding relationship between the feature points in different images is established. After clarifying the corresponding relationship between the feature points between images and the internal and external parameters of the image capture device, the three-dimensional coordinates of the feature points are calculated using triangulation (using the known projection positions of the feature points in different images and the parameters of the capture device, and using the geometric relationship of triangles to calculate the coordinates of the feature points in three dimensions).

[0042] Step 1213: Aggregate the three-dimensional spatial coordinates of all feature points to generate three-dimensional point cloud data of the real product display space, wherein the three-dimensional point cloud data includes coordinate information of product outline feature points and environmental decoration feature points.

[0043] Furthermore, the three-dimensional spatial coordinates of all the calculated feature points are collected and aggregated, and the coordinates of each feature point are combined to form a data set containing the information of all feature points, thereby generating three-dimensional point cloud data of the real product display space. The three-dimensional point cloud data not only contains the coordinate information of the contour feature points of products such as mobile phones, but also contains the coordinate information of environmental decoration feature points such as background decorations, fully presenting the geometric characteristics of the real display space.

[0044] Step 122: Perform spatial coordinate reconstruction based on the three-dimensional point cloud data to determine the position coordinate relationship and display hierarchy relationship of the products in the real display space, wherein the display hierarchy relationship includes the covering and occlusion relationship between the upper display products and the lower display products.

[0045] When reconstructing spatial coordinates using 3D point cloud data, the precise coordinate relationships of the products in real space, as well as the display hierarchical relationships between products, can be determined through calculation and analysis based on the coordinates of the feature points in the 3D point cloud data. This allows the user to determine which products are on the upper and lower levels, as well as the degree of overlap or occlusion between them. This allows the user to determine the coordinate relationships and display hierarchical relationships of products such as mobile phones in the real display space. For example, the user can determine that mobile phone M1 is on the lower level, while mobile phone M2 is on the upper level and partially obscures mobile phone M1.

[0046] Step 123: Perform virtual space mapping processing on the position coordinate relationship and the display level relationship to generate three-dimensional space layout information of the products in the virtual scene.

[0047] In an embodiment of the present invention, the position coordinate relationship and display hierarchy relationship determined in the real display space can be converted into the virtual space, the corresponding position and hierarchy of the goods in the virtual scene can be determined, and then the three-dimensional spatial layout information of the goods in the virtual scene can be generated, so that the virtual scene can accurately reflect the layout of the real goods.

[0048] Step 124: extracting the ambient lighting condition records in the physical scene visual data from the basic data set, analyzing the lighting intensity values ​​and light source direction information in different time periods, and generating lighting intensity distribution information in the virtual scene.

[0049] It can be understood that the ambient lighting condition records in the physical scene visual data are taken out from the basic data set, and the light intensity values ​​and light source direction information recorded in different time periods are analyzed. According to the light intensity and direction data at different times, the light intensity distribution information of each position in the virtual scene is generated through calculation and interpolation methods, and the lighting changes in the real environment are simulated, thereby generating the light intensity distribution information in the virtual scene, such as determining that some areas in the virtual scene have stronger lighting and some areas have weaker lighting.

[0050] Step 125: extracting texture features of the background decoration from the multi-angle image records, performing resolution enhancement processing on the texture features, and generating material texture information of the background decoration in the virtual scene.

[0051] In this embodiment of the present invention, background decorations are located in multi-angle image records and their texture features are extracted. These texture features are then subjected to resolution enhancement, such as through image processing techniques, to enhance detail and increase resolution, making the texture clearer and more realistic. This generates material texture information for the background decorations in the virtual scene, making the background decorations in the virtual scene more realistic.

[0052] Step 126: The three-dimensional space layout information, light intensity distribution information and material texture information are subjected to scene element fusion processing to generate an initial virtual scene model containing the three-dimensional space layout information of the product and the visual characteristics of the environmental elements. The scene element fusion processing is achieved by associating and binding the three-dimensional space coordinates with the light intensity value and the material texture parameters.

[0053] It can be understood that the generated three-dimensional space layout information, light intensity distribution information and material texture information are integrated, for example, the three-dimensional space coordinates are associated with the light intensity value and material texture parameters, and different types of information are integrated to form a complete virtual scene model, thereby generating an initial virtual scene model that includes the three-dimensional space layout information of the product and the visual characteristics of the environmental elements, and completing the modeling of the initial virtual scene.

[0054] Step 130: Perform visual interaction feature analysis on the initial virtual scene model to generate user interaction focus area features and interaction operation preference features.

[0055] The embodiment of the present invention analyzes user interaction characteristics for an established initial virtual scene model. In an optional embodiment, the user interaction focus area characteristics are used to indicate the product display area that the user frequently focuses on in the virtual scene, and the interaction operation preference characteristics are used to indicate the user's preferred viewing method for virtual products. The visual interaction feature analysis processing of the initial virtual scene model to generate the user interaction focus area characteristics and interaction operation preference characteristics includes: Step 131: extract browsing path data, click operation data and dwell time data from the user's historical interactive behavior data from the basic data set, and construct a user behavior sequence model. The user behavior sequence model uses timestamps as indexes to associate user operations with corresponding product display areas.

[0056] From the basic data set, we retrieve browsing path data, click operation data, and dwell time data from historical user interaction behavior data. Using timestamps as clues, we utilize a pre-set user behavior sequence construction algorithm (which chronologically associates user browsing, click, and dwell operations with corresponding product display areas to form an ordered behavior sequence model) to construct a user behavior sequence model. For example, in the case of an electronics e-commerce platform, based on user A's recorded actions on products such as mobile phones M1 and M2 at different times, we construct a behavior sequence model for user A, clearly presenting their behavioral trajectory.

[0057] Step 132: Performing spatial coordinate mapping processing on the user behavior sequence model and the three-dimensional spatial layout information of the commodities in the initial virtual scene model to determine the coordinates of the commodity display area visited by the user in the virtual scene.

[0058] In one implementation, performing spatial coordinate mapping processing on the user behavior sequence model and the three-dimensional spatial layout information of the products in the initial virtual scene model to determine the coordinates of the product display area visited by the user in the virtual scene includes: Step 1320: extract the product identification information visited by the user from the user behavior sequence model; extract the product location coordinate relationship corresponding to the product identification information from the three-dimensional spatial layout information of the products in the initial virtual scene model; convert the product location coordinate relationship into a regional coordinate range in the virtual scene; perform time association processing on the product identification information visited by the user and the corresponding regional coordinate range according to the timestamp sequence in the user behavior sequence model, and generate virtual scene regional coordinate records visited by the user at different time points; perform de-duplication and merging processing on the regional coordinate records, eliminate the repeatedly visited regional coordinates, and generate a set of product display area coordinates visited by the user in the virtual scene; perform spatial clustering analysis on the product display area coordinate set, and identify the concentrated area coordinates frequently visited by the user as the product display area coordinates visited by the user in the virtual scene.

[0059] The product identifiers of the user's visits are identified from the user behavior sequence model, such as mobile phone identifiers M1 and M2. The corresponding location coordinates of these products are then found from the three-dimensional spatial layout of the products in the initial virtual scene model and converted into regional coordinate ranges in the virtual scene. Following the timestamp sequence in the user behavior sequence model, the product identifier information and regional coordinate ranges are temporally associated to generate visit records at different time points. These records are then de-duplicated and merged to remove duplicate regional coordinates. Finally, through spatial clustering (clustering spatially similar regional coordinates to identify concentrated areas frequently visited by users), the coordinates of these frequently visited concentrated areas are identified and used as the coordinates of the product display areas visited by users in the virtual scene.

[0060] Step 133: Count the visit frequencies of different product display area coordinates, and generate product display areas that users frequently focus on in the virtual scene as user interaction focus area features.

[0061] In this step, the number of visits to the coordinates of different product display areas can be counted, and the visit frequency of each area can be calculated. The visit frequency of each product display area coordinate can be statistically analyzed. Areas with high frequency are the product display areas where users frequently focus in the virtual scene, and these areas are considered to be the user interaction attention area features. For example, statistics show that the display area where mobile phone M1 is located has a high visit frequency, so this area is determined to be part of the user interaction attention area features.

[0062] Step 134: Analyze the click position coordinates and click duration in the click operation data to identify the user's click operation mode on the virtual product, where the click operation mode includes a single quick click mode and a multiple continuous click mode.

[0063] Specifically, the click operation data is analyzed for click location coordinates and click duration. Based on the distribution of click location coordinates and click duration, the user's click operation pattern is determined, distinguishing between single, rapid clicks and multiple, continuous clicks. This allows the user's click operation pattern for virtual goods to be identified. For example, if the user's click locations are concentrated and the click duration is short, it may be a single, rapid click pattern; if the click locations show a certain pattern and the click duration is long, it may be a multiple, continuous click pattern.

[0064] Step 135: Analyze the area identification information and corresponding stay duration records in the stay time data to identify the user's stay preference for the product display area, where the stay preference includes a long-stay preference for the product details area and a short-stay preference for the product thumbnail area.

[0065] It can be understood that by analyzing the area identification information and corresponding dwell time records in the dwell time data and comparing the dwell time in different areas, it can be determined which areas users tend to spend more time in and which areas they spend less time in, thereby determining their dwelling preferences for the product details area and the product thumbnail area. For example, it was found that users generally spend more time in the product details area of ​​a mobile phone and less time in the product thumbnail area, thereby identifying users' preference for long dwell time in the product details area and short dwell time in the product thumbnail area.

[0066] Step 136: Generate a user's preferred way of viewing virtual goods based on the click operation mode and the stay preference and use it as an interactive operation preference feature. The user's preferred way of viewing virtual goods includes a way of quickly switching goods by continuously clicking and a way of browsing goods in detail by staying for a long time.

[0067] In an embodiment of the present invention, based on the identified click operation mode and dwell preference, the click operation mode and dwell preference are comprehensively analyzed. If the user has multiple continuous click patterns and prefers short dwell times in the thumbnail area, the user tends to quickly switch the viewing method of the product through continuous clicks; if the user prefers long dwell times in the product details area, the user tends to browse the product in detail by lingering for a long time. In this way, the user's preferred virtual product viewing method can be generated and used as an interactive operation preference feature. For example, in the example of an electronic product e-commerce platform, if the user often clicks on the thumbnails of different mobile phones in rapid succession, and stays in the thumbnail area for a short time, but stays for a long time after entering the product details page, it is determined that the user has two viewing methods: quickly switching products through continuous clicks and browsing products in detail by lingering for a long time.

[0068] Step 140: performing interactive optimization and adjustment processing on the initial virtual scene model according to the user interaction focus area characteristics and the interactive operation preference characteristics to generate an e-commerce virtual scene that supports dynamic visual interaction.

[0069] Based on the user interaction characteristics obtained from the previous analysis, an embodiment of the present invention optimizes and adjusts the initial virtual scene model to generate an e-commerce virtual scene that supports dynamic visual interaction. In another example, the dynamic visual interaction includes the user's multi-angle rotation viewing operation of virtual goods and zooming and focusing operations on the product display area. The interactive optimization and adjustment processing of the initial virtual scene model based on the user's interactive focus area characteristics and the interactive operation preference characteristics to generate an e-commerce virtual scene that supports dynamic visual interaction includes: Step 141: extract the coordinates of the frequently focused product display area in the user interaction attention area features, and perform visual highlighting processing on the frequently focused product display area in the initial virtual scene model, wherein the visual highlighting processing includes increasing the regional lighting intensity and optimizing the product arrangement density.

[0070] The coordinates of frequently focused product display areas are identified from the characteristics of user interaction attention areas. For example, in a virtual scene on an electronics e-commerce platform, the display area of ​​mobile phone M1 is determined to be a frequently focused area. In the initial virtual scene model, a visual prominence algorithm is used (for this area, the lighting intensity is increased according to certain rules, such as increasing the light intensity value by a certain percentage; the product arrangement density is optimized, and the spacing between products is calculated and adjusted to make the product arrangement more reasonable and attractive). This area is visually highlighted in the virtual scene, making it more eye-catching.

[0071] Step 142: extracting the user's preferred virtual commodity viewing method from the interactive operation preference feature, and setting a multi-angle rotation viewing interactive interface for the corresponding commodity in the initial virtual scene model, wherein the interactive interface includes a rotation control button and a perspective lock function.

[0072] Furthermore, the user's preferred viewing method for virtual products is obtained from the interactive operation preference characteristics. If it is found that the user has the need to view products from multiple angles, interactive interface elements such as rotation control buttons and perspective lock functions are added to the initial virtual scene model for the relevant products. These functions are associated with the products through programming and interface design, allowing users to easily operate rotation and perspective lock. In this way, a multi-angle rotation viewing interactive interface is set up to meet the user's needs to view products from different angles.

[0073] Step 143: for the long-stay preference product details area in the stay preference, expand the display space of the area in the initial virtual scene model and increase the hierarchical display level of product detail information.

[0074] It can be understood that for product detail areas with medium to long stay preferences, such as the mobile phone product detail area, the spatial scope of this area is expanded in the initial virtual scene model to increase the display space; at the same time, based on factors such as the category and importance of product detail information, a solution is designed to increase the layered display level, and different types of detail information are displayed at different levels. This can expand the display space of this area and increase the layered display level of product detail information, making it easier for users to understand the product more comprehensively and deeply.

[0075] Step 144: For the multiple consecutive click mode products in the click operation mode, a quick view interaction path is set in the initial virtual scene model, wherein the quick view interaction path includes a direct jump link from the thumbnail area to the detail area.

[0076] Optionally, for products that require multiple consecutive clicks, such as some popular mobile phone products, in the initial virtual scene model, by establishing link relationships and interface navigation design, a direct jump link is set between the thumbnail area and the detail area to ensure that users can quickly enter the detail area from the thumbnail when clicking. This can set a quick viewing interaction path and improve the efficiency of users viewing product information.

[0077] Step 145: Conduct an interactive fluency test on the adjusted initial virtual scene model to verify the response delay time of the multi-angle rotation viewing operation and the product display area zoom and focus operation, optimize and adjust the computing resource allocation of the interactive interface based on the response delay time, and generate an e-commerce virtual scene that supports dynamic visual interaction. The computing resource allocation optimization adjustment includes prioritizing the allocation of graphics rendering resources to high-frequency interactive areas and reducing the rendering accuracy of non-interactive areas.

[0078] After adjusting the initial virtual scene model, it is necessary to conduct an interactive fluency test on it. In a preferred embodiment, the interactive fluency test on the adjusted initial virtual scene model verifies the response delay time of multi-angle rotation viewing operations and product display area zoom and focus operations. Based on the response delay time, the computing resource allocation of the interactive interface is optimized and adjusted to generate an e-commerce virtual scene that supports dynamic visual interaction, including: Step 1451: Simulate the user's multi-angle rotation viewing operation on the virtual product, and record the time interval from the user triggering the rotation command to the virtual product completing the rotation action as the rotation response delay time; simulate the user's zoom and focus operation on the product display area, and record the time interval from the user triggering the zoom command to the virtual scene completing the zoom rendering as the zoom response delay time.

[0079] In this step, a pre-set simulation test process (simulating real user actions according to certain rules, initiating rotation and zoom commands, and accurately recording the time interval from command triggering to the completion of the virtual product rotation and the completion of the zoom rendering of the virtual scene) is used to simulate the user's multi-angle rotation of the virtual product and zoom and focus operations on the product display area. The rotation response delay and zoom response delay are recorded respectively. For example, the user's rotation operation on the virtual model of the mobile phone is simulated multiple times, and the response delay of each operation is recorded.

[0080] Step 1452: Repeat the simulation operation for a preset number of times, and calculate the average value and standard deviation of the rotation response delay time as the rotation operation smoothness index; repeat the simulation operation for a preset number of times, and calculate the average value and standard deviation of the zoom response delay time as the zoom operation smoothness index.

[0081] It can be understood that the above simulation operation is repeated a preset number of times. Using preset statistical calculations (calculating the multiple recorded rotation response delay times, calculating the average and standard deviation to measure the smoothness of the rotation operation; performing the same calculation on the zoom response delay times to obtain the zoom operation smoothness index), the average and standard deviation of the rotation response delay times are calculated as the rotation operation smoothness index, and the average and standard deviation of the zoom response delay times are calculated as the zoom operation smoothness index. These indices can be used to quantitatively evaluate the smoothness of the interactive operation.

[0082] Step 1453: Compare the differences between the rotation operation fluency index and the zoom operation fluency index and the preset fluency threshold, determine the type of interactive operation that needs to be optimized, and generate a difference analysis result; generate an interactive fluency test report based on the difference analysis result, and the interactive fluency test report includes the delay time distribution and optimization priority recommendations of different interactive operations.

[0083] Optionally, the rotation and zoom fluency indicators are compared with a pre-set fluency threshold, and the difference between the indicator and the threshold is calculated to determine whether the difference exceeds an acceptable range. This determines the type of interactive operation that needs to be optimized and generates a difference analysis result. Based on this result, the delay time distribution of different interactive operations is sorted and analyzed, and optimization priority recommendations are given based on the degree of difference to form an interactive fluency test report. For example, if the average value of the rotation operation fluency indicator is greater than the preset threshold and the standard deviation is large, it indicates that the rotation operation needs to be optimized, which is clearly stated in the report and corresponding recommendations are given.

[0084] Step 1454: Extract the interactive operation types whose delay time exceeds a preset threshold in the interactive fluency test report; allocate additional graphics processing unit computing resources to the interactive interface corresponding to the interactive operation type, and the graphics processing unit computing resources include texture rendering threads and geometric transformation threads; reduce the computing resource allocation ratio of non-critical interactive operation interfaces, and the non-critical interactive operation interfaces include auxiliary function buttons that are rarely used by users.

[0085] Optionally, identify interaction types with latency exceeding a preset threshold from the interaction fluency test report. For the interaction interfaces corresponding to these interaction types that require optimization, use the preset resource allocation process (allocating additional graphics processing unit computing resources, such as increasing the number of texture rendering threads and geometry transformation threads to improve processing power; at the same time, reducing the computing resource allocation ratio for non-critical interaction interfaces such as auxiliary function buttons that are rarely used by users to balance overall resources) to reallocate computing resources.

[0086] Step 1455: Perform a secondary interactive fluency test on the adjusted computing resource allocation scheme to verify whether the response delay time of the optimized interactive operation meets the preset fluency threshold; if the secondary test result meets the preset fluency threshold, the adjusted initial virtual scene model is determined as an e-commerce virtual scene that supports dynamic visual interaction; if the secondary test result does not meet the preset fluency threshold, repeat the computing resource allocation adjustment and testing process until the response delay time of the interactive operation meets the preset fluency threshold.

[0087] In this step, the adjusted computing resource allocation scheme is re-applied using the pre-set simulation test process to conduct a second interactive fluency test to verify whether the response delay time of the optimized interactive operation meets the preset fluency threshold. If the secondary test results meet the threshold requirements, the adjustment is effective, and the adjusted initial virtual scene model is determined to be an e-commerce virtual scene that supports dynamic visual interaction. If not, the computing resource allocation adjustment and testing process is repeated according to the previous method until the response delay time of the interactive operation meets the preset fluency threshold, ensuring that the generated e-commerce virtual scene has good interactive fluency.

[0088] As a non-limiting embodiment, after generating the e-commerce virtual scene supporting dynamic visual interaction, the method further includes: Perform real-time user interaction behavior perception processing on dynamic visual interactive e-commerce virtual scenes, and generate real-time interaction feedback data including interaction path continuity features and interaction focus drift features; Compare the real-time interaction feedback data with the user's historical interaction behavior data to identify stable feature clusters and dynamically changing feature clusters of user interaction behavior; Based on stable feature clusters, visual stability enhancement is performed on high-frequency interaction areas of dynamic visual interactive e-commerce virtual scenes. This visual stability enhancement process includes maintaining the consistency of regional lighting distribution and the fixedness of product display levels. Based on dynamically changing feature clusters, the low-frequency interaction area of ​​the dynamic visual interactive e-commerce virtual scene is adaptively expanded. The adaptive expansion process includes adding related product display positions according to the focus drift direction and adjusting the dynamic switching frequency of background decoration textures. The processed e-commerce virtual scene is associated with the real-time interactive feedback data and stored to generate an e-commerce virtual scene with self-evolution capabilities.

[0089] During the operation of a dynamic visual interactive e-commerce virtual scene on an electronics e-commerce platform, real-time user interaction can be monitored and analyzed within the virtual scene, recording the user's interaction path and focus changes. This allows for real-time user interaction behavior perception within the virtual scene. For example, by monitoring user clicks, swipes, and other actions while browsing a virtual mobile phone display, real-time interaction feedback data can be generated, including interaction path continuity features and interaction focus drift characteristics.

[0090] Next, real-time interaction feedback data is compared in detail with historical user interaction data, analyzing operational patterns and focus areas to identify stable and dynamically changing features. This allows for the identification of stable and dynamically changing feature clusters of user interaction behavior. For high-frequency interaction areas, visual stability enhancement is performed using a pre-set visual stabilization algorithm based on stable feature clusters (maintaining a consistent lighting pattern in the area and ensuring a fixed, non-volatile display hierarchy). For low-frequency interaction areas, adaptive expansion is performed using a pre-set expansion algorithm based on dynamically changing feature clusters (analyzing focus drift, increasing the display space for related products according to specific rules, and adjusting the frequency of dynamic switching of background decorative textures to attract more user attention).

[0091] Finally, the processed e-commerce virtual scene and real-time interactive feedback data are associated and stored through a preset associative storage algorithm (establishing an association relationship between the data, storing the real-time feedback data and relevant parameters of the virtual scene to facilitate subsequent analysis and further optimization), so that the e-commerce virtual scene has the ability to continuously evolve according to user behavior.

[0092] As a non-limiting embodiment, after generating the e-commerce virtual scene supporting dynamic visual interaction, the method further includes: Extract user interaction focus area features and interaction operation preference features in dynamic visual interactive e-commerce virtual scenes as the benchmark interaction feature set; Obtain a target basic data set corresponding to a target e-commerce scenario, where the target basic data set includes user historical interaction behavior data and physical scene visual data that are co-structured with the basic data set; Perform feature matching evaluation on the baseline interaction feature set and the user historical interaction behavior data in the target basic data set to generate a feature migration adaptation coefficient; Perform interactive feature enhancement on the target basic data set based on the feature migration adaptation coefficient. The interactive feature enhancement includes supplementing the focus area identification and operation preference labeling of the target user's historical interactive behavior data. Based on the enhanced target basic data set, the steps from digital twin scene modeling to interaction optimization and adjustment are repeatedly executed to generate a target e-commerce virtual scene with interaction mode consistency with the dynamic visual interactive e-commerce virtual scene.

[0093] From a generated dynamic visual interactive e-commerce virtual scene for an electronic product e-commerce platform, user interaction focus area features and interaction operation preference features are extracted to form a baseline interaction feature set. For a target e-commerce scene, such as one for a new electronic product line, the corresponding target basic data set is obtained, which also contains historical user interaction behavior data and physical scene visual data. A feature matching evaluation algorithm is used (by comparing the baseline interaction feature set with the historical user interaction behavior data in the target basic data set, calculating the similarity between the features from multiple dimensions to derive a feature transfer adaptation coefficient) to evaluate the feature matching between the two and generate a feature transfer adaptation coefficient. Based on this coefficient, a feature enhancement algorithm is used (by supplementing the target user historical interaction behavior data with more accurate focus area identifiers and annotating operation preference labels to make the data more targeted and instructive) to enhance the interaction features of the target basic data set.

[0094] Afterwards, the process from digital twin scene modeling to interaction optimization and adjustment is repeated based on the enhanced target basic data set. For example, data processing, scene modeling, interaction analysis, and optimization steps are repeated. Ultimately, a target e-commerce virtual scene is generated that maintains the same interaction model as the original dynamic visual interaction e-commerce virtual scene, ensuring that the new scene better meets user needs based on the previous interaction model.

[0095] Based on the above, during the data collection phase, the user behavior tracking system can utilize event-based browser APIs (such as JavaScript's click and mousemove events) to record screen coordinates and operation timestamps. It can also map click locations to specific product display areas through page DOM tree parsing, ensuring the consistency of the screen coordinate system and the page layout model (both using pixel units). The visual acquisition phase draws on photogrammetry principles, using camera calibration techniques such as the Zhang Zhengyou calibration method to obtain device internal and external parameters. A spatial coordinate calibration process based on checkerboard markers is employed to ensure that the conversion from image pixel coordinates to real-world spatial coordinates has a computable projection matrix relationship.

[0096] During 3D reconstruction, feature point matching can utilize scale-invariant feature descriptor algorithms such as SIFT and SURF, and the Random Sample Consensus Algorithm (RANSAC) can be used to eliminate mismatched points, ensuring that similarity matching of feature description vectors meets an achievable similarity threshold (e.g., a Euclidean distance less than 0.7 times the nearest neighbor distance). The triangulation process can rely on known camera pose parameters and epipolar geometry constraints to solve for 3D point coordinates using the least squares method. For user behavior mapping, the transformation of e-commerce page coordinates to real-world space can be accomplished by establishing an explicit perspective projection model or affine transformation matrix, using a database of correspondences between web page elements and physical products as the basis for the transformation.

[0097] Spatial clustering algorithms for interactive feature analysis can employ specific methods such as DBSCAN or K-means, with a neighborhood radius (e.g., virtual scene unit distance) and a minimum point count threshold defined to define "frequently visited areas." Click operation pattern recognition can incorporate time window analysis (e.g., three consecutive clicks within two seconds are considered a continuous pattern) and coordinate aggregation metrics (e.g., a standard deviation of click positions less than 50 pixels) to avoid relying solely on subjective descriptions. In the interactive optimization phase, graphics rendering resource allocation utilizes GPU pipeline scheduling technology, with asynchronous compute shaders used in the texture rendering thread and instanced rendering used in the geometry transformation thread. Precision adjustment is achieved through rendering target resolution grading (e.g., reducing the LOD of non-interactive areas to 50% of the original resolution).

[0098] Furthermore, a feature cluster recognition mechanism can be added to process self-evolving scenarios: stable feature clusters can identify persistent high-frequency areas through sliding window mean filtering (window size 30 operation events), while dynamic feature clusters utilize a Markov chain-based focus transfer probability model. Background texture switching frequency adjustment can be linked to the duration of focus dwell (e.g., increasing the switching frequency to 2Hz for dwell times exceeding 5 seconds), and texture streaming technology can be used for dynamic loading. Feature matching assessment in the feature migration process can pre-define quantitative similarity metrics, such as the Jaccard coefficient to calculate focus region overlap and the KL divergence to measure differences in the distribution of operational preferences. This provides a mathematical definition for the "feature migration adaptation coefficient" (e.g., a weighted average exceeding 0.75 indicates migration).

[0099] It's worth noting that all time-related operations (such as response latency testing) in this embodiment of the present invention are timed uniformly using the millisecond system clock. Light intensity is recorded explicitly in lux and the sensor calibration method is noted. The spatial coordinate system uses a right-handed Cartesian coordinate system with a declared origin (e.g., the lower left corner of a display shelf is defined as (0, 0, 0)). By incorporating computer graphics' frustum clipping technology to optimize the rendering of non-interactive areas and incorporating Fitts's Law from human-computer interaction to design button size and spacing, the implementation path of each algorithm node can be further optimized.

[0100] The embodiment of the present invention obtains a basic data set of e-commerce scenarios and then uses digital twins to perform scene modeling, thereby generating an initial virtual scene model that includes the three-dimensional spatial layout of goods and visual features of environmental elements; performs visual interaction feature analysis on the initial model, and innovatively mines user interaction focus area features and interactive operation preference features that serve as key bases for optimizing virtual scenes; based on these features, the initial model is interactively optimized and adjusted to ultimately generate an e-commerce virtual scene that supports dynamic visual interaction, comprehensively improving the user interaction experience of the e-commerce scenario, achieving a leap from traditional static display to dynamic, intelligent interactive experience, and realizing the construction of an immersive and interactive e-commerce virtual scene.

[0101] Furthermore, Figure 2 This is a structural diagram of an e-commerce virtual scene generation system 200 provided by an embodiment of the present invention. Figure 2 The e-commerce virtual scene generation system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present invention.

[0102] Alternatively, as Figure 2 As shown, the e-commerce virtual scene generation system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present invention.

[0103] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0104] Alternatively, as Figure 2 As shown, the e-commerce virtual scene generation system 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices. Specifically, it may send information or data to other devices, or receive information or data sent by other devices.

[0105] Optionally, the e-commerce virtual scene generation system 200 can implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or the device deployed with the storage engine in each method of the embodiments of the present invention. For the sake of brevity, they are not repeated here.

[0106] It should be understood that the processor in the embodiment of the present invention may be an integrated circuit chip with signal processing capabilities.

[0107] It is understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. It should be noted that the memory of the systems and methods described herein is intended to include but is not limited to suitable types of memory.

[0108] Based on the above, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0109] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present invention is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0110] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a more preferred embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a computer software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0111] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the purpose of the present invention and the scope of protection of the present invention, all of which are protected by the present invention.

Claims

1. A method for generating e-commerce virtual scenes based on digital twins and visual interaction, characterized in that: The method comprises: Obtain the basic data set corresponding to the e-commerce scenario; Performing digital twin scene modeling processing based on the basic data set to generate an initial virtual scene model that includes product three-dimensional spatial layout information and visual features of environmental elements; Performing visual interaction feature analysis on the initial virtual scene model to generate user interaction focus area features and interaction operation preference features; The initial virtual scene model is interactively optimized and adjusted according to the user interaction focus area characteristics and the interactive operation preference characteristics to generate an e-commerce virtual scene that supports dynamic visual interaction.

2. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 1 is characterized in that: The basic data set includes user historical interaction behavior data and physical scene visual data. The user historical interaction behavior data consists of user browsing path data, click operation data, and dwell time data on the e-commerce platform. The physical scene visual data consists of multi-angle image records of the real product display space and ambient lighting condition records. The basic data set corresponding to the e-commerce scene is obtained, including: The user behavior tracking system of the e-commerce platform collects the browsing path data of users on the product details page. The browsing path data includes the order of product pages visited by users and the page jump timestamp; the click event monitoring module records the user's click operation data on the product thumbnails. The click operation data includes the click location coordinates and click duration; the page dwell time statistics module collects the user's dwell time data in different product display areas. The dwell time data includes area identification information and corresponding dwell time records; Use visual acquisition equipment to capture multi-angle images of the real product display space, generating multi-angle image records that include product position relationships and environmental decoration details; use light sensors to monitor the ambient light conditions of the real product display space in real time, generating ambient light condition records that include light intensity values ​​and light source direction information for different time periods; The browsing path data, the click operation data, the dwell time data, the multi-angle image record and the ambient lighting condition record are processed in a unified data format to generate a basic data set with timestamp alignment and spatial coordinate alignment.

3. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 2 is characterized in that: The step of uniformly processing the browsing path data, the click operation data, the dwell time data, the multi-angle image records, and the ambient lighting condition records to generate a basic data set with timestamp alignment and spatial coordinate alignment includes: Performing timestamp standardization on the browsing path data, click operation data, and dwell time data, converting timestamps from different sources into a unified platform time format; Performing timestamp matching processing on the multi-angle image records and the ambient light condition records to associate the image capture time with the light monitoring time; Performing spatial coordinate calibration processing on the multi-angle image records, by setting reference marking points in the real commodity display space to convert the image pixel coordinates into real space coordinates; Performing spatial coordinate conversion processing on the click position coordinates in the user's historical interaction behavior data to map the screen coordinates of the e-commerce page to real space coordinates; Perform correlation verification on the standardized timestamp data and the converted spatial coordinate data to eliminate abnormal data with missing timestamps or spatial coordinate deviations exceeding the preset range; The verified timestamp data and spatial coordinate data are bound and stored with the original data records to generate a basic data set with timestamp alignment and spatial coordinate alignment.

4. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 1 is characterized in that: The three-dimensional spatial layout information of the products includes the position coordinate relationship and display hierarchy relationship of the products in the virtual scene, and the visual features of the environmental elements include the material texture information and light intensity distribution information of the background decoration in the virtual scene. The digital twin scene modeling processing based on the basic data set to generate an initial virtual scene model including the three-dimensional spatial layout information of the products and the visual features of the environmental elements includes: Extracting multi-angle image records from the physical scene visual data from the basic data set, performing feature point matching processing on the multi-angle image records, and generating three-dimensional point cloud data of the real product display space; Performing spatial coordinate reconstruction based on the three-dimensional point cloud data to determine the position coordinate relationship and display hierarchy relationship of the products in the real display space, wherein the display hierarchy relationship includes the covering and occlusion relationship between the upper display products and the lower display products; Performing virtual space mapping processing on the position coordinate relationship and the display hierarchy relationship to generate three-dimensional spatial layout information of the products in the virtual scene; Extracting ambient lighting condition records in the physical scene visual data from the basic data set, analyzing the lighting intensity values ​​and light source direction information in different time periods, and generating lighting intensity distribution information in the virtual scene; Extracting texture features of the background decoration from the multi-angle image records, performing resolution enhancement processing on the texture features, and generating material texture information of the background decoration in the virtual scene; The three-dimensional space layout information, light intensity distribution information and material texture information are subjected to scene element fusion processing to generate an initial virtual scene model containing the three-dimensional space layout information of the product and the visual characteristics of the environmental elements. The scene element fusion processing is achieved by associating and binding the three-dimensional space coordinates with the light intensity value and the material texture parameters.

5. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 4 is characterized in that: The performing feature point matching processing on the multi-angle image records to generate three-dimensional point cloud data of the real commodity display space includes: Performing grayscale preprocessing on the multi-angle image record; extracting feature points with scale invariance from the grayscale image using a feature point detection algorithm; Performing descriptor generation processing on the feature points to generate feature description vectors containing feature point position, scale and direction information; Through similarity matching processing of feature description vectors, the correspondence between feature points of different images is established; Calculating the three-dimensional spatial coordinates of the feature points using a triangulation method based on the corresponding relationship and the internal and external parameters of the image capture device; The three-dimensional spatial coordinates of all feature points are aggregated to generate three-dimensional point cloud data of the real product display space, which contains coordinate information of product outline feature points and environmental decoration feature points.

6. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 1 is characterized in that: The user interaction focus area feature is used to indicate the product display area that the user frequently focuses on in the virtual scene, and the interactive operation preference feature is used to indicate the user's preferred way of viewing virtual products. The visual interaction feature analysis and processing of the initial virtual scene model to generate the user interaction focus area feature and the interactive operation preference feature includes: Extracting browsing path data, click operation data, and dwell time data from the user's historical interactive behavior data from the basic data set, and constructing a user behavior sequence model, wherein the user behavior sequence model associates user operations with corresponding product display areas using timestamps as indexes; Performing spatial coordinate mapping processing on the user behavior sequence model and the three-dimensional spatial layout information of the commodities in the initial virtual scene model to determine the coordinates of the commodity display area visited by the user in the virtual scene; Count the visit frequencies of different product display area coordinates and generate product display areas where users frequently focus in the virtual scene as user interaction attention area features; Analyzing the click position coordinates and click duration in the click operation data to identify the user's click operation mode on the virtual product, wherein the click operation mode includes a single quick click mode and a multiple continuous click mode; Analyzing the area identification information and corresponding stay duration records in the stay time data to identify the user's stay preference for the product display area, wherein the stay preference includes a long stay preference for the product details area and a short stay preference for the product thumbnail area; A user-preferred way of viewing virtual goods is generated based on the click operation mode and the stay preference and used as an interactive operation preference feature. The user-preferred way of viewing virtual goods includes a way of quickly switching goods by continuous clicks and a way of browsing goods in detail by staying for a long time.

7. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 6 is characterized in that: The performing spatial coordinate mapping processing on the user behavior sequence model and the three-dimensional spatial layout information of the commodities in the initial virtual scene model to determine the coordinates of the commodity display area visited by the user in the virtual scene includes: Extracting identification information of commodities accessed by the user from the user behavior sequence model; Extracting the commodity position coordinate relationship corresponding to the commodity identification information from the commodity three-dimensional space layout information of the initial virtual scene model; Converting the product location coordinate relationship into a region coordinate range in a virtual scene; According to the timestamp sequence in the user behavior sequence model, the product identification information visited by the user is temporally associated with the corresponding area coordinate range to generate the coordinate records of the virtual scene area visited by the user at different time points; Performing de-duplication and merging processing on the area coordinate records, eliminating the coordinates of areas visited repeatedly, and generating a coordinate set of the product display areas visited by the user in the virtual scene; A spatial cluster analysis is performed on the product display area coordinate set to identify concentrated area coordinates frequently visited by users as the product display area coordinates visited by users in the virtual scene.

8. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 6 is characterized in that: The dynamic visual interaction includes a user's multi-angle rotation viewing operation on a virtual commodity and a zooming and focusing operation on a commodity display area. The interactive optimization and adjustment processing of the initial virtual scene model based on the user's interactive focus area characteristics and the interactive operation preference characteristics to generate an e-commerce virtual scene supporting dynamic visual interaction includes: Extracting coordinates of frequently focused product display areas from the user interaction attention area features, and performing visual highlighting processing on the frequently focused product display areas in the initial virtual scene model, wherein the visual highlighting processing includes increasing regional lighting intensity and optimizing product arrangement density; Extracting the user's preferred virtual commodity viewing mode from the interactive operation preference feature, and setting a multi-angle rotation viewing interactive interface for the corresponding commodity in the initial virtual scene model, wherein the interactive interface includes a rotation control button and a viewing angle locking function; For the long-stay preference product details area in the stay preference, expand the display space of the area in the initial virtual scene model and add a hierarchical display level of product detail information; For a product in a multiple-continuous-click mode in the click operation mode, a quick view interaction path is set in the initial virtual scene model, wherein the quick view interaction path includes a direct jump link from a thumbnail area to a detail area; The adjusted initial virtual scene model is tested for interactive fluency to verify the response delay time of multi-angle rotation viewing operations and product display area zoom and focus operations. The computing resource allocation of the interactive interface is optimized and adjusted based on the response delay time to generate an e-commerce virtual scene that supports dynamic visual interaction. The computing resource allocation optimization adjustment includes prioritizing the allocation of graphics rendering resources to high-frequency interactive areas and reducing the rendering accuracy of non-interactive areas.

9. The e-commerce virtual scene generation method based on digital twin and visual interaction according to claim 8 is characterized in that: The interactive fluency test is performed on the adjusted initial virtual scene model to verify the response delay time of the multi-angle rotation viewing operation and the product display area zoom and focus operation, and the computing resource allocation of the interactive interface is optimized and adjusted according to the response delay time to generate an e-commerce virtual scene that supports dynamic visual interaction, including: Simulate the user's multi-angle rotation viewing operation of the virtual product, and record the time interval from the user triggering the rotation command to the virtual product completing the rotation action as the rotation response delay time; simulate the user's zoom and focus operation on the product display area, and record the time interval from the user triggering the zoom command to the virtual scene completing the zoom rendering as the zoom response delay time; Repeating the simulation operation for a preset number of times, calculating the average value and standard deviation of the rotation response delay time as a rotation operation smoothness indicator; repeating the simulation operation for a preset number of times, calculating the average value and standard deviation of the zoom response delay time as a zoom operation smoothness indicator; Comparing the differences between the rotation operation fluency index and the zoom operation fluency index and a preset fluency threshold, determining the type of interactive operation that needs to be optimized, and generating a difference analysis result; generating an interactive fluency test report based on the difference analysis result, the interactive fluency test report including the delay time distribution of different interactive operations and optimization priority recommendations; Extracting the interactive operation types whose delay time exceeds a preset threshold in the interactive fluency test report; Allocate additional graphics processing unit computing resources to the interactive interface corresponding to the interactive operation type, the graphics processing unit computing resources including texture rendering threads and geometry transformation threads; reduce the proportion of computing resources allocated to non-critical interactive operation interfaces, the non-critical interactive operation interfaces including auxiliary function buttons that are rarely used by users; Conduct a secondary interactive fluency test on the adjusted computing resource allocation scheme to verify whether the response delay time of the optimized interactive operation meets the preset fluency threshold; If the secondary test result meets the preset fluency threshold, the adjusted initial virtual scene model is determined as an e-commerce virtual scene that supports dynamic visual interaction; if the secondary test result does not meet the preset fluency threshold, the computing resource allocation adjustment and testing process is repeated until the response delay time of the interactive operation meets the preset fluency threshold.

10. An e-commerce virtual scene generation system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 9.

Images