Search

CN-122023738-A - Landscape design dynamic display method and system based on Virtual Reality (VR)

CN122023738ACN 122023738 ACN122023738 ACN 122023738ACN-122023738-A

Abstract

The invention discloses a landscape design dynamic display method and system based on Virtual Reality (VR), which comprises the steps of collecting real-time environment data and plant growth period information through a multi-source sensor, processing through a particle filtering algorithm to obtain a preliminary time-space aligned dynamic data set, training an interaction relation among key characteristic parameters through a neural network model according to the preliminary time-space aligned dynamic data set, determining a coupling coefficient matrix, obtaining a logic rule of environment interaction from the coupling coefficient matrix to obtain an interaction map containing time-space consistency, updating a dynamic state value of each node through an iteration simulation device based on a node connection structure in the interaction map to obtain an optimized scene evolution sequence, determining a collaborative driving model if a part with inconsistent data exists in the optimized scene evolution sequence, generating a virtual display rendering data stream according to the collaborative driving model, and generating a continuous landscape dynamic result.

Inventors

  • ZHANG LILI
  • SU BEILE
  • SU YUNQIAO

Assignees

  • 山东建筑大学

Dates

Publication Date
20260512
Application Date
20260127

Claims (9)

  1. 1. A Virtual Reality (VR) based landscape design dynamic display method, comprising: Collecting real-time environment data and plant growth period information through a multi-source sensor, and processing by adopting a particle filtering algorithm to obtain a preliminary space-time aligned dynamic data set; According to the preliminary time-space aligned dynamic data set, extracting key characteristic parameters by utilizing characteristic extraction equipment, training the mutual influence relation among the key characteristic parameters by adopting a neural network model, and determining a coupling coefficient matrix; acquiring a logic rule of environment interaction from the coupling coefficient matrix, and constructing a node connection structure under a unified frame by adopting a graph neural network to obtain an interaction map containing space-time consistency; Based on the node connection structure in the interaction map, updating the dynamic state value of each node through iterative simulation equipment, judging whether the updated state value accords with the constraint of a physical rule, and obtaining an optimized scene evolution sequence; If the optimized scene evolution sequence has a part with inconsistent data, acquiring the supplementary information from the multi-source heterogeneous data source again and fusing the supplementary information into the sequence to determine a collaborative driving model; And generating a virtual display rendering data stream according to the collaborative driving model, and processing the rendering data stream by adopting a real-time rendering engine to generate a continuous landscape dynamic result.
  2. 2. The Virtual Reality (VR) based landscape design dynamic display method of claim 1, wherein the acquiring real-time environmental data and plant growth cycle information by the multi-source sensor, the processing by the particle filter algorithm, the obtaining the preliminary space-time aligned dynamic dataset comprises: collecting real-time environment data and plant growth period information through a multi-source sensor; Extracting outline information of plant canopy and stalk in the plant growth period information by adopting a semantic segmentation network, and constructing an initial multidimensional data record containing temperature, humidity, illumination intensity and canopy area by combining environmental data; Calculating the information entropy of the current data point according to the historical fluctuation range of each dimension in the initial multidimensional data record, and marking the corresponding data point as a potential abnormal point to be processed if the information entropy exceeds a preset confidence interval; Carrying out state estimation and weight update on the data marked as potential abnormal points according to a state transition equation of the plant growth model by adopting a particle filtering algorithm to obtain a modified smooth data sequence; And establishing a unified space-time axis by taking the global time service signal as a reference, and resampling the smooth data sequences of different sources by a cubic spline interpolation method to obtain the preliminary space-time aligned dynamic data set.
  3. 3. The Virtual Reality (VR) based landscape design dynamic display method of claim 1, wherein extracting key feature parameters from the preliminary spatiotemporal alignment dynamic dataset using feature extraction equipment, training the relationship of interactions between the key feature parameters using a neural network model, determining a coupling coefficient matrix comprises: Extracting key coupling coefficients from the preliminary space-time aligned dynamic data set to generate an inter-feature correlation strength table; Constructing a characteristic interaction network diagram according to the association strength table, and determining a dynamic coupling mode of wind strength and illumination change; If the complexity of the dynamic coupling mode exceeds a preset threshold, grouping by a cluster analysis method to obtain a simplified coupling mode set; optimizing parameters of the neural network model of the multi-layer perceptron through simplifying the coupling mode set to obtain an optimized model; adopting an optimized model to predict the interaction trend of wind intensity and illumination change of a new input data set, and obtaining a predicted interaction result; And extracting significant trend characteristics from the predicted interaction result, and determining the coupling coefficient matrix.
  4. 4. A Virtual Reality (VR) based landscape design dynamic presentation method in accordance with claim 3, wherein extracting salient trend features from the predicted interaction results, determining the coupling coefficient matrix comprises: If a certain element value in the coupling coefficient matrix exceeds a preset threshold value, acquiring an element set exceeding the threshold value through threshold value judgment, and determining a dynamic element list to be adjusted; according to the dynamic element list to be adjusted, grouping the dynamic elements by adopting a K-means clustering algorithm to obtain element grouping results; according to the grouping result, sorting each group of dynamic elements through a priority sorting algorithm to obtain a sorted element sequence; According to the ordered element sequence, the positions of corresponding elements in the coupling coefficient matrix are adjusted to obtain a preliminarily adjusted matrix; If element values in the preliminarily adjusted matrix still exceed a preset threshold value, locally adjusting the matrix through a gradient descent algorithm to obtain an optimized matrix; And verifying whether all the element values meet the preset threshold condition according to the optimized matrix to obtain a final coupling coefficient matrix.
  5. 5. The Virtual Reality (VR) -based landscape design dynamic display method of claim 1, wherein obtaining the logic rules of the environment interactions from the coupling coefficient matrix, and constructing the node connection structure under the unified frame using the graph neural network, the obtaining the interaction map including the space-time consistency comprises: Obtaining a strong coupling relation list according to the coupling coefficient matrix; analyzing co-occurrence modes of the entity pairs in the time dimension by adopting an Apriori algorithm according to the strong coupling relation list and the associated entity state data thereof, and generating a primary interaction rule set for describing the association influence among the entities; abstracting each independent entity into a graph node by analyzing the logic relationship in the preliminary interaction rule set, converting the pointing relationship among the entities in the rule into directed connection among the nodes, and obtaining an initial topological graph; According to each graph node in the initial topological graph, corresponding space position information and historical state time sequence data are acquired, the space position information is used as a static attribute of the node, and the time sequence data is used as a dynamic attribute of the node, so that a space-time node set is obtained; Inputting a topological graph containing a space-time node set into a graph annotation meaning network model to obtain new node embedding containing dynamic evolution information; And calculating the similarity between the new node embeddings by adopting cosine similarity, and generating a weighted directed graph, wherein the nodes represent environmental entities, and the weights of the sides represent interaction strength, so as to obtain an interaction graph containing space-time consistency.
  6. 6. The Virtual Reality (VR) -based landscape design dynamic display method of claim 1, wherein updating the dynamic state value of each node through the iterative simulation device based on the node connection structure in the interaction map, and determining whether the updated state value meets the constraint of the physical rule, and obtaining the optimized scene evolution sequence includes: acquiring an initial node set, connectivity and structure information of an interaction map, and endowing each node with an initial state value to form an initial scene map; extracting a constraint set related to the physical quantity of the node from a pre-established rule base according to the initial scene graph, and constructing a discriminant for state evolution; calculating potential change rate of each node state value according to the connectivity among the nodes by adopting an iterative simulation method, and generating a scene graph to be tested containing temporary dynamic values; Carrying out compliance test on the temporary dynamic value of each node in the scene graph to be tested according to the discriminant, judging the state as effective if the temporary dynamic value accords with the constraint set, and judging the state as ineffective if the temporary dynamic value does not accord with the constraint set; updating the evolution tree according to the judging result of the effective state and the ineffective state, taking the scene graph to be tested containing the effective state as a new branch node of the evolution tree, and discarding the scene graph to be tested containing the ineffective state; When the evolution tree reaches a preset simulation depth, extracting a plurality of complete state evolution paths from the root node to all leaf nodes to form a candidate sequence set; And evaluating the cumulative prize value of each evolution path in the candidate sequence set by adopting a Q-learning algorithm, and determining the evolution path with the highest cumulative prize value as a final optimized scene evolution sequence.
  7. 7. The Virtual Reality (VR) based landscape design dynamic presentation method of claim 1, wherein if there is a data inconsistent portion in the optimized scene evolution sequence, re-acquiring the supplemental information from the multi-source heterogeneous data source and fusing the supplemental information into the sequence, determining the collaborative driving model comprises: Scanning continuous time nodes in the optimized scene evolution sequence by adopting a sequential logic checking method, judging whether data breakpoints or logic conflicts exist according to a preset logic rule, and obtaining position information of inconsistent fragments; according to the position information of the inconsistent fragments, matching entities and relations in a predefined knowledge graph, and directionally acquiring supplementary data streams related to corresponding fragment time stamps and event types from a plurality of heterogeneous supplementary sources; the data set to be fused is obtained by carrying out structural analysis and semantic alignment on the supplementary data stream and converting the supplementary data stream into a standardized information unit with the same data protocol as the original scene evolution sequence; calculating the contribution weight of each standardized information unit in the data set to be fused to inconsistent fragments by adopting a fusion strategy based on an attention mechanism, and generating an initial calibration sequence in a weighted summation mode; If the initial calibration sequence passes the consistency check in the time sequence logic check, the corresponding sequence is determined to be a final calibration sequence, and if the initial calibration sequence does not pass the consistency check, the weight parameters of the attention mechanism are adjusted and fusion is re-executed until the final calibration sequence is obtained; And according to the final calibration sequence, model training is carried out by adopting a long-term and short-term memory network, and the collaborative driving model is obtained by learning the state transition and the collaborative relationship in the sequence.
  8. 8. The Virtual Reality (VR) based landscape design dynamics presentation method of claim 1, wherein generating a virtual presented rendered data stream from the collaborative driving model, processing the rendered data stream with a real-time rendering engine to produce a continuous landscape dynamics result comprises: Acquiring a driving instruction set output by the collaborative driving model, and analyzing the driving instruction set one by one through a grammar analyzer to obtain a structured scene parameter set containing entity positions, postures and environment states; According to the entity identifier in the structured scene parameter set, corresponding three-dimensional geometric data and surface texture map information are retrieved from a digital asset library, and an original asset set required for constructing a scene is obtained; binding the original asset set with a predefined physical field parameter by adopting a physical simulation module, and obtaining an enhanced scene description with the dynamic attribute added through calculation; Uniformly encoding geometrical vertex, texture map index and physical field vector in the enhanced scene description by a serialization processor to generate a continuous rendering data stream containing time stamps; receiving the continuous rendering data stream by adopting a real-time rendering engine, and unpacking in a rendering pipeline to reconstruct a layered scene tree if the data frame integrity of the data stream passes the verification; And according to the node information of the layered scene tree and the associated illumination field data, performing coloring calculation and post-processing, generating a single frame image, and adding the single frame image into an output frame sequence to obtain a final high-fidelity simulation view.
  9. 9. The Virtual Reality (VR) -based landscape design dynamic display system implemented by the method of any one of claims 1-8, comprising a data acquisition module, a feature extraction module, a map construction module, a scene evolution sequence optimization module, a model construction module, and a result output module; The data acquisition module is used for acquiring real-time environment data and plant growth period information through a multisource sensor, and processing the real-time environment data and the plant growth period information through a particle filtering algorithm to obtain a preliminary time-space aligned dynamic data set; the feature extraction module is used for extracting key feature parameters by using feature extraction equipment according to the preliminary space-time aligned dynamic data set, training the mutual influence relationship among the key feature parameters by using a neural network model, and determining a coupling coefficient matrix; the map construction module is used for acquiring logic rules of environment interaction from the coupling coefficient matrix, and constructing a node connection structure under a unified frame by adopting a graph neural network to obtain an interaction map containing space-time consistency; The scene evolution sequence optimizing module is used for updating the dynamic state value of each node through iterative simulation equipment based on the node connection structure in the interaction map, judging whether the updated state value accords with the constraint of a physical rule or not, and obtaining an optimized scene evolution sequence; the model construction module is used for acquiring the supplementary information again from the multi-source heterogeneous data source and fusing the supplementary information into the sequence if the optimized scene evolution sequence has a part with inconsistent data, and determining a collaborative driving model; And the result output module is used for generating a virtual display rendering data stream according to the collaborative driving model, and processing the rendering data stream by adopting a real-time rendering engine to generate a continuous landscape dynamic result.

Description

Landscape design dynamic display method and system based on Virtual Reality (VR) Technical Field The invention belongs to the technical field of virtual reality, and particularly relates to a landscape design dynamic display method and system based on Virtual Reality (VR). Background The landscape design is used as a key link for constructing a harmonious living environment, and has the core value of foreseeing and shaping the vitality and aesthetic feeling of the space in time circulation. With the development of digital technology, design expression forms have evolved from traditional drawings to three-dimensional visualization, however, the existing virtual display method still has a deep bottleneck in expressing the "life feeling" of landscapes. These approaches are not just lack of dynamic effects, but rather are more limited in that each dynamic element in the landscape is treated as an isolated, independently programmable module, ignoring the complex coupling relationships of interrelations, interactions between elements in the real world, resulting in a lack of inherent logic uniformity in the simulated scene. The processing mode of the fracturing introduces a core technical problem of ensuring the space-time consistency of the multi-source heterogeneous dynamic data. The reality of a landscape scene is derived from the fact that all changes inside the landscape scene follow the unified time and physical laws. For example, the growth state of a plant is not only related to its own biological cycle, but is also affected in real time by environmental data such as light intensity, temperature and humidity. If the season phase change of the plants in the system cannot be accurately synchronized with the simulation data of the environmental climate, or if the wind power data and the swing amplitude of the tree branches and leaves lack accurate physical logic correlation, a false sense of 'apparent' can be generated. The distortion of the dynamic effect is based on the fact that a bottom framework capable of effectively fusing and cooperatively driving dynamic information with different sources and different time scales is not established, so that each dynamic effect is 'dynamic' but not 'alive'. Specifically, when a designer tries to evaluate the design scheme of a water-side area, he needs to not only see the spray pattern of the fountain, but also understand the drift range of the water flow at different wind speeds and the fine influence of water mist on the peripheral microclimate (such as local humidity), and observe the long-term change of stone texture of the water side under sunlight and water vapor erosion. If these dynamic changes are run based on respective independent preset scripts instead of a unified environmental physics engine drive, the evolution of the whole scene will lose confidence and cannot provide reliable basis for design optimization. Therefore, how to construct a virtual landscape system which can make discrete and multi-source dynamic data (such as plant growth period, environmental climate change and physical feedback of materials) perform space-time coordination and convert the discrete and multi-source dynamic data into a unified and logical self-consistent virtual landscape system, so that a designer and a user can intuitively feel and verify real evolution of landscape elements and complex interaction of environments in different time scales, and the virtual landscape system becomes a key problem for pushing the landscape design industry to enter the high-fidelity dynamic simulation era. Disclosure of Invention In order to solve the technical problems, the invention provides a landscape design dynamic display method and a landscape design dynamic display system based on Virtual Reality (VR), wherein the method can construct a virtual landscape system which can perform space-time coordination on discrete and multi-source dynamic data (such as plant growth period, environmental climate change and physical feedback of materials) and convert the discrete and multi-source dynamic data into uniform and logic self-consistent, so that a designer and a user can intuitively feel and verify real evolution of landscape elements and complex interaction of environment in different time scales. In order to achieve the above object, the present invention provides a landscape design dynamic display method based on Virtual Reality (VR), comprising: Collecting real-time environment data and plant growth period information through a multi-source sensor, and processing by adopting a particle filtering algorithm to obtain a preliminary space-time aligned dynamic data set; According to the preliminary time-space aligned dynamic data set, extracting key characteristic parameters by utilizing characteristic extraction equipment, training the mutual influence relation among the key characteristic parameters by adopting a neural network model, and determining a coupling coefficient matrix; acquiri