Interior design results evaluation method and system based on virtual reality technology
By combining images segmentation, Bayesian network, Monte Carlo tree search and Markov chain Monte Carlo algorithm in VR interior design system, in-depth analysis of user interaction behavior is solved, and the existing system cannot deeply analyze user preferences. The problem has achieved design optimization and user experience improvement.
Patent Information
- Application Number
- CN202410712561.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-04
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-06-04
AI Technical Summary
The existing VR interior design system cannot deeply analyze user interaction behaviors and preferences, and lacks a comprehensive analysis of user operation data, resulting in limited results in design optimization and user experience improvement.
Using image segmentation, feature extraction, Bayesian network rendering parameter analysis, Monte Carlo tree search processing, principal component analysis and Markov chain Monte Carlo algorithm, combined with a variety of data processing technologies, detailed user behavior patterns and preferences are generated Data, used to optimize design and enhance user experience.
Through detailed user behavior analysis and data processing, scientific design optimization suggestions are provided, which significantly improves design quality and user satisfaction, and enhances the reality and visual effects of virtual reality scenes.
Smart Images

Figure CN118587344B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to an indoor design achievement evaluation method and system based on virtual reality technology. Background Art
[0002] In the existing interior design field, virtual reality (VR) technology is increasingly being used in design display and user experience. Using VR technology, designers can create a three-dimensional virtual environment that allows users to view and interact with design solutions in an immersive way. This approach not only improves user experience, but also displays design effects more intuitively. However, existing VR interior design systems often only provide static displays and simple interactions, and cannot deeply analyze users' interactive behaviors and preferences, thus failing to provide a scientific basis for design optimization.
[0003] Existing VR interior design evaluation schemes lack in-depth analysis of user behaviors and preferences, and are unable to effectively obtain and utilize user operation data in virtual environments. Usually, only basic user operations, such as movement and rotation, are recorded, but there is a lack of comprehensive analysis of key behavioral characteristics such as user dwell time, operation path complexity, and selection frequency. In addition, existing schemes also lack advanced data analysis and modeling technologies when processing these data, and are unable to fully explore user behavior patterns and preferences, resulting in limited effects on design optimization and user experience improvement. Summary of the invention
[0004] In view of this, an embodiment of the present invention provides an interior design achievement evaluation method and system based on virtual reality technology, which are used to improve the accuracy of interior design achievement evaluation based on virtual reality technology.
[0005] The present invention provides an indoor design achievement evaluation method based on virtual reality technology, comprising: performing image segmentation on an acquired design drawing to obtain a segmented image set, and performing feature extraction on each segmented image in the segmented image set to obtain an element feature set; constructing a three-dimensional house model according to the element feature set, and importing the three-dimensional house model into a preset virtual reality scene to generate an initial virtual scene; inputting the three-dimensional house model into a Bayesian network for rendering parameter analysis to obtain target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain a target virtual scene; performing Monte Carlo tree search processing on user interaction behaviors in the target virtual scene to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, residence time, operation frequency and user feedback data; performing principal component analysis on the user operation data to obtain principal component data, and converting the principal component data into multi-dimensional evaluation index data; inputting the evaluation index data into a Markov chain Monte Carlo algorithm for user behavior recognition to obtain user behavior patterns and user preference data, and generating an evaluation report according to the user behavior patterns and the user preference data.
[0006] In the present invention, the steps of performing image segmentation on the acquired design drawings to obtain a segmented image set, and performing feature extraction on each segmented image in the segmented image set to obtain an element feature set include: performing grayscale conversion on the design drawings to obtain grayscale drawings; performing edge detection on the grayscale drawings to obtain object contour data, wherein the object contour data includes: wall contours, door and window contours, and furniture contours; performing region segmentation on the grayscale drawings according to the object contour data to obtain the segmented image set, wherein the segmented image set includes multiple segmented images; performing shape feature extraction on each segmented image to obtain an invariant moment of each segmented image, wherein the invariant moment is used to describe the shape features of the image; performing color feature extraction on each segmented image to obtain a color feature of each segmented image; performing texture feature extraction on each segmented image by a local binary pattern algorithm to obtain a texture feature of each segmented image; and merging the invariant moment of each segmented image, the color feature of each segmented image, and the texture feature of each segmented image into the element feature set.
[0007] In the present invention, the steps of constructing a three-dimensional house model according to the element feature set, and importing the three-dimensional house model into a preset virtual reality scene to generate an initial virtual scene include: constructing a house model according to the invariant moment of each segmented image to obtain a house sub-model set, wherein the house sub-model set includes: a wall sub-model, a floor sub-model, a ceiling sub-model, a door and window sub-model, and a furniture sub-model; mapping the color features of each segmented image and the texture features of each segmented image to the house sub-model set to obtain an initial three-dimensional house model; performing model scale correction on the initial three-dimensional house model to obtain the three-dimensional house model; initializing a virtual reality scene, and importing the three-dimensional house model into the virtual reality scene to obtain a scene to be processed; performing interactive function matching on the scene to be processed to obtain interactive function parameters, wherein the interactive function parameters include: a moving function and an angle switching function; and performing interactive function embedding on the scene to be processed according to the interactive function parameters to obtain the initial virtual scene.
[0008] In the present invention, the step of inputting the three-dimensional model of the house into a Bayesian network for rendering parameter analysis to obtain target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain the target virtual scene includes: performing rendering parameter matching on the three-dimensional model of the house to obtain a rendering parameter set, wherein the rendering parameter set includes: light intensity, light source position, material reflectivity, and shadow effect; inputting the rendering parameter set into the Bayesian network to perform Bayesian network node construction to obtain multiple Bayesian network nodes; performing dependency data analysis on the multiple Bayesian network nodes to obtain dependency data corresponding to each Bayesian network node; performing Bayesian graph structure construction on the multiple Bayesian network nodes according to the dependency data corresponding to each Bayesian network node to obtain a Bayesian graph structure; performing node conditional probability analysis on the Bayesian graph structure by a maximum likelihood estimation algorithm to obtain conditional probability data of each Bayesian network node; performing rendering parameter mapping according to the conditional probability data of each Bayesian network node to obtain the target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain the target virtual scene.
[0009] In the present invention, the user interaction behavior in the target virtual scene is processed by Monte Carlo tree search to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, dwell time, operation frequency and user feedback data, including: extracting interaction actions from the user interaction behavior to obtain an interaction action set, wherein the interaction action set includes: moving, rotating, scaling and selecting objects; extracting the user initial position and user perspective information corresponding to the user interaction behavior, and inputting the user initial position and the user perspective information into a preset Monte Carlo tree search algorithm to match the user operation path to obtain operation path data; extracting dwell points from the operation path data to obtain multiple dwell points; matching the dwell time of each dwell point to obtain the dwell time; calculating the operation frequency based on the operation path data and the multiple dwell points to obtain the operation frequency; collecting user feedback data in real time, and merging the user feedback data, operation path data, dwell time and the operation frequency into the user operation data.
[0010] In the present invention, the steps of performing principal component analysis on the user operation data to obtain principal component data, and converting the principal component data into multi-dimensional evaluation index data include: performing operation feature extraction on the user operation data to obtain an operation feature set; performing covariance matrix calculation on the operation feature set to obtain a covariance matrix corresponding to the operation feature set, wherein the covariance matrix is used to represent the linear correlation between different operation features; performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors; performing principal component data matching on the user operation data according to the eigenvalues and the eigenvectors to obtain the principal component data; performing evaluation index matching on the principal component data to obtain an evaluation index set, wherein the evaluation index set includes: the length of stay in the area of interest, the path complexity, and the operation frequency interval; performing data conversion on the principal component data according to the evaluation index set to obtain the multi-dimensional evaluation index data.
[0011] In the present invention, the step of inputting the evaluation index data into a Markov chain Monte Carlo algorithm for user behavior identification, obtaining a user behavior pattern and user preference data, and generating an evaluation report based on the user behavior pattern and the user preference data includes: performing Markov chain starting point matching on the evaluation index data to obtain a Markov chain starting point, wherein the Markov chain starting point represents a user initial state and a user initial behavior pattern; matching a transition probability according to the Markov chain starting point, and based on the transition probability, inputting the Markov chain starting point into a Markov chain Monte Carlo algorithm for posterior distribution analysis to obtain posterior distribution data; performing user preference data analysis on the user initial state according to the posterior distribution data to obtain user preference data; performing pattern extraction on the user initial behavior pattern according to the posterior distribution data to obtain the user behavior pattern; performing data fusion on the user preference data according to the user behavior pattern to obtain fused behavior data, and generating the evaluation report according to the fused behavior data.
[0012] The present invention also provides an indoor design achievement evaluation system based on virtual reality technology, comprising:
[0013] A segmentation module is used to perform image segmentation on the acquired design drawing to obtain a segmented image set, and perform feature extraction on each segmented image in the segmented image set to obtain an element feature set;
[0014] A construction module, used to construct a three-dimensional model of the house according to the element feature set, and import the three-dimensional model of the house into a preset virtual reality scene to generate an initial virtual scene;
[0015] An analysis module, used for inputting the three-dimensional model of the house into a Bayesian network to perform rendering parameter analysis to obtain target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain a target virtual scene;
[0016] A processing module, used for performing Monte Carlo tree search processing on the user interaction behavior in the target virtual scene to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, residence time, operation frequency and user feedback data;
[0017] A conversion module, used to perform principal component analysis on the user operation data to obtain principal component data, and convert the principal component data into multi-dimensional evaluation index data;
[0018] The identification module is used to input the evaluation index data into the Markov chain Monte Carlo algorithm to perform user behavior identification, obtain user behavior patterns and user preference data, and generate an evaluation report based on the user behavior patterns and the user preference data.
[0019] In the technical solution provided by the present invention, the user's operation data is processed by principal component analysis (PCA) technology to extract the principal component data reflecting the user's behavior characteristics. By recording and analyzing the user's operation behaviors such as movement, rotation, scaling and selection in a virtual reality environment in detail, the relationship between these operation behaviors can be quantified to form a comprehensive operation feature set. Then, the linear correlation and important principal components of the operation feature set are further extracted by covariance matrix calculation and eigenvalue decomposition technology, thereby simplifying the data structure and improving the analysis efficiency. This process not only ensures the accuracy and efficiency of data processing, but also effectively identifies the key behavior patterns and operation habits of users in a virtual environment. On this basis, the present invention uses a Bayesian network to analyze and optimize the rendering parameters of a three-dimensional model of a house. By constructing Bayesian network nodes, defining the conditional dependency between rendering parameters, and using the maximum likelihood estimation algorithm to determine the conditional probability data of each node, it is possible to achieve accurate control and optimization of key rendering parameters such as light intensity, light source position, material reflectivity and shadow effect. This not only improves the realism and visual effects of the virtual reality scene, but also can be personalized according to the user's preferences to enhance the user's immersive experience and satisfaction. In addition, the Monte Carlo Tree Search (MCTS) algorithm is used to match and optimize the user's interactive behavior in the virtual reality environment, which can accurately identify the user's operation path and stay point, thereby deeply analyzing the user's behavior pattern and interest area. Through the detailed analysis of the operation path data, stay time and operation frequency, the user's focus and operation habits in the virtual environment can be fully understood, providing valuable reference for design optimization. In addition, the present invention also performs posterior distribution analysis on the user's behavior pattern through the Markov Chain Monte Carlo algorithm (MCMC), and further identifies and quantifies the user's behavior pattern and preference data. By inputting the user's initial state and transition probability into the MCMC algorithm, a state sequence that conforms to the user's operation habits can be generated, and the posterior distribution data of the user's behavior pattern can be obtained. These data not only reflect the user's behavior probability and preference under different operation states, but also can form a comprehensive user behavior and preference data set through pattern extraction and data fusion. Based on these data sets, a detailed evaluation report can be generated to describe the user's operation process, focus and satisfaction in the virtual environment, provide scientific optimization suggestions for designers, and significantly improve the design quality and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0021] Figure 1 It is a flow chart of an indoor design achievement evaluation method based on virtual reality technology in an embodiment of the present invention;
[0022] Figure 2 Schematic diagram of an interior design achievement evaluation system based on virtual reality technology in an embodiment of the present invention. DETAILED DESCRIPTION
[0023] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0024] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.
[0025] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0026] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , Figure 1 is a flow chart of an indoor design achievement evaluation method based on virtual reality technology according to an embodiment of the present invention. Figure 1 As shown, the following steps are included:
[0027] S101, performing image segmentation on the acquired design drawing to obtain a segmented image set, and performing feature extraction on each segmented image in the segmented image set to obtain an element feature set;
[0028] S102, constructing a three-dimensional model of the house according to the element feature set, and importing the three-dimensional model of the house into a preset virtual reality scene to generate an initial virtual scene;
[0029] S103, inputting the three-dimensional model of the house into the Bayesian network to perform rendering parameter analysis to obtain target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain a target virtual scene;
[0030] S104, performing a Monte Carlo tree search process on the user interaction behavior in the target virtual scene to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, residence time, operation frequency and user feedback data;
[0031] S105, performing principal component analysis on the user operation data to obtain principal component data, and converting the principal component data into multi-dimensional evaluation index data;
[0032] S106: Input the evaluation index data into the Markov Chain Monte Carlo algorithm to perform user behavior recognition, obtain user behavior patterns and user preference data, and generate an evaluation report based on the user behavior patterns and user preference data.
[0033] Specifically, starting from the acquired interior design drawings, image segmentation is performed on these design drawings. This process includes converting the design drawings into grayscale images to simplify data processing, and using edge detection algorithms (such as Canny edge detection) to identify the outlines of objects in the drawings, which may include walls, doors, windows, and furniture. The image is segmented by the identified object outlines to obtain multiple segmented images. Each segmented image represents an independent element in the design drawing, such as a wall, a window, or a piece of furniture. Next, feature extraction is performed on these segmented images. Feature extraction includes shape features (describing the shape of the image by calculating invariant moments), color features (extracting the color distribution of the image by color histograms), and texture features (capturing the texture information of the image by local binary pattern algorithms). These features are combined to form an element feature set.
[0034] For example, suppose there is a design drawing that includes the layout of a room with clearly identified walls, windows, and furniture. After grayscale conversion and edge detection, the outlines of the walls are identified and used as independent segmented images. By calculating the shape features, color features, and texture features of these segmented images, detailed feature descriptions of these elements can be obtained, such as the height, color, and texture of a wall.
[0035] Next, a three-dimensional model of the house is constructed based on these element feature sets. First, based on the invariant moment information of each segmented image, the corresponding house sub-model is constructed, including the wall sub-model, floor sub-model, ceiling sub-model, door and window sub-model, and furniture sub-model. The extracted color features and texture features are mapped to these sub-models to form an initial three-dimensional model of the house. Then, the initial three-dimensional model of the house is scaled to ensure that the size and proportion of the model are consistent with the actual design, and finally a complete three-dimensional model of the house is obtained. Next, this three-dimensional model of the house is imported into the preset virtual reality scene to generate the initial virtual scene. This step includes initializing the virtual reality scene and embedding interactive function parameters, such as movement function and angle switching function, so that users can freely observe and operate in the virtual reality environment.
[0036] For example, if the design is a living room, the above steps build a 3D model of furniture including sofas, coffee tables, TV cabinets, etc., and import it into the virtual reality scene. Users can enter this scene through virtual reality equipment, freely move the perspective, and observe the design effects from different angles.
[0037] Then, the three-dimensional model of the house is input into the Bayesian network for rendering parameter analysis. First, the rendering parameters of the three-dimensional model are matched to obtain rendering parameters such as light intensity, light source position, material reflectivity, and shadow effect. These rendering parameter sets are input into the Bayesian network to construct Bayesian network nodes, and dependency data analysis is performed to determine the dependency of each Bayesian network node. By constructing the Bayesian graph structure and the maximum likelihood estimation algorithm, the node conditional probability is analyzed to obtain the conditional probability data of each node. Finally, the rendering parameter mapping is performed based on these conditional probability data to obtain the optimized target rendering parameters, which are then applied to the initial virtual scene for rendering optimization to obtain the target virtual scene.
[0038] In the optimized target virtual scene, users can start interactive operations. The Monte Carlo Tree Search (MCTS) algorithm is used to process the user's interactive behavior. First, the user's interactive actions, such as moving, rotating, scaling, and selecting objects, are extracted, and the user's initial position and perspective information are recorded. This information is input into the MCTS algorithm to match the user's operation path to obtain detailed operation path data. The operation path data is further analyzed to extract the user's stay time and operation frequency at each stop point, and user feedback data is collected in real time. All data are combined to form a user operation data set.
[0039] For example, when a user moves to a certain area in a virtual reality scene and stays there for a long time, it means that the user has a high interest in the area; frequently adjusting the position of a piece of furniture may mean that the user has made multiple attempts and received unsatisfactory feedback on the placement of the furniture.
[0040] Then, principal component analysis (PCA) is performed on the user operation data. Through PCA, the high-dimensional data is reduced in dimension and the principal component data is extracted. First, feature extraction is performed on the user operation data to obtain the operation feature set, and the covariance matrix is calculated to represent the linear correlation between different operation features. Then, the covariance matrix is decomposed by eigenvalue to obtain eigenvalues and eigenvectors, and the user operation data is matched with principal component data based on these eigenvalues and eigenvectors to obtain principal component data. Finally, the principal component data is converted into multi-dimensional evaluation index data, such as the length of stay in the area of interest, path complexity, and operation frequency interval.
[0041] For example, suppose a user performs multiple operations in multiple virtual scenes. Through PCA analysis, it is found that the user stays longer in certain specific areas and has higher path complexity. This information can be converted into evaluation indicators of the user's high interest in these areas.
[0042] Finally, the evaluation index data is input into the Markov Chain Monte Carlo (MCMC) algorithm for user behavior identification. First, the evaluation index data is matched with the starting point of the Markov chain to determine the initial state and initial behavior pattern of the user. Then, according to the initial state and transition probability, the posterior distribution analysis is performed through the MCMC algorithm to obtain the posterior distribution data of the user behavior pattern. With these data, the user's preferences and behavior patterns can be further analyzed, and a detailed evaluation report can be generated based on these analyses.
[0043] For example, by analyzing the user's operation path and dwell time in the virtual scene, it can be identified that the user prefers bright areas, prefers modern minimalist furniture layout, and prefers left-side layout when choosing furniture. Based on these behavior patterns and preference data, a detailed evaluation report can be generated, including user preference suggestions on scene layout, lighting effects, furniture style, etc., to help designers further optimize the design and improve user satisfaction.
[0044] By executing the above steps, the user's operation data is processed using the principal component analysis (PCA) technology to extract the principal component data reflecting the user's behavior characteristics. By recording and analyzing the user's operation behaviors such as moving, rotating, scaling and selecting in the virtual reality environment in detail, the relationship between these operation behaviors can be quantified to form a comprehensive operation feature set. Then, the linear correlation and important principal components of the operation feature set are further extracted using covariance matrix calculation and eigenvalue decomposition technology, thereby simplifying the data structure and improving the analysis efficiency. This process not only ensures the accuracy and efficiency of data processing, but also effectively identifies the key behavior patterns and operation habits of users in the virtual environment. On this basis, the present invention uses a Bayesian network to analyze and optimize the rendering parameters of the three-dimensional model of the house. By constructing Bayesian network nodes, defining the conditional dependency between rendering parameters, and using the maximum likelihood estimation algorithm to determine the conditional probability data of each node, it is possible to achieve accurate control and optimization of key rendering parameters such as light intensity, light source position, material reflectivity and shadow effect. This not only improves the realism and visual effects of the virtual reality scene, but also can be personalized according to the user's preferences to enhance the user's immersive experience and satisfaction. In addition, the Monte Carlo Tree Search (MCTS) algorithm is used to match and optimize the user's interactive behavior in the virtual reality environment, which can accurately identify the user's operation path and stay point, thereby deeply analyzing the user's behavior pattern and interest area. Through the detailed analysis of the operation path data, stay time and operation frequency, the user's focus and operation habits in the virtual environment can be fully understood, providing valuable reference for design optimization. In addition, the present invention also performs posterior distribution analysis on the user's behavior pattern through the Markov Chain Monte Carlo algorithm (MCMC), and further identifies and quantifies the user's behavior pattern and preference data. By inputting the user's initial state and transition probability into the MCMC algorithm, a state sequence that conforms to the user's operation habits can be generated, and the posterior distribution data of the user's behavior pattern can be obtained. These data not only reflect the user's behavior probability and preference under different operation states, but also can form a comprehensive user behavior and preference data set through pattern extraction and data fusion. Based on these data sets, a detailed evaluation report can be generated to describe the user's operation process, focus and satisfaction in the virtual environment, provide scientific optimization suggestions for designers, and significantly improve the design quality and user satisfaction.
[0045] In a specific embodiment, the process of executing step S101 may specifically include the following steps:
[0046] (1) Convert the design drawing into grayscale to obtain a grayscale drawing;
[0047] (2) performing edge detection on the grayscale drawing to obtain object contour data, wherein the object contour data includes: wall contour, door and window contour, and furniture contour;
[0048] (3) performing regional segmentation on the grayscale drawing according to the object contour data to obtain a segmented image set, wherein the segmented image set includes a plurality of segmented images;
[0049] (4) extracting shape features from each segmented image to obtain the invariant moment of each segmented image, which is used to describe the shape features of the image;
[0050] (5) extracting color features from each segmented image to obtain color features of each segmented image;
[0051] (6) extracting texture features from each segmented image using a local binary pattern algorithm to obtain texture features of each segmented image;
[0052] (7) The invariant moment of each segmented image, the color feature of each segmented image, and the texture feature of each segmented image are combined into an element feature set.
[0053] Specifically, starting from the acquired design drawings, these design drawings are converted to grayscale. The main purpose of this step is to convert the color design drawings into grayscale images to simplify the subsequent image processing process. Grayscale images only contain brightness information, but no color information, which makes edge detection and segmentation easier and more accurate. By using image processing tools, the image file of the design drawing can be read and converted into a grayscale image, resulting in an image containing only brightness values.
[0054] Next, edge detection is performed on the grayscale image to identify the outlines of objects in the drawing. Edge detection is a technique used to identify significantly changed areas in an image, which usually correspond to the boundaries of objects. The Canny edge detection algorithm can be used, which is a common and efficient edge detection method. By applying the Canny edge detection algorithm, the contour data of the object can be obtained, which includes the outline of the wall, the outline of the door and window, and the outline of the furniture. These contour data provide the basis for subsequent region segmentation.
[0055] Based on the obtained object contour data, the grayscale drawing can be segmented to separate different areas in the image. The purpose of region segmentation is to divide the image into multiple regions with independent meanings, each of which corresponds to an independent element in the design drawing. By using the contour extraction algorithm, the contours in the image can be identified, and the image can be segmented according to these contours to obtain multiple segmented images. These segmented images constitute a segmented image set, each of which represents an independent element in the design drawing, such as a wall, a window, or a piece of furniture.
[0056] For example, suppose there is a design drawing that includes the layout of a room with clearly marked walls, windows, and furniture. After grayscale conversion and edge detection, the outlines of the walls are identified and used as independent segmented images. By calculating the shape features, color features, and texture features of these segmented images, a detailed feature description of these elements can be obtained.
[0057] Next, shape features are extracted for each segmented image. Shape features are important indicators for describing the geometric shape of objects in an image, and these features can be obtained by calculating invariant moments. Invariant moments are a type of shape descriptor that is invariant to scaling, rotation, and flipping, making them very suitable for describing shape features in an image. By calculating the invariant moments of each segmented image, the shape features of these segmented images can be obtained.
[0058] After obtaining the shape features, color features are extracted for each segmented image. Color features are important indicators for describing the color distribution in an image and can be obtained by calculating a color histogram. A color histogram represents the frequency distribution of different colors in an image and can be used to describe the color characteristics of an image. Each segmented image can be converted into a hue, saturation, and brightness (HSV) color space and its color histogram is calculated to obtain the color features of each segmented image.
[0059] In addition to shape and color features, texture features are also extracted for each segmented image. Texture features are important indicators for describing the surface structure and pattern of an image, and can be obtained through the local binary pattern (LBP) algorithm. LBP is an effective texture descriptor that captures the texture information of an image by comparing the grayscale value of a pixel with the grayscale value of its neighboring pixels. By applying the LBP algorithm, the texture features of each segmented image can be obtained.
[0060] After extracting the invariant moments, color features, and texture features of each segmented image, these features need to be merged to form a complete element feature set. The element feature set contains detailed feature descriptions of all independent elements in the design drawings, providing basic data for subsequent 3D modeling and virtual reality scene generation.
[0061] For example, if the design is a living room, the above steps build a 3D model of furniture including sofas, coffee tables, TV cabinets, etc., and import it into the virtual reality scene. Users can enter this scene through virtual reality equipment, freely move the perspective, and observe the design effects from different angles. By analyzing the user's operation path and stay time in the virtual reality scene, it is identified that the user prefers bright areas, prefers modern and simple style furniture layout, and prefers the left layout when choosing furniture. Based on these behavior patterns and preference data, a detailed evaluation report can be generated, including user preference suggestions for scene layout, lighting effects, furniture style, etc., to help designers further optimize the design and improve user satisfaction.
[0062] In a specific embodiment, the process of executing step S102 may specifically include the following steps:
[0063] (1) constructing a house model according to the invariant moment of each segmented image to obtain a house sub-model set, wherein the house sub-model set includes: a wall sub-model, a floor sub-model, a ceiling sub-model, a door and window sub-model, and a furniture sub-model;
[0064] (2) Mapping the color features of each segmented image and the texture features of each segmented image to a set of house sub-models to obtain an initial three-dimensional model of the house;
[0065] (3) Correcting the model proportion of the initial three-dimensional house model to obtain a three-dimensional house model;
[0066] (4) Initializing the virtual reality scene and importing the three-dimensional model of the house into the virtual reality scene to obtain the scene to be processed;
[0067] (5) performing interactive function matching on the scene to be processed to obtain interactive function parameters, wherein the interactive function parameters include: a movement function and an angle switching function;
[0068] (6) Embed interactive functions into the scene to be processed according to the interactive function parameters to obtain an initial virtual scene.
[0069] Specifically, the house model is constructed starting from the invariant moment of each segmented image. Invariant moment is a shape descriptor that can be used to capture the geometric features of an image and has the property of being invariant to scaling, rotation, and flipping. By using invariant moments, the shape features of each segmented image can be accurately described, and these features can be used to construct a three-dimensional model of the house. Specifically, based on the invariant moment information of each segmented image, a set of sub-models of the house can be constructed. These sub-models include wall sub-models, floor sub-models, ceiling sub-models, door and window sub-models, and furniture sub-models. In this way, a preliminary skeleton structure of a three-dimensional model of the house can be obtained.
[0070] For example, suppose there is a design drawing of a room, which includes walls, floors, ceilings, doors, windows, and furniture. After image segmentation and moment invariant extraction, a detailed shape description of each element is obtained. Next, these shape descriptions are used to construct the corresponding 3D sub-models. For example, the wall sub-model can be constructed based on the wall's moment invariant information to ensure that its shape and size are consistent with the design drawing. Similarly, the sub-models of the floor, ceiling, doors, windows, and furniture can also be constructed based on their respective shape features to form a complete set of house sub-models.
[0071] Next, the color features and texture features of each segmented image need to be mapped to these house sub-models. Color features and texture features describe the surface color and texture information of each element, respectively. By mapping these features to the three-dimensional model, a more realistic and detailed initial three-dimensional model of the house can be obtained. For example, the invariant moment of the wall has determined the shape of the wall, while the color features can describe the color of the wall (for example, white or light gray), and the texture features can describe the surface texture of the wall (for example, rough brick walls or smooth plaster walls). By applying these features to the wall sub-model, a realistic wall model can be obtained. Similarly, other elements such as floors, ceilings, doors, windows, and furniture can also be mapped with color features and texture features to obtain realistic three-dimensional models.
[0072] After obtaining the initial 3D model of the house, the model is scaled to ensure that the size and proportion of the model match the actual design. This step is very important because the initial model may have some scale errors during the construction process. By scaling the model, you can ensure that the size and position of all elements are accurate, thus obtaining the final 3D model of the house.
[0073] Next, initialize the VR scene and import the 3D model of the house into the VR scene to generate the scene to be processed. The initialization of the VR scene includes setting the basic parameters of the scene, such as lighting, background, and ground. Then, import the 3D model of the house into the scene and ensure that the position and orientation of the model are correct. The purpose of this step is to create a VR environment containing the 3D model of the house so that users can interact and experience it in this environment.
[0074] For example, if you are designing a living room, you can build a 3D model of furniture including sofas, coffee tables, TV cabinets, etc. through the above steps and import it into the virtual reality scene. Users can enter this scene through virtual reality equipment, freely move the perspective, and observe the design effects from different angles.
[0075] After the scene to be processed is generated, the scene is matched with interactive functions to obtain interactive function parameters. The interactive function parameters include the user's movement function and angle switching function in the virtual reality scene. The movement function allows the user to move freely in the scene and change the observation position; the angle switching function allows the user to change the observation direction and view the house model from different angles. These functions can be realized through the built-in functions of the virtual reality platform.
[0076] Finally, according to the interactive function parameters, the interactive function is embedded in the scene to be processed to generate the initial virtual scene. Interactive function embedding includes adding movement and angle switching functions to the scene, allowing users to freely explore and operate in the virtual reality environment. For example, users can move around the scene through controllers or gestures to view the design and layout of the house from different angles. In this way, users can better understand and experience the design plan and provide valuable feedback.
[0077] In a specific embodiment, the process of executing step S103 may specifically include the following steps:
[0078] (1) matching rendering parameters of the three-dimensional model of the house to obtain a rendering parameter set, wherein the rendering parameter set includes: light intensity, light source position, material reflectivity, and shadow effect;
[0079] (2) inputting the rendering parameter set into the Bayesian network to construct the Bayesian network nodes, thereby obtaining a plurality of Bayesian network nodes;
[0080] (3) performing dependency data analysis on multiple Bayesian network nodes to obtain dependency data corresponding to each Bayesian network node;
[0081] (4) constructing a Bayesian graph structure for multiple Bayesian network nodes according to the dependency data corresponding to each Bayesian network node to obtain a Bayesian graph structure;
[0082] (5) Performing node conditional probability analysis on the Bayesian graph structure using the maximum likelihood estimation algorithm to obtain conditional probability data for each Bayesian network node;
[0083] (6) Rendering parameters are mapped according to the conditional probability data of each Bayesian network node to obtain target rendering parameters, and the initial virtual scene is rendered and optimized according to the target rendering parameters to obtain the target virtual scene.
[0084] Specifically, starting from the rendering parameter matching of the 3D model of the house, the rendering parameter set is determined. These parameters include light intensity, light source position, material reflectivity, and shadow effect. These rendering parameters are key factors affecting the visual effect of the virtual scene. For example, light intensity determines the brightness of the scene, light source position affects the distribution of light and shadow, material reflectivity affects the glossiness of the object surface, and shadow effect determines the three-dimensional sense and realism of the scene. By using 3D modeling and rendering software, these parameters can be extracted and preliminarily matched to form a rendering parameter set containing all necessary parameters.
[0085] Next, the rendering parameter set is input into the Bayesian network for node construction. A Bayesian network is a probabilistic graph model used to represent conditional dependencies between variables. In this process, each rendering parameter (such as light intensity, light source position, etc.) is regarded as a node, and through the Bayesian network, the dependency between these parameters can be established. For example, the light source position may affect the light intensity, while the material reflectivity may be associated with the shadow effect. By constructing the Bayesian network nodes, the conditional dependencies between these parameters can be preliminarily defined.
[0086] Next, dependency data analysis is performed on multiple Bayesian network nodes to determine the specific dependency data of each node. This step involves analyzing the mutual influence between various rendering parameters and quantifying this influence. For example, through experimental data or historical data, the influence of different light intensities on the shadow effect under a certain light source position can be determined. Through this dependency data analysis, the interaction between rendering parameters can be described more accurately.
[0087] Once the dependency data is determined, the Bayesian network nodes are graphed according to the data to form a complete Bayesian graph structure. The Bayesian graph structure is a directed acyclic graph in which nodes represent rendering parameters and edges represent conditional dependencies between parameters. By constructing the Bayesian graph structure, the complex relationship between rendering parameters can be intuitively represented and understood.
[0088] After having the Bayesian graph structure, the maximum likelihood estimation algorithm is used to perform node conditional probability analysis on the Bayesian graph structure. Maximum likelihood estimation is a statistical method used to estimate model parameters so that the probability of the model generating data is maximized under given data. Through maximum likelihood estimation, the conditional probability data of each Bayesian network node can be determined, which reflects the probability distribution of various rendering parameters under different conditions. For example, given the light source position and light intensity, the probability distribution of reflectivity and shadow effects of different materials can be calculated.
[0089] After obtaining the conditional probability data, the rendering parameters are mapped according to these data to determine the final target rendering parameters. The target rendering parameters are the optimal parameter combinations obtained through mapping and optimization based on Bayesian network analysis. These parameter combinations can achieve the best visual effects and realism in the virtual scene. For example, by mapping the conditional probability data, the best combination of light intensity and material reflectivity can be determined at a specific light source position, thereby achieving the best light and shadow effects and surface glossiness of the object.
[0090] Finally, the initial virtual scene is rendered and optimized according to the target rendering parameters to obtain the target virtual scene. In the rendering optimization process, the target rendering parameters are applied to the virtual scene, and the parameters such as lighting, material and shadow are adjusted to achieve the best visual effect of the scene. This step can be achieved by the rendering engine of the virtual reality platform, which will perform real-time rendering according to the target rendering parameters to generate a high-quality virtual scene.
[0091] For example, suppose a modern-style living room is being designed. The above steps determine the rendering parameters such as light intensity, light source position, material reflectivity, and shadow effect. Through Bayesian network analysis, it is found that under a certain light source position, medium light intensity and high material reflectivity can achieve the best shadow effect and surface glossiness of objects. After applying these target rendering parameters to the initial virtual scene, a virtual living room scene with sufficient light, natural shadows, and smooth and shiny object surfaces is obtained. Users can enter this scene through virtual reality devices and experience a real sense of space and visual effects, thereby providing a more realistic and intuitive design evaluation.
[0092] In a specific embodiment, the process of executing step S104 may specifically include the following steps:
[0093] (1) Extracting interactive actions from user interactive behaviors to obtain an interactive action set, where the interactive action set includes: moving, rotating, scaling, and selecting objects;
[0094] (2) extracting the user's initial position and user's perspective information corresponding to the user's interactive behavior, and inputting the user's initial position and user's perspective information into a preset Monte Carlo tree search algorithm to perform user operation path matching to obtain operation path data;
[0095] (3) extracting stop points from the operation path data to obtain multiple stop points;
[0096] (4) Match the stay time of each stay point to obtain the stay time;
[0097] (5) calculating the operation frequency based on the operation path data and the plurality of stop points to obtain the operation frequency;
[0098] (6) Collect user feedback data in real time, and merge the user feedback data, operation path data, dwell time, and operation frequency into user operation data.
[0099] Specifically, starting from the user's interactive behavior in the virtual reality scene, the interactive actions are extracted. The user's interactive actions in the virtual reality environment mainly include moving, rotating, scaling, and selecting objects. These actions reflect the user's operation methods and habits in the scene. By using the interactive data recording function of the virtual reality platform, every operation detail of the user can be captured and recorded. For example, when the user moves the position, rotates the view, scales the scene, or selects a piece of furniture through the controller, these operations will be recorded by the system to form a detailed interactive action set.
[0100] For example, suppose a user is in a virtual living room scene, first moves to the sofa area with a controller, then rotates the view to view different angles, then zooms the view to carefully examine the details of the sofa, and finally selects and drags a coffee table to a different position. All these operations will be recorded as interactive actions of moving, rotating, scaling, and selecting objects.
[0101] Next, it is necessary to extract the user's initial position and user perspective information corresponding to the user's interactive behavior. This information is the starting point and initial perspective of the user when entering the virtual scene, which can help understand the user's operation path and line of sight trajectory. By analyzing the user's initial position and perspective information, these data can be input into the preset Monte Carlo Tree Search (MCTS) algorithm for user operation path matching. MCTS is a search algorithm based on random sampling, which is widely used in fields such as decision trees and path optimization. By inputting the user's initial position and perspective information into the MCTS algorithm, the user's possible operation path in the scene can be simulated, and the optimal path can be selected for matching, thereby obtaining detailed operation path data.
[0102] The operation path data can be used to further extract the stay points. Stay points are locations where users stay for a long time in the scene, and usually reflect the user's interest in a specific area or object. For example, if a user moves to the sofa in the living room scene and stays there for a long time, it means that the user is interested in the sofa area. By analyzing the operation path data, these stay points can be identified and the specific location of the user at each stay point can be recorded.
[0103] Next, the dwell time of each dwell point is matched to obtain the dwell time of the user at each dwell point. Dwell time is an important indicator to measure the user's interest in a certain area or object. By calculating the dwell time of the user at each dwell point, we can understand the key areas and objects that the user pays attention to in the virtual scene. For example, if the user stays in the sofa area for a long time, it may mean that the user pays special attention to the design and layout of the sofa.
[0104] Then, based on the operation path data and multiple stop points, the user's operation frequency can be calculated. Operation frequency refers to the number of times a user performs certain interactive actions within a specific period of time. For example, the number of times a user rotates the view, moves the position, and selects an object within a certain period of time. By calculating the operation frequency, the user's operation habits and behavior patterns can be further analyzed.
[0105] On this basis, it is necessary to collect user feedback data in real time. User feedback data is the user's subjective evaluation of the virtual scene and design plan during the interaction process, which is usually recorded through user ratings, comments, and expressions. For example, a user may give a satisfaction rating when viewing a sofa, or provide comments when selecting furniture. By collecting this feedback data in real time, it can be combined with the operation path data, dwell time, and operation frequency to form a complete user operation data set.
[0106] For example, suppose a user takes a detailed look at the design and layout of the sofa in the virtual living room, stays at the stop point for a long time, and provides high ratings and positive comments. By combining this data, we can get the user's high attention and positive comments on the sofa area, thereby providing valuable feedback to the designer and helping to optimize the design plan.
[0107] In a specific embodiment, the process of executing step S105 may specifically include the following steps:
[0108] (1) Extracting operation features from user operation data to obtain an operation feature set;
[0109] (2) Calculating the covariance matrix of the operating feature set to obtain the covariance matrix corresponding to the operating feature set. The covariance matrix is used to represent the linear correlation between different operating features.
[0110] (3) Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors;
[0111] (4) performing principal component data matching on the user operation data according to the eigenvalues and eigenvectors to obtain principal component data;
[0112] (5) matching the principal component data with evaluation indicators to obtain an evaluation indicator set, wherein the evaluation indicator set includes: the length of stay in the area of interest, the complexity of the path, and the operation frequency interval;
[0113] (6) The principal component data is transformed according to the evaluation index set to obtain multi-dimensional evaluation index data.
[0114] Specifically, starting from the user operation data, it is necessary to extract operation features from these data. User operation data usually includes various interactive behaviors of users in the virtual reality environment, such as moving, rotating, scaling, and selecting objects. Each operation behavior can be further subdivided, such as the distance and direction of movement, the angle and axis of rotation, the scale and frequency of scaling, and the number and type of object selection. By analyzing these operation behaviors, a set of operation features can be extracted, which comprehensively describe the user's behavior pattern in the virtual environment.
[0115] For example, suppose a user performs multiple operations in a virtual living room scene, including multiple moves, rotations, zooming, and furniture selections. By recording and analyzing these operations, specific operation features can be extracted, such as the total distance the user moves, the total angle of rotation, the number and proportion of zooms, and the frequency and type of furniture selection. These operation features constitute the operation feature set, which fully reflects the user's interactive behavior in the virtual environment.
[0116] Next, we need to calculate the covariance matrix for the set of operation features. The covariance matrix is a tool in statistics that is used to describe the linear correlation between different variables. In the example, the covariance matrix is used to represent the correlation between different operation features, such as the relationship between movement distance and rotation angle. By calculating the covariance matrix, we can quantify the mutual influence between these operation features and understand the user's operation habits and preferences in the virtual environment.
[0117] Once the covariance matrix is obtained, eigenvalue decomposition is performed. Eigenvalue decomposition is a technique in linear algebra that is used to decompose the covariance matrix into eigenvalues and eigenvectors. The eigenvalues represent the importance of each principal component, and the eigenvectors represent the direction of each principal component. Through eigenvalue decomposition, a set of principal components can be extracted, which are linear combinations of the original operating characteristics that can best explain the variance of the data.
[0118] For example, suppose that several main principal components are obtained through eigenvalue decomposition, corresponding to different operation modes of users in the virtual environment. The first principal component may be mainly composed of movement distance and rotation angle, indicating the user's exploration behavior in the scene; the second principal component may be mainly composed of zoom times and object selection frequency, indicating the user's attention to a specific object. By analyzing these principal components, we can better understand the user's operation behavior.
[0119] Next, according to the eigenvalues and eigenvectors, the user operation data is matched with the principal component data to obtain the principal component data. The principal component data is the projection of the original operation features on the principal component direction, which can simplify the data structure and retain the most important information. For example, by projecting the user's operation data onto the first two principal components, a two-dimensional principal component data can be obtained, which can effectively describe the user's operation mode in the virtual environment.
[0120] Then, the principal component data is matched with evaluation indicators to obtain an evaluation indicator set. The evaluation indicator set includes key behavioral characteristics of users in the virtual environment, such as the length of time spent in the area of interest, path complexity, and operation frequency interval. The length of time spent in the area of interest indicates the time the user stays in a specific area, the path complexity indicates the complexity of the path the user moves in the scene, and the operation frequency interval indicates the frequency with which the user performs an operation within a specific time. By matching these evaluation indicators, the principal component data can be converted into specific evaluation indicator data, which is convenient for subsequent analysis and report generation.
[0121] Finally, the principal component data is transformed according to the evaluation index set to obtain multi-dimensional evaluation index data. These data fully reflect the behavioral characteristics of users in the virtual environment and can be used for in-depth analysis and optimization design. For example, by analyzing the length of time users stay in different interest areas, we can understand the degree of user attention to different areas, thereby optimizing the scene layout and design; by analyzing the complexity of the path, we can understand the user's exploration behavior in the scene, thereby improving the navigation and interaction design; by analyzing the operation frequency range, we can understand the user's operation habits, thereby optimizing the interaction interface and function settings.
[0122] For example, suppose in a virtual living room scene, the user's multi-dimensional evaluation index data is analyzed and found that the user stays in the sofa area the longest, the path complexity is the highest, and the operation frequency is high when selecting furniture. This information shows that the user is particularly interested in the sofa area and may pay more attention to the design and layout of the sofa; at the same time, the high path complexity indicates that the user is exploring different perspectives and positions and may be looking for the best layout plan; the high operation frequency indicates that the user is very concerned about the selection and placement of furniture and may be trying different combinations and combinations. Based on these analysis results, a detailed evaluation report can be generated to provide designers with valuable feedback to help them further optimize the design plan and improve user satisfaction.
[0123] In a specific embodiment, the process of executing step S106 may specifically include the following steps:
[0124] (1) Matching the evaluation index data with the starting point of the Markov chain to obtain the starting point of the Markov chain, which represents the initial state of the user and the initial behavior pattern of the user;
[0125] (2) matching the transition probability according to the starting point of the Markov chain, and based on the transition probability, inputting the starting point of the Markov chain into the Markov chain Monte Carlo algorithm for posterior distribution analysis to obtain posterior distribution data;
[0126] (3) Analyze the user preference data of the user's initial state according to the posterior distribution data to obtain the user preference data;
[0127] (4) Extracting the user's initial behavior pattern based on the posterior distribution data to obtain the user behavior pattern;
[0128] (5) Perform data fusion on user preference data according to user behavior patterns to obtain fused behavior data, and generate an evaluation report based on the fused behavior data.
[0129] Specifically, starting from the evaluation index data, these data are matched with the starting point of the Markov chain. The evaluation index data include various behavioral characteristics of users in the virtual reality environment, such as the length of stay in the area of interest, path complexity, and operation frequency range. These index data can fully reflect the user's behavior pattern in the virtual environment. By analyzing these evaluation index data, the user's initial state and initial behavior pattern can be identified and used as the starting point of the Markov chain. The starting point of the Markov chain describes the user's initial state when entering the virtual scene, such as the user's initial position, initial perspective, and initial operation preference.
[0130] For example, suppose a user performs multiple operations in a virtual living room scene, including multiple moves, rotations, zooming, and furniture selections. By recording and analyzing these operations, specific evaluation index data can be extracted, such as the time the user stays in the sofa area, the complexity of the movement path, and the frequency of furniture selection. By matching these data with the starting point of the Markov chain, the user's initial state and initial behavior pattern can be obtained. For example, when the user enters the scene, he first focuses on the sofa area and performs multiple operations around the sofa.
[0131] Next, match the transition probability according to the starting point of the Markov chain. The transition probability represents the probability of a user transitioning from one state to another, and these probabilities reflect the user's operating habits and behavior patterns in the virtual environment. These transition probabilities can be quantified by analyzing historical data or experimental data. For example, in a certain initial state, the user may have a higher probability of moving from the sofa area to the TV cabinet area, and a lower probability of moving directly from the sofa area to the kitchen area. Through this transition probability analysis, a complete Markov chain model can be constructed to describe the user's behavior path in the virtual environment.
[0132] Once the transition probability is obtained, the starting point of the Markov chain and the transition probability are input into the Markov Chain Monte Carlo algorithm (MCMC) for posterior distribution analysis. MCMC is an algorithm based on random sampling, which is used to estimate the posterior distribution of complex probability distributions. In this process, the MCMC algorithm generates a state sequence that conforms to the transition probability through continuous sampling and iteration. These state sequences represent the possible behavior paths of users in the virtual environment. Through the MCMC algorithm, the posterior distribution of user behavior patterns can be estimated and the posterior distribution data can be obtained. These data reflect the behavior probabilities and preferences of users in different states.
[0133] Based on the posterior distribution data, the user's initial state can be analyzed for user preference data. User preference data refers to the user's preference for different elements and operations in the virtual environment, such as the user's preference for a certain piece of furniture, satisfaction with a certain layout, etc. By analyzing the posterior distribution data, these user preferences can be quantified. For example, in a virtual living room scene, the user may be more interested in modern-style furniture and more satisfied with an open layout. Through this preference data analysis, detailed preference information of the user in the virtual environment can be obtained.
[0134] At the same time, based on the posterior distribution data, the user's initial behavior pattern can also be extracted. The behavior pattern is a pattern that describes the user's operating habits and path selection in the virtual environment. By analyzing these patterns, the user's typical operation path and habits can be identified. For example, the user's behavior pattern in the virtual living room scene may include first looking at the sofa area, then moving to the TV cabinet area, and then choosing to adjust the position of the coffee table. Through this pattern extraction, the user's operation process and preferences in the virtual environment can be understood.
[0135] Next, the user preference data is fused according to the user behavior pattern to obtain fused behavior data. Fusion behavior data combines the user's behavior pattern and preference data to form a data set that comprehensively describes the user's behavior and preferences in the virtual environment. For example, by combining the user's stay time in different areas, operation frequency, and rating of furniture, detailed behavior and preference information of the user in the virtual living room scene can be obtained. Through this data fusion, the user's operating habits and preferences in the virtual environment can be fully understood, providing valuable reference for design optimization.
[0136] Finally, an evaluation report is generated based on the fused behavior data. The evaluation report is a comprehensive analysis of the user's behavior and preferences in the virtual environment, including detailed descriptions of behavior patterns, preference data statistics, and optimization suggestions. For example, by analyzing the user's operation path and dwell time in the virtual living room scene, we can understand the user's attention to different areas, thereby optimizing the scene layout and design; by analyzing the user's selection frequency and rating of furniture, we can understand the user's furniture style preferences, thereby improving furniture design and layout; by analyzing the user's operation frequency and path complexity, we can understand the user's operation habits, thereby optimizing the interactive interface and function settings. Based on these analysis results, a detailed evaluation report can be generated to provide designers with valuable feedback to help them further optimize the design plan and improve user satisfaction.
[0137] The embodiment of the present invention also provides an indoor design achievement evaluation system based on virtual reality technology. Figure 2 As shown, the interior design achievement evaluation system based on virtual reality technology specifically includes:
[0138] The segmentation module 201 is used to perform image segmentation on the acquired design drawing to obtain a segmented image set, and perform feature extraction on each segmented image in the segmented image set to obtain an element feature set;
[0139] A construction module 202 is used to construct a three-dimensional model of the house according to the element feature set, and import the three-dimensional model of the house into a preset virtual reality scene to generate an initial virtual scene;
[0140] The analysis module 203 is used to input the three-dimensional model of the house into the Bayesian network to perform rendering parameter analysis to obtain target rendering parameters, and to optimize the rendering of the initial virtual scene according to the target rendering parameters to obtain a target virtual scene;
[0141] The processing module 204 is used to perform Monte Carlo tree search processing on the user interaction behavior in the target virtual scene to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, residence time, operation frequency and user feedback data;
[0142] The conversion module 205 is used to perform principal component analysis on the user operation data to obtain principal component data, and convert the principal component data into multi-dimensional evaluation index data;
[0143] The identification module 206 is used to input the evaluation index data into the Markov Chain Monte Carlo algorithm to perform user behavior identification, obtain user behavior patterns and user preference data, and generate an evaluation report based on the user behavior patterns and the user preference data.
[0144] Through the collaborative work of the above modules, the principal component analysis (PCA) technology is used to process the user's operation data and extract the principal component data reflecting the user's behavior characteristics. By recording and analyzing the user's operation behaviors such as movement, rotation, scaling and selection in the virtual reality environment in detail, the relationship between these operation behaviors can be quantified to form a comprehensive operation feature set. Then, the linear correlation and important principal components of the operation feature set are further extracted by using covariance matrix calculation and eigenvalue decomposition technology, thereby simplifying the data structure and improving the analysis efficiency. This process not only ensures the accuracy and efficiency of data processing, but also effectively identifies the key behavior patterns and operation habits of users in the virtual environment. On this basis, the present invention uses a Bayesian network to analyze and optimize the rendering parameters of the three-dimensional model of the house. By constructing Bayesian network nodes, defining the conditional dependency between rendering parameters, and using the maximum likelihood estimation algorithm to determine the conditional probability data of each node, it is possible to achieve accurate control and optimization of key rendering parameters such as light intensity, light source position, material reflectivity and shadow effect. This not only improves the realism and visual effects of the virtual reality scene, but also can be personalized according to the user's preferences to enhance the user's immersive experience and satisfaction. In addition, the Monte Carlo Tree Search (MCTS) algorithm is used to match and optimize the user's interactive behavior in the virtual reality environment, which can accurately identify the user's operation path and stay point, thereby deeply analyzing the user's behavior pattern and interest area. Through the detailed analysis of the operation path data, stay time and operation frequency, the user's focus and operation habits in the virtual environment can be fully understood, providing valuable reference for design optimization. In addition, the present invention also performs posterior distribution analysis on the user's behavior pattern through the Markov Chain Monte Carlo algorithm (MCMC), and further identifies and quantifies the user's behavior pattern and preference data. By inputting the user's initial state and transition probability into the MCMC algorithm, a state sequence that conforms to the user's operation habits can be generated, and the posterior distribution data of the user's behavior pattern can be obtained. These data not only reflect the user's behavior probability and preference under different operation states, but also can form a comprehensive user behavior and preference data set through pattern extraction and data fusion. Based on these data sets, a detailed evaluation report can be generated to describe the user's operation process, focus and satisfaction in the virtual environment, provide scientific optimization suggestions for designers, and significantly improve the design quality and user satisfaction.
[0145] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.
Claims
1. A method for evaluating interior design results based on virtual reality technology. It is characterized in that include: Performing image segmentation on the acquired design drawing to obtain a segmented image set, and performing feature extraction on each segmented image in the segmented image set to obtain an element feature set; Constructing a three-dimensional model of the house according to the element feature set, and importing the three-dimensional model of the house into a preset virtual reality scene to generate an initial virtual scene; Inputting the three-dimensional model of the house into a Bayesian network to perform rendering parameter analysis to obtain target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain a target virtual scene; Performing Monte Carlo tree search processing on the user interaction behavior in the target virtual scene to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, residence time, operation frequency and user feedback data; Performing principal component analysis on the user operation data to obtain principal component data, and converting the principal component data into multi-dimensional evaluation index data; The evaluation index data is input into a Markov chain Monte Carlo algorithm to perform user behavior recognition, obtain user behavior patterns and user preference data, and generate an evaluation report based on the user behavior patterns and the user preference data.
2. According to the method for evaluating interior design results based on virtual reality technology in claim 1, It is characterized in that The step of performing image segmentation on the acquired design drawing to obtain a segmented image set, and performing feature extraction on each segmented image in the segmented image set to obtain an element feature set comprises: Performing grayscale conversion on the design drawing to obtain a grayscale drawing; Performing edge detection on the grayscale drawing to obtain object contour data, wherein the object contour data includes: wall contours, door and window contours, and furniture contours; Performing regional segmentation on the grayscale drawing according to the object contour data to obtain the segmented image set, wherein the segmented image set includes a plurality of segmented images; Perform shape feature extraction on each segmented image to obtain an invariant moment of each segmented image, wherein the invariant moment is used to describe the shape feature of the image; Perform color feature extraction on each segmented image to obtain the color feature of each segmented image; The texture features of each segmented image are extracted by using the local binary pattern algorithm to obtain the texture features of each segmented image; The invariant moment of each segmented image, the color feature of each segmented image, and the texture feature of each segmented image are combined into the element feature set.
3. The indoor design achievement evaluation method based on virtual reality technology according to claim 2, It is characterized in that The step of constructing a three-dimensional model of the house according to the element feature set, and importing the three-dimensional model of the house into a preset virtual reality scene to generate an initial virtual scene includes: Constructing a house model according to the invariant moment of each segmented image to obtain a house sub-model set, wherein the house sub-model set includes: a wall sub-model, a floor sub-model, a ceiling sub-model, a door and window sub-model, and a furniture sub-model; Mapping the color features of each segmented image and the texture features of each segmented image to the house sub-model set to obtain an initial house three-dimensional model; Correcting the model proportion of the initial three-dimensional house model to obtain the three-dimensional house model; Initializing a virtual reality scene, and importing the three-dimensional model of the house into the virtual reality scene to obtain a scene to be processed; Performing interactive function matching on the scene to be processed to obtain interactive function parameters, wherein the interactive function parameters include: a movement function and an angle switching function; The interactive function is embedded into the scene to be processed according to the interactive function parameters to obtain the initial virtual scene.
4. The method for evaluating interior design results based on virtual reality technology according to claim 1, It is characterized in that The step of inputting the three-dimensional model of the house into a Bayesian network to perform rendering parameter analysis to obtain target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain the target virtual scene comprises: Performing rendering parameter matching on the three-dimensional model of the house to obtain a rendering parameter set, wherein the rendering parameter set includes: light intensity, light source position, material reflectivity, and shadow effect; Inputting the rendering parameter set into the Bayesian network to construct a Bayesian network node to obtain a plurality of Bayesian network nodes; Performing dependency data analysis on the multiple Bayesian network nodes to obtain dependency data corresponding to each Bayesian network node; According to the dependency data corresponding to each Bayesian network node, a Bayesian graph structure is constructed for multiple Bayesian network nodes to obtain a Bayesian graph structure; Performing node conditional probability analysis on the Bayesian graph structure by using a maximum likelihood estimation algorithm to obtain conditional probability data for each Bayesian network node; Rendering parameter mapping is performed according to the conditional probability data of each Bayesian network node to obtain the target rendering parameters, and rendering optimization is performed on the initial virtual scene according to the target rendering parameters to obtain the target virtual scene.
5. The indoor design achievement evaluation method based on virtual reality technology according to claim 1, It is characterized in that The step of performing a Monte Carlo tree search on the user interaction behavior in the target virtual scene to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, residence time, operation frequency and user feedback data, comprises: Extracting interactive actions from the user interactive behavior to obtain an interactive action set, wherein the interactive action set includes: moving, rotating, scaling, and selecting an object; Extracting the user initial position and user perspective information corresponding to the user interaction behavior, and inputting the user initial position and the user perspective information into a preset Monte Carlo tree search algorithm to perform user operation path matching to obtain operation path data; Extracting stop points from the operation path data to obtain a plurality of stop points; Matching the stay time of each stay point respectively to obtain the stay time; Calculate the operation frequency based on the operation path data and the plurality of stay points to obtain the operation frequency; User feedback data is collected in real time, and the user feedback data, operation path data, dwell time and operation frequency are combined into the user operation data.
6. The indoor design achievement evaluation method based on virtual reality technology according to claim 1, It is characterized in that The step of performing principal component analysis on the user operation data to obtain principal component data and converting the principal component data into multi-dimensional evaluation index data includes: Extracting operation features from the user operation data to obtain an operation feature set; Performing covariance matrix calculation on the operation feature set to obtain a covariance matrix corresponding to the operation feature set, wherein the covariance matrix is used to represent linear correlations between different operation features; Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors; Performing principal component data matching on the user operation data according to the eigenvalue and the eigenvector to obtain the principal component data; Performing evaluation index matching on the principal component data to obtain an evaluation index set, wherein the evaluation index set includes: the length of stay in the area of interest, the complexity of the path, and the operation frequency interval; The principal component data is converted according to the evaluation index set to obtain the multi-dimensional evaluation index data.
7. The method for evaluating interior design results based on virtual reality technology according to claim 6, It is characterized in that The step of inputting the evaluation index data into a Markov chain Monte Carlo algorithm to perform user behavior recognition, obtain a user behavior pattern and user preference data, and generate an evaluation report according to the user behavior pattern and the user preference data includes: Performing Markov chain starting point matching on the evaluation index data to obtain a Markov chain starting point, wherein the Markov chain starting point represents an initial state of the user and an initial behavior pattern of the user; Matching the transition probability according to the starting point of the Markov chain, and based on the transition probability, inputting the starting point of the Markov chain into the Markov chain Monte Carlo algorithm to perform posterior distribution analysis to obtain posterior distribution data; Performing user preference data analysis on the user's initial state according to the posterior distribution data to obtain user preference data; Performing pattern extraction on the user's initial behavior pattern according to the posterior distribution data to obtain the user behavior pattern; The user preference data is fused according to the user behavior pattern to obtain fused behavior data, and the evaluation report is generated according to the fused behavior data.
8. An indoor design achievement evaluation system based on virtual reality technology, used to execute the indoor design achievement evaluation method based on virtual reality technology as claimed in any one of claims 1 to 7, It is characterized in that include: A segmentation module is used to perform image segmentation on the acquired design drawing to obtain a segmented image set, and perform feature extraction on each segmented image in the segmented image set to obtain an element feature set; A construction module, used to construct a three-dimensional model of the house according to the element feature set, and import the three-dimensional model of the house into a preset virtual reality scene to generate an initial virtual scene; An analysis module, used for inputting the three-dimensional model of the house into a Bayesian network to perform rendering parameter analysis to obtain target rendering parameters, and performing rendering optimization on the initial virtual scene according to the target rendering parameters to obtain a target virtual scene; A processing module, used for performing Monte Carlo tree search processing on the user interaction behavior in the target virtual scene to obtain user operation data of the user in the target virtual scene, wherein the user operation data includes: operation path data, residence time, operation frequency and user feedback data; A conversion module, used to perform principal component analysis on the user operation data to obtain principal component data, and convert the principal component data into multi-dimensional evaluation index data; The identification module is used to input the evaluation index data into the Markov chain Monte Carlo algorithm to perform user behavior identification, obtain user behavior patterns and user preference data, and generate an evaluation report based on the user behavior patterns and the user preference data.
