Live-action three-dimensional geographic entity and space planning display system

Abstract

The invention belongs to the technical field of image data processing, and discloses a live-action three-dimensional geographic entity and space planning display system. The system comprises a first processing module which constructs a three-dimensional space tile data set corresponding to N precision hierarchies based on multi-density sampling real scenic spot cloud data; the second processing module is used for carrying out operation behavior feature extraction based on a zooming operation behavior in the three-dimensional display system to obtain operation behavior feature data; the behavior diagnosis module is used for inputting the operation behavior characteristic data into an operation behavior diagnosis model to obtain a scaling behavior mode; the three-dimensional display system carries out intelligent display based on the operation behavior characteristic data, the zooming behavior mode and the three-dimensional space tile data set corresponding to the N precision levels; according to the method, an operation behavior feature extraction and zooming behavior pattern recognition mechanism is introduced, and intelligent self-adaptive scheduling display is performed, so that the problems of zooming dislocation and visual jump are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and more particularly to a real-scene three-dimensional geographic entity and space planning display system. Background Art

[0002] In existing technology, when displaying and analyzing real-world 3D geographic entity data and planning model data at multiple scales, the real-world point cloud data is typically downsampled to a uniform granularity to generate a single-level 3D spatial tile set. Data sets of different resolutions are then switched based on a fixed zoom ratio to accommodate changes in viewport. Furthermore, to enhance the interactive zooming experience, some systems have introduced zoom rate detection, adjusting tile loading priorities based on zoom speed to mitigate rendering delays caused by continuous zooming.

[0003] However, existing technologies still have significant shortcomings in handling scenarios with large-scale scaling changes, rapid scaling, or dramatic changes in spatial extent. Due to the lack of unified standards for hierarchical division and error control between real-world point clouds and spatial planning models, the scaling process is prone to problems such as feature dislocation, model overlap, or visual jumps, affecting display continuity and spatial logic consistency. Furthermore, existing technologies are unable to fully extract scaling behavior characteristics and dynamically adapt tile scheduling strategies, resulting in rigid resource scheduling and delayed loading responses when the system is rapidly zoomed in or out. This can easily lead to large-scale tile instantaneous loading bottlenecks or detail loading delays, making it difficult to balance scaling fluidity with the final visual detail presentation.

[0004] Therefore, there is an urgent need to propose a display system that can dynamically identify scaling modes based on operational behavior characteristics and intelligently optimize loading scheduling, so as to improve the overall display continuity, system response performance and spatial expression accuracy of three-dimensional scenes under multi-scale scaling. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned objectives, the present invention provides the following technical solution: a real-scene three-dimensional geographic entity and spatial planning display system, comprising:

[0006] The first processing module constructs a 3D spatial tile data set corresponding to N levels of accuracy based on multi-density sampled real-world point cloud data;

[0007] The second processing module extracts operation behavior features based on the zoom operation behavior in the three-dimensional display system to obtain operation behavior feature data;

[0008] A behavior diagnosis module inputs the operation behavior characteristic data into the operation behavior diagnosis model to obtain a scaling behavior pattern;

[0009] Intelligent display module, the three-dimensional display system performs intelligent display based on the operation behavior feature data, scaling behavior pattern and the three-dimensional space tile data set corresponding to N precision levels.

[0010] Furthermore, the zoom behavior mode includes slow zooming and fast zooming; the slow zooming includes slow zooming in and slow zooming out; the fast zooming includes fast zooming in and fast zooming out; and the method for intelligent display based on operation behavior feature data, zoom behavior mode, and three-dimensional space tile data sets corresponding to N precision levels includes:

[0011] If the zoom behavior mode is slow zoom, the 3D display system performs intelligent display based on the slow zoom;

[0012] If the zoom behavior pattern is rapid zoom, the operation behavior feature data and the zoom behavior pattern are input into the accuracy level prediction model to obtain the prediction accuracy level; the three-dimensional display system performs intelligent display for rapid zoom.

[0013] Furthermore, the method for intelligently displaying the slow zooming of the 3D display system includes:

[0014] Obtaining a second zoom ratio value corresponding to the completion of the zoom operation from the operation behavior feature data, matching the second zoom ratio value with a pre-built zoom ratio-precision level matching table to obtain a corresponding precision level, which is recorded as a slow zoom target precision level; the zoom ratio-precision level matching table includes zoom ratio values ​​and corresponding precision levels;

[0015] Selecting a 3D space tile data set corresponding to a slowly zoomed target precision level from the 3D space tile data sets corresponding to the N precision levels, and recording the result as a target 3D space tile data set;

[0016] The three-dimensional presentation system performs a rendering presentation operation based on the target three-dimensional space tile data set.

[0017] Furthermore, the method for intelligent display of the 3D display system for rapid zooming includes:

[0018] Obtaining a first zoom ratio value corresponding to the start of the zoom operation from the operation behavior feature data, matching the first zoom ratio value with a pre-built zoom ratio-precision level matching table to obtain a corresponding precision level, which is recorded as a fast zoom target precision level;

[0019] A clustering algorithm is used to compare each 3D spatial tile data in the 3D spatial tile data set of the fast scaling target accuracy level with all 3D spatial tile data in the 3D spatial tile data set of the prediction accuracy level to obtain corresponding tile similarities; for each 3D spatial tile data in the 3D spatial tile data set of the fast scaling target accuracy level, the 3D spatial tile data with the highest tile similarity is selected from the 3D spatial tile data set of the prediction accuracy level to form a one-to-one corresponding 3D spatial tile data pair;

[0020] A tile similarity threshold 1 and a tile similarity threshold 2 are preset; the tile similarity threshold 1 is greater than the tile similarity threshold 2;

[0021] The three-dimensional space tile data pairs whose tile similarity is greater than or equal to the tile similarity threshold 1 are divided into the preload set; the three-dimensional space tile data pairs whose tile similarity is less than the tile similarity threshold 1 and greater than or equal to the tile similarity threshold 2 are divided into the progressive supplement set; the three-dimensional space tile data pairs whose tile similarity is less than the tile similarity threshold 2 are divided into the unload set;

[0022] The three-dimensional display system performs loading, delayed replenishment and unloading rendering operations based on the preload set, the progressive replenishment set and the unloaded set respectively.

[0023] Furthermore, the method for obtaining the operation behavior characteristic data includes:

[0024] When a zoom operation is detected, a system timestamp and a zoom ratio value at the start of the zoom operation are obtained, which are recorded as a first system timestamp and a first zoom ratio value respectively;

[0025] When it is detected that the zoom operation is completed, a system timestamp and a zoom ratio value at the time of completion of the zoom operation are obtained, and the values ​​are recorded as a second system timestamp and a second zoom ratio value respectively;

[0026] Subtracting the second scaling ratio value from the first scaling ratio value to obtain a scaling change amplitude; if the scaling change amplitude is greater than zero, setting the scaling direction indicator to an enlargement indicator; if the scaling change amplitude is less than zero, setting the scaling direction indicator to a reduction indicator;

[0027] Subtracting the second system timestamp from the first system timestamp to obtain a scaling change time interval;

[0028] The absolute value of the scaling change amplitude is taken and divided by the scaling change time interval to obtain the scaling change rate;

[0029] The second zoom ratio value, the zoom change amplitude, the zoom direction identifier, the zoom change time interval and the zoom change rate are constructed into operation behavior characteristic data.

[0030] Furthermore, a method for constructing a three-dimensional space tile data set corresponding to N precision levels based on multi-density sampled real-world point cloud data includes:

[0031] Obtaining real-time point cloud data, which contains three-dimensional spatial coordinates;

[0032] Construct the real-time point cloud data into a bounding box of the real-time point cloud data;

[0033] Inputting the pre-collected multi-scale precision control requirements of the 3D display system into the precision parameter setting model to obtain precision setting parameters, wherein the precision setting parameters include N precision levels and corresponding stratified sampling granularities; the multi-scale precision control requirements include a minimum magnification ratio, a maximum magnification ratio, a minimum reduction ratio, and a maximum reduction ratio;

[0034] Based on N precision levels and corresponding hierarchical sampling granularity, the bounding box of the actual point cloud data is downsampled to obtain a downsampled point cloud data set corresponding to N precision levels;

[0035] Based on the hierarchical sampling granularity of N precision levels and the downsampled point cloud data set, spatial slicing processing is performed to obtain a three-dimensional spatial tile data set corresponding to N precision levels.

[0036] Furthermore, the method of constructing the on-site point cloud data into a bounding box of the on-site point cloud data includes:

[0037] Obtain all three-dimensional coordinate points in the real-world point cloud data, and based on all three-dimensional coordinate points, determine the boundary values ​​of the real-world point cloud data in three dimensions, including the maximum and minimum coordinate values ​​in the X-axis direction, the maximum and minimum coordinate values ​​in the Y-axis direction, and the maximum and minimum coordinate values ​​in the Z-axis direction;

[0038] The length of the bounding box of the real-world point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the X-axis direction, the width of the bounding box of the real-world point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the Y-axis direction, and the height of the bounding box of the real-world point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the Z-axis direction. The center coordinates of the bounding box of the point cloud data are obtained based on the length, width, and height of the bounding box of the real-world point cloud data.

[0039] The length, width, height and center coordinates of the bounding box of the real-time point cloud data are used as the parameters of the bounding box of the real-time point cloud data to construct a geometric boundary volume that surrounds the entire spatial range of the real-time point cloud data, that is, the bounding box of the real-time point cloud data is obtained.

[0040] Furthermore, the method of downsampling the bounding box of the field point cloud data based on N precision levels and corresponding hierarchical sampling granularity to obtain a downsampled point cloud data set corresponding to the N precision levels includes:

[0041] S100: let the initial value of n be 1, and the value range of n be 1 to N;

[0042] S101: Using the stratified sampling granularity of the nth precision level as the precision, divide along the X-axis Point cloud data grid units, divided along the Y axis Point cloud data grid units are divided along the Z axis point cloud data grid units, that is, a total of Total n Point cloud data grid unit; Indicates rounding up; CD bwh is the length of the bounding box of the field point cloud data, KD bwh is the width of the bounding box of the real-time point cloud data, GD bwh is the height of the bounding box of the field point cloud data, LD n is the stratified sampling granularity of the nth precision level;

[0043] S102: Total n point cloud data grid units are constructed into a downsampled point cloud data set of the nth accuracy level; the downsampled point cloud data set maintains the same spatial boundary range as the bounding box of the actual point cloud data;

[0044] S103: Let n=n+1. If n is less than or equal to N, continue executing S101 to S102; if n is greater than N, end the current process.

[0045] Furthermore, a method for performing spatial slicing processing based on the hierarchical sampling granularity and downsampled point cloud data set of N precision levels to obtain a three-dimensional spatial tile data set corresponding to N precision levels includes:

[0046] S200: Let the initial value of n be 1, and the value range of n be 1 to N;

[0047] S201: Get the total number of point cloud data grid cells in the nth downsampled point cloud data set n Total n Each point cloud data grid unit constructs the corresponding downsampled point cloud data sub-bounding box; the downsampled point cloud data sub-bounding box is the three-dimensional space tile, and the Total n Three-dimensional space tiles;

[0048] S202: Total n Extract the length, width, height and center coordinates of the three-dimensional space tiles to get the Total nThe 3D tile length, 3D tile width, 3D tile height and 3D tile center coordinates corresponding to the 3D tile; β times the stratified sampling granularity of the nth precision level is used as the 3D tile error parameter, where β is a constant greater than 0;

[0049] S203: Total n The 3D tile error parameters, 3D tile length, 3D tile width, 3D tile height and 3D tile center coordinates corresponding to each 3D tile are constructed into Total n The three-dimensional tile index information corresponding to the three-dimensional tile; Total n The 3D tiles and the corresponding 3D tile index information are constructed into Total n 3D space tile data, Total n The three-dimensional space tile data is constructed into a three-dimensional space tile data set;

[0050] S204: Let n=n+1. If n is less than or equal to N, continue to execute S201 to S103; if n is greater than N, end the current process.

[0051] Furthermore, the training method of the operational behavior diagnosis model includes:

[0052] Pre-constructing an operation behavior diagnostic data set, the operation behavior diagnostic data set including XW groups of operation behavior diagnostic data and scaling behavior patterns corresponding to the XW groups of operation behavior diagnostic data, where XW is a positive integer greater than 0, and the operation behavior diagnostic data includes operation behavior feature data;

[0053] The operational behavior diagnosis dataset is divided into a training set and a validation set. The training set is used for parameter learning, and the validation set is used to evaluate the generalization ability of the model.

[0054] A support vector machine model is used as the operation behavior diagnosis model. The operation behavior feature data is converted into feature vector input, nonlinear features are extracted through the hidden layer, and the scaling behavior pattern corresponding to the maximum probability is output in the output layer as the final prediction result; the training goal is to minimize the cross entropy loss function, and training is stopped when the accuracy of the verification set reaches the preset threshold.

[0055] Compared with the existing technology, the technical effects and advantages of the real-scene three-dimensional geographic entity and spatial planning display system of the present invention are as follows:

[0056] The real-life three-dimensional geographic entity and spatial planning display system provided by the present invention, based on a multi-density sampling mechanism, constructs a three-dimensional spatial tile data set covering different levels of precision for real-life point cloud data. By combining the matching relationship between scaling ratio and precision level, and through downsampling and spatial slicing processing, it effectively realizes the structured organization and index optimization of multi-scale three-dimensional data.

[0057] On this basis, this application introduces an operation behavior feature extraction and zoom behavior pattern recognition mechanism, which can judge in real time whether the user's zoom behavior belongs to fast zoom in, slow zoom in, fast zoom out or slow zoom out mode based on feature data such as zoom change amplitude, zoom direction, and zoom rate, and then dynamically adapt the corresponding three-dimensional space tile scheduling strategy. Among them, for slow zooming, the system accurately switches to the three-dimensional space tile set of the target accuracy level based on the zoom ratio when the operation is completed, ensuring display continuity and system stability; for fast zooming, based on the zoom starting ratio and the predicted accuracy level, a clustering algorithm is used to match tile similarity, divide the preload set, progressive supplement set and unload set, and execute loading, delayed supplement and resource recovery in order according to priority, effectively alleviating the loading mutation and performance bottleneck caused by fast zooming.

[0058] Through the above mechanism, this application not only solves the problems of scaling dislocation, visual jump and analysis distortion caused by inconsistent hierarchical definitions and inconsistent error control between the real-world scenic spot cloud and the planning model in the existing technology, but also significantly improves the rendering continuity, spatial logic consistency and resource scheduling efficiency of the three-dimensional display system under multi-scale scaling. At the same time, it takes into account the comprehensive requirements of refinement, real-time and large-scale data processing in urban three-dimensional reconstruction, spatial planning display and digital twin applications, and has significant technical effects with clear system structure, superior response performance and strong scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Schematic diagram of a real-scene three-dimensional geographic entity and spatial planning display system according to Example 1 of the present invention;

[0060] Figure 2 This is a flow chart of a method for displaying real-scene three-dimensional geographic entities and spatial planning according to embodiment 2 of the present invention;

[0061] Figure 3 A flow chart of a method for intelligently displaying a three-dimensional display system for slow zooming;

[0062] Figure 4 A flow chart of a method for intelligent display of a 3D display system for rapid scaling;

[0063] Figure 5 A schematic diagram of the system architecture for displaying real-life 3D geographic entities and spatial planning;

[0064] Figure 6 A mind map of real-life 3D geographic entities and spatial planning display methods. DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present invention will be described in detail, clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. It should be noted that the specific embodiments described below are only used to better illustrate and describe the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and essence of the present invention, those skilled in the art may modify, adjust or make equivalent replacements based on the contents disclosed in the present invention, and these should all be regarded as the scope of protection of the present invention.

[0066] Example 1

[0067] See also Figure 1 As shown, this embodiment discloses a real-life three-dimensional geographic entity and space planning display system, including a first processing module, a second processing module, a behavior diagnosis module and an intelligent display module. Each module realizes data transmission through wired and / or wireless connections.

[0068] The first processing module constructs a three-dimensional space tile data set corresponding to N accuracy levels based on the multi-density sampled real-world point cloud data.

[0069] The method for constructing a three-dimensional space tile data set corresponding to N levels of accuracy based on multi-density sampled real-world point cloud data includes:

[0070] Obtain real-time point cloud data, which includes three-dimensional spatial coordinates (X, Y, Z); the real-time point cloud data format can be LAS, LAZ, PCD or BIN, etc.

[0071] Construct the real-time point cloud data into a bounding box of the real-time point cloud data;

[0072] Inputting the pre-collected multi-scale precision control requirements of the 3D display system into the precision parameter setting model to obtain precision setting parameters, wherein the precision setting parameters include N precision levels and corresponding stratified sampling granularities; the multi-scale precision control requirements include a minimum magnification ratio, a maximum magnification ratio, a minimum reduction ratio, and a maximum reduction ratio;

[0073] Based on N precision levels and corresponding hierarchical sampling granularity, the bounding box of the actual point cloud data is downsampled to obtain a downsampled point cloud data set corresponding to N precision levels;

[0074] Based on the hierarchical sampling granularity of N precision levels and the downsampled point cloud data set, spatial slicing processing is performed to obtain a three-dimensional spatial tile data set corresponding to N precision levels.

[0075] The method of constructing the field point cloud data into the field point cloud data bounding box includes:

[0076] Obtain all three-dimensional coordinate points in the real-world point cloud data, and based on all three-dimensional coordinate points, determine the boundary values ​​of the real-world point cloud data in three dimensions, including the maximum and minimum coordinate values ​​in the X-axis direction, the maximum and minimum coordinate values ​​in the Y-axis direction, and the maximum and minimum coordinate values ​​in the Z-axis direction;

[0077] The length of the bounding box of the real-time point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the X-axis direction, the width of the bounding box of the real-time point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the Y-axis direction, and the height of the bounding box of the real-time point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the Z-axis direction.

[0078] The center coordinates of the bounding box of the real-world point cloud data are constructed by the mean of the maximum and minimum coordinate values ​​in the X-axis direction, the mean of the maximum and minimum coordinate values ​​in the Y-axis direction, and the mean of the maximum and minimum coordinate values ​​in the Z-axis direction;

[0079] The length, width, height and center coordinates of the real-time point cloud data bounding box are used as the parameters of the real-time point cloud data bounding box to construct a geometric boundary volume that surrounds the entire real-time point cloud data spatial range, that is, the real-time point cloud data bounding box is obtained.

[0080] The method for obtaining the center coordinates of the bounding box of the real-world point cloud data includes:

[0081]

[0082] Among them, ZB center is the center coordinate of the bounding box of the real-time point cloud data, X max is the maximum coordinate value in the X-axis direction, X min is the minimum coordinate value in the X-axis direction, Y max is the maximum coordinate value in the Y-axis direction, Y min is the minimum coordinate value in the Y-axis direction, Z max is the maximum coordinate value in the Z-axis direction, Z min The minimum coordinate value in the Z-axis direction.

[0083] The training method of the precision parameter setting model includes:

[0084] Pre-constructing a precision parameter setting data set, the precision parameter setting data set including JD group precision parameter setting data and precision setting parameters corresponding to the JD group precision parameter setting data, JD is a positive integer greater than 0, the precision parameter setting data including multi-scale precision control requirements; the multi-scale precision control requirements include a minimum magnification ratio, a maximum magnification ratio, a minimum reduction ratio, and a maximum reduction ratio; dividing the precision parameter setting data set into a precision parameter setting data training set and a precision parameter setting data validation set, wherein the precision parameter setting data training set is used for parameter learning of a precision parameter setting model, and the precision parameter setting data validation set is used for real-time evaluation of the generalization ability of the precision parameter setting model;

[0085] During the training process of the precision parameter setting model, the precision parameter setting model is trained using a Bayesian network model, the precision parameter setting data is converted into a feature vector as input, the nonlinear features in the data are extracted through the hidden layer, and finally the probability distribution of the precision setting parameters is generated in the output layer using the softmax activation function, and the precision setting parameters corresponding to the maximum probability are output as the final prediction result; the training process aims to minimize the cross entropy loss function, and at the same time introduces an early stopping strategy to monitor the performance of the precision parameter setting data verification set. When the prediction accuracy on the precision parameter setting data verification set reaches the preset accuracy, it is determined that the precision parameter setting model has converged and the training stops immediately.

[0086] Examples of accuracy setting parameters are shown in Table 1:

[0087] Table 1 Precision setting parameters

[0088]

[0089]

[0090] The method for downsampling the bounding box of the field point cloud data based on N precision levels and corresponding hierarchical sampling granularity to obtain a downsampled point cloud data set corresponding to the N precision levels includes:

[0091] S100: let the initial value of n be 1, and the value range of n be 1 to N;

[0092] S101: Using the stratified sampling granularity of the nth precision level as the precision, divide along the X-axis Point cloud data grid units, divided along the Y axis Point cloud data grid units are divided along the Z axis Point cloud data grid unit; Indicates rounding up; CD bwh is the length of the bounding box of the field point cloud data, KD bwh is the width of the bounding box of the real-time point cloud data, GDbwh is the height of the bounding box of the field point cloud data, LD n is the hierarchical sampling granularity of the nth level of accuracy; that is, the bounding box of the field point cloud data is divided into voxels to obtain the Total n Point cloud data grid units, Total n for and Multiply to get;

[0093] It should be noted that

[0094] S102: Total n point cloud data grid units are constructed into a downsampled point cloud data set of the nth accuracy level; the downsampled point cloud data set maintains the same spatial boundary range as the bounding box of the actual point cloud data;

[0095] S103: Let n=n+1. If n is less than or equal to N, continue executing S101 to S102; if n is greater than N, end the current process.

[0096] A method for performing spatial slicing processing based on a hierarchical sampling granularity and a downsampled point cloud data set at N precision levels to obtain a three-dimensional spatial tile data set corresponding to N precision levels includes:

[0097] S200: Let the initial value of n be 1, and the value range of n be 1 to N;

[0098] S201: Get the total number of point cloud data grid cells in the nth downsampled point cloud data set n Total n Each point cloud data grid unit constructs the corresponding downsampled point cloud data sub-bounding box; the downsampled point cloud data sub-bounding box is the three-dimensional space tile, and the Total n Three-dimensional space tiles;

[0099] S202: Total n Extract the length, width, height and center coordinates of the three-dimensional space tiles to get the Total n The 3D tile length, 3D tile width, 3D tile height, and 3D tile center coordinates corresponding to each 3D tile; β times the stratified sampling granularity of the nth precision level is used as the 3D tile error parameter, where β is a constant greater than 0; for example, in this embodiment, the value of β can be set to 2;

[0100] S203: Total nThe 3D tile error parameters, 3D tile length, 3D tile width, 3D tile height and 3D tile center coordinates corresponding to each 3D tile are constructed into Total n The three-dimensional tile index information corresponding to the three-dimensional tile; Total n The 3D tiles and the corresponding 3D tile index information are constructed into Total n 3D space tile data, Total n The three-dimensional space tile data is constructed into a three-dimensional space tile data set;

[0101] S204: Let n=n+1. If n is less than or equal to N, continue to execute S201 to S103; if n is greater than N, end the current process.

[0102] It should be noted that this application generates N levels of three-dimensional spatial tiles and their three-dimensional spatial tile index information based on multi-density sampled real-world point cloud data, which can structure the original unstructured high-density real-world point cloud data according to the multi-scale precision control requirements of the three-dimensional display system, and automatically generates precision levels and layered sampling granularity through the precision parameter setting model to achieve multi-level downsampling of point cloud data within the spatial range, and constructs three-dimensional spatial tiles and corresponding index information with spatial boundary description and error control parameters based on the downsampling results, thereby forming a precision layered point cloud data structure that can be called on demand by the three-dimensional visualization system; this method can not only effectively improve the organizational efficiency and spatial schedulability of point cloud data, support multi-level continuous display from a global perspective to local details, but also reduce system resource consumption, improve rendering performance and interactive response speed, and is particularly suitable for application scenarios with high requirements for point cloud precision control and visualization performance, such as urban three-dimensional reconstruction, spatial planning display, digital twin platform, etc., with significant advantages such as clear structure, high processing efficiency, strong system adaptability and good scalability.

[0103] It should be noted that although the existing technology can perform multi-level simplification and error unification processing on the real-life point cloud data and spatial planning models respectively, and generate three-dimensional spatial tiles that meet the spatial alignment requirements, due to the lack of a coordinated mechanism in the front-end three-dimensional engine for the loading scheduling logic, error interpretation standards, etc. of different data sources in the actual scaling operation, it will still cause the real-life point cloud data and spatial planning models to have geometric jumps, loading lags, display dislocations and scaling level imbalances during the scaling process, thereby affecting the continuity and accuracy of the final three-dimensional scene display, resulting in inconsistent spatial analysis results, and making it difficult to meet practical engineering needs.

[0104] The second processing module extracts operation behavior features based on the zoom operation behavior in the three-dimensional display system to obtain operation behavior feature data.

[0105] The method for acquiring the operation behavior characteristic data includes:

[0106] When a zoom operation is detected, a system timestamp and a zoom ratio value at the start of the zoom operation are obtained, which are recorded as a first system timestamp and a first zoom ratio value respectively;

[0107] When it is detected that the zoom operation is completed, a system timestamp and a zoom ratio value at the time of completion of the zoom operation are obtained, and the values ​​are recorded as a second system timestamp and a second zoom ratio value respectively;

[0108] Subtracting the second scaling ratio value from the first scaling ratio value to obtain a scaling change amplitude; if the scaling change amplitude is greater than zero, setting the scaling direction indicator to an enlargement indicator; if the scaling change amplitude is less than zero, setting the scaling direction indicator to a reduction indicator;

[0109] It should be noted that the zoom direction flag can also be represented by a numerical value. For example, the zoom direction flag can be set to -1 and 1, where -1 represents zooming out and 1 represents zooming in.

[0110] Subtracting the second system timestamp from the first system timestamp to obtain a scaling change time interval;

[0111] The absolute value of the scaling change amplitude is taken and divided by the scaling change time interval to obtain the scaling change rate;

[0112] The second zoom ratio value, the zoom change amplitude, the zoom direction identifier, the zoom change time interval and the zoom change rate are constructed into operation behavior characteristic data.

[0113] The method for calculating the scaling variation includes:

[0114] BHFD = DEBL-DYBL;

[0115] Wherein, BHFD is the scaling change amplitude, DEBL is the second scaling ratio value, and DYBL is the first scaling ratio value.

[0116] The method for calculating the scaling change time interval includes:

[0117] SJJG = DESJ-DYSJ;

[0118] Among them, SJJG is the scaling change time interval, DESJ is the second system timestamp, and DYSJ is the first system timestamp.

[0119] The method for calculating the scaling change rate includes:

[0120]

[0121] Where SFSL is the scaling rate of change.

[0122] The behavior diagnosis module inputs the operation behavior characteristic data into the operation behavior diagnosis model to obtain a zoom behavior pattern; the zoom behavior pattern includes slow zooming and fast zooming; the slow zooming includes slow zooming in and slow zooming out; the fast zooming includes fast zooming in and fast zooming out.

[0123] The training method of the operational behavior diagnosis model includes:

[0124] Pre-constructing an operation behavior diagnostic dataset, wherein the operation behavior diagnostic dataset includes XW groups of operation behavior diagnostic data and scaled behavior patterns corresponding to the XW groups of operation behavior diagnostic data, where XW is a positive integer greater than 0, and the operation behavior diagnostic data includes operation behavior feature data; dividing the operation behavior diagnostic dataset into an operation behavior diagnostic data training set and an operation behavior diagnostic data validation set, wherein the operation behavior diagnostic data training set is used for parameter learning of an operation behavior diagnostic model, and the operation behavior diagnostic data validation set is used for real-time evaluation of the generalization ability of the operation behavior diagnostic model;

[0125] During the training process of the operational behavior diagnosis model, the operational behavior diagnosis model adopts a support vector machine model, converts the operational behavior diagnosis data into a feature vector as input, extracts nonlinear features in the data through the hidden layer, and finally uses the softmax activation function in the output layer to generate the probability distribution of the scaling behavior pattern, and outputs the scaling behavior pattern corresponding to the maximum probability as the final prediction result; the training process aims to minimize the cross-entropy loss function, and at the same time introduces an early stopping strategy to monitor the performance of the operational behavior diagnosis data verification set. When the prediction accuracy on the operational behavior diagnosis data verification set reaches the preset accuracy, the operational behavior diagnosis model is determined to have converged and the training stops immediately.

[0126] It should be noted that by collecting operational behavior feature data such as the zoom change amplitude, zoom direction identifier, zoom change time interval, and zoom change rate, the action characteristics and change trends of the three-dimensional display system during the zoom operation can be effectively characterized. By training the machine learning model based on the above-mentioned operational behavior feature data, the machine learning model can learn the characteristic distribution characteristics corresponding to different zoom behavior patterns, so that in actual applications, it can accurately identify whether the current zoom behavior belongs to a fast zoom in, slow zoom in, fast zoom out, or slow zoom out mode based on the real-time extracted operational behavior feature data. By introducing a machine learning model for zoom behavior pattern recognition, it can effectively avoid the problems of low accuracy and poor adaptability in complex interactive environments caused by traditional judgments based on a single threshold rule, and further improve the accuracy and robustness of zoom behavior recognition. In addition, based on the identified zoom behavior pattern, it can provide an accurate basis for the subsequent dynamic adjustment of the loading strategy of the three-dimensional space tiles, thereby achieving resource scheduling optimization and interactive experience improvement during the zoom process.

[0127] Intelligent display module, the three-dimensional display system performs intelligent display based on the operation behavior feature data, scaling behavior pattern and the three-dimensional space tile data set corresponding to N precision levels.

[0128] The method for intelligently displaying the three-dimensional space tile data set based on the operation behavior feature data, the scaling behavior pattern, and N precision levels includes:

[0129] If the zoom behavior mode is slow zoom, the 3D display system performs intelligent display based on the slow zoom;

[0130] If the zoom behavior pattern is rapid zoom, the operation behavior feature data and the zoom behavior pattern are input into the accuracy level prediction model to obtain the prediction accuracy level; the three-dimensional display system performs intelligent display for rapid zoom.

[0131] like Figure 3 As shown, the method for intelligently displaying the slow zooming of the 3D display system includes:

[0132] Obtaining a second zoom ratio value corresponding to the completion of the zoom operation from the operation behavior feature data, matching the second zoom ratio value with a pre-built zoom ratio-precision level matching table to obtain a corresponding precision level, which is recorded as a slow zoom target precision level; the zoom ratio-precision level matching table includes zoom ratio values ​​and corresponding precision levels;

[0133] Selecting a 3D space tile data set corresponding to a slowly zoomed target precision level from the 3D space tile data sets corresponding to the N precision levels, and recording the result as a target 3D space tile data set;

[0134] The three-dimensional display system performs rendering and display operations based on the target three-dimensional space tile data set, realizes dynamic switching of display accuracy according to the zoom ratio, and improves visual continuity and system response efficiency during the zoom interaction process.

[0135] like Figure 4 As shown, the method for intelligent display of the 3D display system for rapid scaling includes:

[0136] Obtaining a first zoom ratio value corresponding to the start of the zoom operation from the operation behavior feature data, matching the first zoom ratio value with a pre-built zoom ratio-precision level matching table to obtain a corresponding precision level, which is recorded as a fast zoom target precision level;

[0137] A clustering algorithm is used to compare each 3D spatial tile data in the 3D spatial tile data set of the fast scaling target accuracy level with all 3D spatial tile data in the 3D spatial tile data set of the prediction accuracy level to obtain corresponding tile similarities; for each 3D spatial tile data in the 3D spatial tile data set of the fast scaling target accuracy level, the 3D spatial tile data with the highest tile similarity is selected from the 3D spatial tile data set of the prediction accuracy level to form a one-to-one corresponding 3D spatial tile data pair;

[0138] It should be noted that clustering algorithms that can be used to obtain similarity include K-Means clustering algorithm, K-Medoids clustering algorithm and DBSCAN density clustering algorithm.

[0139] A tile similarity threshold 1 and a tile similarity threshold 2 are preset; the tile similarity threshold 1 is greater than the tile similarity threshold 2;

[0140] The three-dimensional space tile data pairs whose tile similarity is greater than or equal to the tile similarity threshold 1 are divided into the preload set; the three-dimensional space tile data pairs whose tile similarity is less than the tile similarity threshold 1 and greater than or equal to the tile similarity threshold 2 are divided into the progressive supplement set; the three-dimensional space tile data pairs whose tile similarity is less than the tile similarity threshold 2 are divided into the unload set;

[0141] The 3D display system performs loading, delayed replenishment and unloading rendering operations based on the preloaded set, progressive replenishment set and unloaded set, respectively, to achieve smooth display and resource optimization scheduling during rapid scaling.

[0142] It should be noted that tile similarity measures the degree of feature similarity between three-dimensional tile data at different precision levels. The tile similarity range is defined as [0, 1]. A tile similarity value closer to 1 indicates closer spatial positions, dimensional parameters, and error characteristics between the tile data. A tile similarity value closer to 0 indicates greater feature differences between the tile data. In practical applications, both the first and second tile similarity thresholds are set within the range of [0, 1]. For example, tile similarity threshold 1 can be set to 0.7, and tile similarity threshold 2 can be set to 0.4.

[0143] The preloaded collection prioritizes loading target tile data with highly similar local features to the original viewport when rapidly zooming in or out. By quickly loading tiles from the preloaded collection, the new viewport is instantly filled after the zoom operation, ensuring scene continuity and smooth rendering, avoiding visible holes or noticeable lag.

[0144] Progressive Supplementation is used to gradually load tile data that has a certain degree of similarity to the original local features but does not meet the immediate loading requirements after the initial rapid scaling is completed. The existence of progressive supplementation can refine display details during the stable scaling phase, improving visual quality while balancing scaling responsiveness and final image accuracy.

[0145] The offload set is used to identify tile data that has no valid similarity with the current predicted viewport and is redundantly occupying resources during rapid zooming. By promptly offloading tile data from the offload set, system memory, cache, and bandwidth resources can be effectively freed up, improving overall system performance, reducing unnecessary resource waste, and ensuring efficient and stable operation of the 3D display system.

[0146] An example of the scaling-precision level matching table is shown in Table 2:

[0147] Table 2 Scaling ratio-precision level matching table

[0148]

[0149]

[0150] It should be noted that the overall architecture diagram of the real-scene 3D geographic entity and spatial planning display system provided by this application is as follows Figure 5 As shown in the figure, it mainly includes three hierarchical structures, namely hardware device layer, data processing layer and intelligent interaction layer.

[0151] The hardware device layer includes: 3D data acquisition equipment, such as the RIEGL VZ-400 lidar, for acquiring real-world point cloud data; computing equipment, such as the NVIDIA DGX A100 server, for high-performance point cloud processing, cluster analysis, and intelligent scheduling calculations; and display terminals, such as the Planar 8K 3D stereo display, for high-resolution 3D visualization rendering.

[0152] The data processing layer includes a first processing module and a second processing module: the first processing module is used to perform spatial tile construction and precision level division, and to build a three-dimensional spatial tile structure suitable for multi-scale display based on a multi-density sampling mechanism; the second processing module is used to perform tile screening and organizational reconstruction based on the scaling behavior pattern, and dynamically generate a loadable tile set based on the predicted scaling results and precision requirements.

[0153] The intelligent interaction layer includes a behavior diagnosis module and an intelligent display module: the behavior diagnosis module is used to call the behavior diagnosis model based on user operation behavior feature data to determine whether the zoom behavior is fast zoom in, slow zoom in, fast zoom out, or slow zoom out; the intelligent display module is used to perform dynamic adjustment and progressive tile loading strategies based on the recognition results to achieve smooth control and optimal resource utilization of the system under different zoom behaviors.

[0154] The real-scene three-dimensional geographic entity and spatial planning display method provided by this application is as follows Figure 6 As shown in the figure, the whole process includes four stages: data preprocessing, interactive feature extraction, intelligent decision-making and dynamic rendering. Among them, the data preprocessing stage realizes spatial structure extraction and multi-scale reconstruction of the original real-world point cloud data through point cloud bounding box construction, multi-level downsampling and timestamp recording; the interactive feature extraction stage extracts the zoom rate and direction identification based on the user zoom operation to characterize the interactive behavior characteristics; the intelligent decision-making stage intelligently identifies the zoom behavior and plans resource scheduling through behavioral pattern classification and resource prediction model; the dynamic rendering stage generates preloaded sets, progressive supplementary sets and tile unloading sets according to the decision results, and dynamically executes the loading, refinement and release operations of three-dimensional space tiles, thereby realizing the continuous display, resource optimization and precision matching of three-dimensional scenes during the zoom interaction process.

[0155] Example 2

[0156] See also Figure 2 As shown, this embodiment provides a method for displaying real-life three-dimensional geographic entities and spatial planning, including:

[0157] Construct a 3D spatial tile data set corresponding to N levels of accuracy based on multi-density sampled real-world cloud data;

[0158] Extracting operation behavior features based on the zoom operation behavior in the three-dimensional display system to obtain operation behavior feature data;

[0159] Inputting the operational behavior characteristic data into the operational behavior diagnosis model to obtain a scaling behavior pattern;

[0160] The 3D display system performs intelligent display based on operation behavior feature data, scaling behavior patterns, and 3D space tile data sets corresponding to N precision levels.

[0161] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

[0162] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Realistic 3D geographic entity and spatial planning display system, characterized by: include: The first processing module constructs a 3D spatial tile data set corresponding to N levels of accuracy based on multi-density sampled real-world point cloud data; The second processing module extracts operation behavior features based on the zoom operation behavior in the three-dimensional display system to obtain operation behavior feature data; A behavior diagnosis module inputs the operation behavior characteristic data into the operation behavior diagnosis model to obtain a scaling behavior pattern; Intelligent display module, the three-dimensional display system performs intelligent display based on the operation behavior feature data, scaling behavior pattern and the three-dimensional space tile data set corresponding to N precision levels.

2. The real-scene three-dimensional geographic entity and spatial planning display system according to claim 1 is characterized in that: The zoom behavior mode includes slow zooming and fast zooming; the slow zooming includes slow zooming in and slow zooming out; the fast zooming includes fast zooming in and fast zooming out; the method for intelligent display based on operation behavior feature data, zoom behavior mode and three-dimensional space tile data sets corresponding to N precision levels includes: If the zoom behavior mode is slow zoom, the 3D display system performs intelligent display based on the slow zoom; If the zoom behavior pattern is rapid zoom, the operation behavior feature data and the zoom behavior pattern are input into the accuracy level prediction model to obtain the prediction accuracy level; the three-dimensional display system performs intelligent display for rapid zoom.

3. The real-scene three-dimensional geographic entity and spatial planning display system according to claim 2, characterized in that: The intelligent display methods of the 3D display system for slow zooming include: Obtaining a second zoom ratio value corresponding to the completion of the zoom operation from the operation behavior feature data, matching the second zoom ratio value with a pre-built zoom ratio-precision level matching table to obtain a corresponding precision level, which is recorded as a slow zoom target precision level; the zoom ratio-precision level matching table includes zoom ratio values ​​and corresponding precision levels; Selecting a 3D space tile data set corresponding to a slowly zoomed target precision level from the 3D space tile data sets corresponding to the N precision levels, and recording the result as a target 3D space tile data set; The three-dimensional presentation system performs a rendering presentation operation based on the target three-dimensional space tile data set.

4. The real-scene three-dimensional geographic entity and spatial planning display system according to claim 2, characterized in that: The intelligent display method of the 3D display system for rapid scaling includes: Obtaining a first zoom ratio value corresponding to the start of the zoom operation from the operation behavior feature data, matching the first zoom ratio value with a pre-built zoom ratio-precision level matching table to obtain a corresponding precision level, which is recorded as a fast zoom target precision level; A clustering algorithm is used to compare each 3D spatial tile data in the 3D spatial tile data set of the fast scaling target accuracy level with all 3D spatial tile data in the 3D spatial tile data set of the prediction accuracy level to obtain corresponding tile similarities; for each 3D spatial tile data in the 3D spatial tile data set of the fast scaling target accuracy level, the 3D spatial tile data with the highest tile similarity is selected from the 3D spatial tile data set of the prediction accuracy level to form a one-to-one corresponding 3D spatial tile data pair; A tile similarity threshold 1 and a tile similarity threshold 2 are preset; the tile similarity threshold 1 is greater than the tile similarity threshold 2; The three-dimensional space tile data pairs whose tile similarity is greater than or equal to the tile similarity threshold 1 are divided into the preload set; the three-dimensional space tile data pairs whose tile similarity is less than the tile similarity threshold 1 and greater than or equal to the tile similarity threshold 2 are divided into the progressive supplement set; the three-dimensional space tile data pairs whose tile similarity is less than the tile similarity threshold 2 are divided into the unload set; The three-dimensional display system performs loading, delayed replenishment and unloading rendering operations based on the preload set, the progressive replenishment set and the unloaded set respectively.

5. The real-scene three-dimensional geographic entity and space planning display system according to claim 1 is characterized in that: The method for acquiring the operation behavior characteristic data includes: When a zoom operation is detected, a system timestamp and a zoom ratio value at the start of the zoom operation are obtained, which are recorded as a first system timestamp and a first zoom ratio value respectively; When it is detected that the zoom operation is completed, a system timestamp and a zoom ratio value at the time of completion of the zoom operation are obtained, and the values ​​are recorded as a second system timestamp and a second zoom ratio value respectively; Subtracting the second scaling ratio value from the first scaling ratio value to obtain a scaling change amplitude; if the scaling change amplitude is greater than zero, setting the scaling direction indicator to an enlargement indicator; if the scaling change amplitude is less than zero, setting the scaling direction indicator to a reduction indicator; Subtracting the second system timestamp from the first system timestamp to obtain a scaling change time interval; The absolute value of the scaling change amplitude is taken and divided by the scaling change time interval to obtain the scaling change rate; The second zoom ratio value, the zoom change amplitude, the zoom direction identifier, the zoom change time interval and the zoom change rate are constructed into operation behavior characteristic data.

6. The real-scene three-dimensional geographic entity and spatial planning display system according to claim 1 is characterized in that: The method for constructing a three-dimensional space tile data set corresponding to N levels of accuracy based on multi-density sampled real-world point cloud data includes: Obtaining real-time point cloud data, which contains three-dimensional spatial coordinates; Construct the real-time point cloud data into a bounding box of the real-time point cloud data; Inputting the pre-collected multi-scale precision control requirements of the 3D display system into the precision parameter setting model to obtain precision setting parameters, wherein the precision setting parameters include N precision levels and corresponding stratified sampling granularities; the multi-scale precision control requirements include a minimum magnification ratio, a maximum magnification ratio, a minimum reduction ratio, and a maximum reduction ratio; Based on N precision levels and corresponding hierarchical sampling granularity, the bounding box of the actual point cloud data is downsampled to obtain a downsampled point cloud data set corresponding to N precision levels; Based on the hierarchical sampling granularity of N precision levels and the downsampled point cloud data set, spatial slicing processing is performed to obtain a three-dimensional spatial tile data set corresponding to N precision levels.

7. The real-scene three-dimensional geographic entity and spatial planning display system according to claim 6, characterized in that: The method of constructing the field point cloud data into the field point cloud data bounding box includes: Obtain all three-dimensional coordinate points in the real-world point cloud data, and based on all three-dimensional coordinate points, determine the boundary values ​​of the real-world point cloud data in three dimensions, including the maximum and minimum coordinate values ​​in the X-axis direction, the maximum and minimum coordinate values ​​in the Y-axis direction, and the maximum and minimum coordinate values ​​in the Z-axis direction; The length of the bounding box of the real-world point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the X-axis direction, the width of the bounding box of the real-world point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the Y-axis direction, and the height of the bounding box of the real-world point cloud data is obtained by subtracting the maximum coordinate value from the minimum coordinate value in the Z-axis direction. The center coordinates of the bounding box of the point cloud data are obtained based on the length, width, and height of the bounding box of the real-world point cloud data. The length, width, height and center coordinates of the bounding box of the real-time point cloud data are used as the parameters of the bounding box of the real-time point cloud data to construct a geometric boundary volume that surrounds the entire spatial range of the real-time point cloud data, that is, the bounding box of the real-time point cloud data is obtained.

8. The real-scene three-dimensional geographic entity and spatial planning display system according to claim 6, characterized in that: The method for downsampling the bounding box of the field point cloud data based on N precision levels and corresponding hierarchical sampling granularity to obtain a downsampled point cloud data set corresponding to the N precision levels includes: S100: let the initial value of n be 1, and the value range of n be 1 to N; S101: Using the stratified sampling granularity of the nth precision level as the precision, divide along the X-axis Point cloud data grid units, divided along the Y axis Point cloud data grid units are divided along the Z axis point cloud data grid units, that is, a total of Total n Point cloud data grid unit; Indicates rounding up; CD bwh is the length of the bounding box of the field point cloud data, KD bwh is the width of the bounding box of the real-time point cloud data, GD bwh is the height of the bounding box of the field point cloud data, LD n is the stratified sampling granularity of the nth precision level; S102: Total n point cloud data grid units are constructed into a downsampled point cloud data set of the nth accuracy level; the downsampled point cloud data set maintains the same spatial boundary range as the bounding box of the actual point cloud data; S103: Let n=n+1. If n is less than or equal to N, continue executing S101 to S102; if n is greater than N, end the current process.

9. The real-scene three-dimensional geographic entity and space planning display system according to claim 6, characterized in that: A method for performing spatial slicing processing based on a hierarchical sampling granularity and a downsampled point cloud data set at N precision levels to obtain a three-dimensional spatial tile data set corresponding to N precision levels includes: S200: Let the initial value of n be 1, and the value range of n be 1 to N; S201: Get the total number of point cloud data grid cells in the nth downsampled point cloud data set n Total n Each point cloud data grid unit constructs the corresponding downsampled point cloud data sub-bounding box; the downsampled point cloud data sub-bounding box is the three-dimensional space tile, and the Total n Three-dimensional space tiles; S202: Total n Extract the length, width, height and center coordinates of the three-dimensional space tiles to get the Total n The 3D tile length, 3D tile width, 3D tile height and 3D tile center coordinates corresponding to the 3D tile; β times the stratified sampling granularity of the nth precision level is used as the 3D tile error parameter, where β is a constant greater than 0; S203: Total n The 3D tile error parameters, 3D tile length, 3D tile width, 3D tile height and 3D tile center coordinates corresponding to each 3D tile are constructed into Total n The three-dimensional tile index information corresponding to the three-dimensional tile; Total n The 3D tiles and the corresponding 3D tile index information are constructed into Total n 3D space tile data, Total n The three-dimensional space tile data is constructed into a three-dimensional space tile data set; S204: Let n=n+1. If n is less than or equal to N, continue to execute S201 to S103; if n is greater than N, end the current process.

10. The real-scene three-dimensional geographic entity and space planning display system according to claim 1, characterized in that: The training method of the operational behavior diagnosis model includes: Pre-constructing an operation behavior diagnostic data set, the operation behavior diagnostic data set including XW groups of operation behavior diagnostic data and scaling behavior patterns corresponding to the XW groups of operation behavior diagnostic data, where XW is a positive integer greater than 0, and the operation behavior diagnostic data includes operation behavior feature data; The operational behavior diagnosis dataset is divided into a training set and a validation set. The training set is used for parameter learning, and the validation set is used to evaluate the generalization ability of the model. A support vector machine model is used as the operation behavior diagnosis model. The operation behavior feature data is converted into feature vector input, nonlinear features are extracted through the hidden layer, and the scaling behavior pattern corresponding to the maximum probability is output in the output layer as the final prediction result; the training goal is to minimize the cross entropy loss function, and training is stopped when the accuracy of the verification set reaches the preset threshold.

Images