A working system for three-dimensional data display based on GIS spatial data

Abstract

The present invention relates to a 3D data display and analysis system based on GIS, which covers five main steps: data collection and preparation, data processing, feature extraction, data analysis and 3D display. Through unified projection transformation, hybrid Kriging interpolation, deep learning model fusion and spatial autocorrelation analysis and other technologies, remote sensing images, digital elevation models and building data are integrated and processed, and finally the combined 3D data is intuitively displayed in 3D GIS software, providing an efficient method for the visualization and in-depth analysis of geographic information.

Description

Technical Field

[0001] The present invention relates to a data processing and visualization method, in particular to a working system for displaying three-dimensional data based on GIS spatial data. Background Art

[0002] With the rapid development of geographic information system (GIS) technology, three-dimensional geographic information system (3D GIS) has become an important branch of geospatial information science. 3D GIS can not only express the spatial location and attributes of geographic information, but also provide more intuitive three-dimensional visualization effects, making the analysis and display of spatial data more vivid and specific. This technology has broad application prospects in urban planning, environmental monitoring, disaster management, architectural design and other fields.

[0003] Traditional two-dimensional GIS can only provide a flat display of geographic information, and cannot fully express three-dimensional information such as terrain undulations and building heights. Three-dimensional data display can truly reproduce the shape, location, and attributes of geographic entities in three-dimensional space, providing users with a more comprehensive and intuitive spatial perception. For example, in urban planning, three-dimensional display can help planners better understand the height and distribution of buildings, so as to carry out scientific and reasonable planning and design. Summary of the invention

[0004] The present invention aims to at least solve the technical problems existing in the prior art, and in particular innovatively proposes a working system for three-dimensional data display based on GIS spatial data, comprising:

[0005] 1. Data acquisition module

[0006] Acquire spatial data, which consists of remote sensing image data, digital elevation model and building data.

[0007] Obtain remote sensing image data through high-resolution satellite images, which provide detailed surface information, including roads, rivers, vegetation cover, etc. in the city;

[0008] The digital elevation model (DEM) is generated by LiDAR scanning technology. LiDAR data can accurately reflect the ups and downs and changes of the terrain. These data are processed and stored in DEM format, which is suitable for subsequent three-dimensional terrain modeling and analysis.

[0009] Obtain building data from existing geographic databases, which already contain basic information about most buildings in the city, such as building heights, outlines, etc.;

[0010] Among them, the remote sensing image is in raster format;

[0011] The terrain data is in grid format;

[0012] Building data is in vector format.

[0013] 2. Data processing module

[0014] 2.1 Data preprocessing:

[0015] Process the obtained spatial data;

[0016] Projection transformation: convert all spatial data into the same coordinate system.

[0017]

[0018] Among them, (x, y, z) is the converted spatial data coordinates, (x′, y′, z′) is the original spatial data coordinates, and Q is the projection transformation matrix;

[0019] The above transformations were applied to each spatial dataset, including remote sensing image data, digital elevation models, and building data, to ensure that all data were in the same coordinate system.

[0020] 2.2 Data cleaning: processing missing values ​​and outliers.

[0021] For missing values ​​and outliers in spatial data, a mixed Kriging interpolation formula is used

[0022]

[0023] where Z(S0) is the estimated value at the point S0(x0,y0,z0) where spatial data is missing, and λ u is a known spatial data point S u The weight, λ u The calculation formula is as follows:

[0024]

[0025] Among them, γ(S u ,S v ) is the variation function, which represents the known spatial data point S u and S v is the spatial correlation of , and μ is the Lagrange multiplier, which is used to satisfy the unbiased condition of Kriging interpolation.

[0026] By S u , S v The relationship between , can list the Kriging equation:

[0027] λ1γ(S1,S1)+λ2γ(S1,S2)+…+λ n γ(S1,S n )+μ=γ(S0,S1)

[0028] λ1γ(S2,S1)+λ2γ(S2,S2)+…+λ n γ(S2,S n )+μ=γ(S0,S2)

[0029]

[0030] λ1γ(S n ,S1)+λ2γ(S n ,S2)+…+λ n γ(S n ,S n )+μ=γ(S0,S n )

[0031] λ1+λ2+…+λ n =1

[0032] Solve the system of equations to get the weight λ u and the value of the Lagrange multiplier μ, which is used to calculate the estimated value Z(S0) of the missing point S0.

[0033] Data clipping: Clip data according to the study area to reduce the amount of calculation.

[0034] Determine the bounding box B of the spatial data d :

[0035] B d ={(x min ,y min ),(x max ,y max )}

[0036] in,

[0037] (x min ,y min ) and (x max ,y max ) are the coordinates of the bounding box;

[0038] (x,y)∈B d retain information;

[0039] Cut off information;

[0040] 3. Feature extraction module

[0041] 3.1 Image feature extraction

[0042] Extracting data features from clipped remote sensing impacts

[0043] 3.2 Object Detection

[0044] B detected ={(xc ,y c ,w,h)|Score(x c ,y c ,w,h)>Threshold}

[0045] Among them, (x c ,y c ) is the center coordinate of the building, w and h are the width and height of the building, Score is the detection score, and Threshold is the threshold.

[0046] 3.3 3D reconstruction of buildings

[0047] B 3D =Reconstruct(F detected )

[0048] B 3D is the reconstructed 3D building model, Reconstruct is the reconstruction function, and F detected is the detected building feature map;

[0049] 4. Data fusion module

[0050] 4.1 Data Fusion

[0051] Fusion of deep learning generated 3D building models with terrain data

[0052] D fused ={(x,y,z D )|z D =α(x,y)·z DL +β(x,y)·z DEM}

[0053] Among them, D fused represents the fused spatial data point set, z D The height of the building after the height is fused with the terrain data, z DL is the building height generated by the deep learning model, z DEM is the height of terrain data generated by the deep learning model, α(x,y) and β(x,y) are the dynamic learning rates of the building height and terrain data height at point (x,y), respectively, which are used to improve the model accuracy.

[0054]

[0055] Among them, α(x,y) and β(x,y) are the dynamic learning rates of the building height and terrain data height at (x,y), respectively, and σ target is the target uncertainty coefficient, which indicates the expected uncertainty level of the system;

[0056] κ is the adjustment coefficient, which is used to adjust the impact of spatial features on the learning rate.

[0057] σ DL (x,y) represents the uncertainty of the building height generated by deep learning at the location (x,y).

[0058] σ DEM (x,y) represents the uncertainty of the height of the terrain data generated by deep learning at the position (x,y).

[0059] G DL (x,y) represents the spatial features of building height generated by deep learning at the location (x,y);

[0060] G DEM (x,y) represents the spatial characteristics of the height of terrain data generated by deep learning at the position (x,y).

[0061] and The uncertainty of building height generated by deep learning and the uncertainty of terrain data height generated by deep learning respectively reflect the impact on the learning rate. In places with low uncertainty, the learning rate is increased, and in places with high uncertainty, the learning rate is reduced to reduce the negative impact on the model.

[0062] and It reflects the impact of the spatial characteristics of building heights generated by deep learning and the spatial characteristics of terrain data heights generated by deep learning on the learning rate. In places where the terrain changes dramatically, the learning rate is reduced to reduce the instability of the model.

[0063] 4.2 Spatial Analysis Module:

[0064] Slope calculation and aspect calculation:

[0065]

[0066] Among them, S is the slope, A is the slope aspect, and are the rates of change of height in the x and y directions respectively.

[0067] Viewshed analysis: Analyze the visible range at different locations.

[0068] Composite viewshed analysis:

[0069]

[0070] f d (A) = cos 2 (θ dA )

[0071] Among them, V composite Represents the joint visual range, V base,dir It is the basic visible range in the direction dir;

[0072] f d (A) is a directional function used to adjust the visible range in the direction dir according to the slope aspect A. This function can be a function based on the angle difference between the slope aspect and the direction dir;

[0073] θ dA is the angle between the slope aspect A and the direction dir, and the direction function f d (A) It reaches the maximum value 1 when the slope aspect A is consistent with the direction dir, and reaches the minimum value 0 when they are perpendicular.

[0074] Composite viewshed analysis is used to determine the common visible range between multiple observation points, so as to accurately display the composite viewshed in 3D GIS applications and conduct further spatial analysis;

[0075] Spatial statistics: Introduce spatial autocorrelation analysis to study the spatial aggregation and distribution patterns of geographical phenomena.

[0076]

[0077] Among them, P i is Moran's I index, indicating spatial autocorrelation, N is the number of observations, and w ij is the spatial weight matrix, h i and h j are the i-th and j-th observations, respectively. is the average of the observations of all observation points, and W0 is the sum of the spatial weight matrices.

[0078] Moran's I index calculates the differences between all observations and combines them with the spatial weight matrix to assess whether these differences are spatially clustered or dispersed;

[0079] When the Moran's I index is positive, it indicates that the observations in adjacent areas are relatively similar and there is a spatial aggregation phenomenon;

[0080] When the Moran's I index is negative, it means that the observations in adjacent areas are quite different and there is a spatial dispersion phenomenon;

[0081] When the Moran's I index is close to zero, it indicates that the observations are randomly distributed in space;

[0082] In the air pollution data displayed by this system, the PM2.5 value recorded at each observation point is the observed value. By analyzing the spatial autocorrelation between these observed values, it can be determined whether pollutants are concentrated in certain areas.

[0083] Spatial weight matrix w ij The calculation method is

[0084]

[0085] Among them, d ij is the distance between the i-th observation point and the j-th observation point, τ is the parameter that controls the exponential decay, λ is the parameter that controls the polynomial decay, and φ is the ratio of the two decay modes. When φ approaches 1, the weight matrix is ​​mainly determined by the exponential decay, and when φ approaches 0, the weight matrix is ​​mainly determined by the polynomial decay.

[0086] 5. 3D data display

[0087] 3D Modeling:

[0088] Convert terrain data into 3D surface models.

[0089] 3D GIS software:

[0090] Use 3D GIS software for data display.

[0091] Attribute Mapping:

[0092] The results of Moran's I index and weight matrix are mapped to the attributes of 3D models to enhance the readability and intuitiveness of the data.

[0093] The combined three-dimensional data after mapping is

[0094] D combined ={(x,y,z D ,P i ,w ij )}

[0095] Moran's I Index P i Using Color Mapping

[0096]

[0097] Weight matrix w ij Using transparency mapping

[0098]

[0099] Color represents the calculated color value, and Opacity represents the calculated transparency;

[0100] Attribute is the attribute value of the current data point. min and Attribute max are the minimum and maximum values ​​of the attribute;

[0101] Color min and Color max are the maximum and minimum values ​​in the color space;

[0102] Opacity min and Opacity max are the minimum and maximum values ​​of transparency;

[0103] The role of color mapping and transparency mapping is to standardize the attribute values, and then interpolate in the color space and transparency range according to the standardized ratio, so as to dynamically change the color and transparency of the data points and improve the intuitiveness of the visualized data.

[0104] D combined is the Moran's I index P i , weight matrix w ij and the three-dimensional data point set D fused The combined three-dimensional data is used to achieve richer three-dimensional data display.

[0105] Combined 3D data D combined Generate a 3D mesh and render it to get a visual 3D model, allowing users to intuitively explore and analyze spatial data.

[0106] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0107] Through unified projection transformation, hybrid Kriging interpolation, deep learning model fusion, spatial autocorrelation analysis and other technologies, remote sensing images, digital elevation models and building data are integrated and processed, and finally the combined three-dimensional data is intuitively displayed in the three-dimensional GIS software, providing an efficient method for the visualization and in-depth analysis of geographic information.

[0108] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0110] Figure 1 It is a system schematic diagram of the present invention.

[0111] Figure 2 It is a three-dimensional display of the buildings in part of the study area.

[0112] Figure 3 It is a plot of building height versus mean square error.

[0113] Figure 4 This is a graph showing the relationship between building density and Moran's I index.

[0114] Figure 5 is a graph of the number of data points versus rendering time and memory usage. DETAILED DESCRIPTION

[0115] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0116] like Figure 1 As shown, the present invention discloses a working system for three-dimensional data display based on GIS spatial data, including the following modules:

[0117] 1. Data acquisition module

[0118] Acquire spatial data, which consists of remote sensing image data, digital elevation model and building data.

[0119] Obtain remote sensing image data through high-resolution satellite images, which provide detailed surface information, including roads, rivers, vegetation cover, etc. in the city;

[0120] The digital elevation model (DEM) is generated by LiDAR scanning technology. LiDAR data can accurately reflect the ups and downs and changes of the terrain. These data are processed and stored in DEM format, which is suitable for subsequent three-dimensional terrain modeling and analysis.

[0121] Obtain building data from existing geographic databases, which already contain basic information about most buildings in the city, such as building heights, outlines, etc.;

[0122] Among them, the remote sensing image is in raster format;

[0123] The terrain data is in grid format;

[0124] Building data is in vector format.

[0125] 2. Data processing module

[0126] 2.1 Data preprocessing:

[0127] Process the obtained spatial data;

[0128] Projection transformation: convert all spatial data into the same coordinate system.

[0129]

[0130] Among them, (x, y, z) is the converted spatial data coordinates, (x′, y′, z′) is the original spatial data coordinates, and Q is the projection transformation matrix;

[0131] The above transformations were applied to each spatial dataset, including remote sensing image data, digital elevation models, and building data, to ensure that all data were in the same coordinate system.

[0132] 2.2 Data cleaning: processing missing values ​​and outliers.

[0133] For missing values ​​and outliers in spatial data, a mixed Kriging interpolation formula is used

[0134]

[0135] where Z(S0) is the estimated value at the point S0(x0,y0,z0) where spatial data is missing, and λ u is a known spatial data point S u The weight, λ u The calculation formula is as follows:

[0136]

[0137] Among them, γ(S u ,S v ) is the variation function, which represents the known spatial data point S u and S v is the spatial correlation of , and μ is the Lagrange multiplier, which is used to satisfy the unbiased condition of Kriging interpolation.

[0138] By S u , S v The relationship between , can list the Kriging equation:

[0139] λ1γ(S1,S1)+λ2γ(S1,S2)+…+λ n γ(S1,S n )+μ=γ(S0,S1)

[0140] λ1γ(S2,S1)+λ2γ(S2,S2)+…+λ n γ(S2,S n )+μ=γ(S0,S2)

[0141]

[0142] λ1γ(S n ,S1)+λ2γ(S n ,S2)+…+λ n γ(S n ,S n )+μ=γ(S0,S n )

[0143] λ1+λ2+…+λ n =1

[0144] Solve the system of equations to get the weight λ u and the value of the Lagrange multiplier μ, which is used to calculate the estimated value Z(S0) of the missing point S0.

[0145] Data clipping: Clip data according to the study area to reduce the amount of calculation.

[0146] Determine the bounding box B of the spatial data d :

[0147] B d ={(x min ,y min ),(x max ,y max )}

[0148] in,

[0149] (x min ,y min ) and (x max ,y max ) are the coordinates of the bounding box;

[0150] (x,y)∈B d retain information;

[0151] Cut off information;

[0152] 3. Feature extraction module

[0153] 3.1 Image feature extraction

[0154] Extracting data features from clipped remote sensing impacts

[0155] 3.2 Object Detection

[0156] B detected ={(x c ,y c ,w,h)|Score(x c ,y c ,w,h)>Threshold}

[0157] Among them, (xc ,y c ) is the center coordinate of the building, w and h are the width and height of the building, Score is the detection score, and Threshold is the threshold.

[0158] 3.3 3D reconstruction of buildings

[0159] B 3D =Reconstruct(F detected )

[0160] B 3D is the reconstructed 3D building model, Reconstruct is the reconstruction function, and F detected is the detected building feature map;

[0161] 4. Data fusion module

[0162] 4.1 Data Fusion

[0163] Fusion of deep learning generated 3D building models with terrain data

[0164] D fused ={(x,y,z D )|z D =α(x,y)·z DL +β(x,y)·z DEM}

[0165] Among them, D fused represents the fused spatial data point set, z D The height of the building after the height is fused with the terrain data, z DL is the building height generated by the deep learning model, z DEM is the height of terrain data generated by the deep learning model, α(x,y) and β(x,y) are the dynamic learning rates of the building height and terrain data height at point (x,y), respectively, which are used to improve the model accuracy.

[0166]

[0167] Among them, α(x,y) and β(x,y) are the dynamic learning rates of the building height and terrain data height at (x,y), respectively, and σ target is the target uncertainty coefficient, which indicates the expected uncertainty level of the system;

[0168] κ is the adjustment coefficient, which is used to adjust the impact of spatial features on the learning rate.

[0169] σ DL (x,y) represents the uncertainty of the building height generated by deep learning at the location (x,y).

[0170] σ DEM (x,y) represents the uncertainty of the height of the terrain data generated by deep learning at the position (x,y).

[0171] G DL (x,y) represents the spatial features of building height generated by deep learning at the location (x,y);

[0172] G DEM (x,y) represents the spatial characteristics of the height of terrain data generated by deep learning at the position (x,y).

[0173] and The uncertainty of building height generated by deep learning and the uncertainty of terrain data height generated by deep learning respectively reflect the impact on the learning rate. In places with low uncertainty, the learning rate is increased, and in places with high uncertainty, the learning rate is reduced to reduce the negative impact on the model.

[0174] and It reflects the impact of the spatial characteristics of building heights generated by deep learning and the spatial characteristics of terrain data heights generated by deep learning on the learning rate. In places where the terrain changes dramatically, the learning rate is reduced to reduce the instability of the model.

[0175] 4.2 Spatial Analysis Module:

[0176] Slope calculation and aspect calculation:

[0177]

[0178] Among them, S is the slope, A is the slope aspect, and are the rates of change of height in the x and y directions respectively.

[0179] Viewshed analysis: Analyze the visible range at different locations.

[0180] Composite viewshed analysis:

[0181]

[0182] f d (A) = cos 2 (θ dA )

[0183] Among them, V composite Represents the joint visual range, V base,dir It is the basic visible range in the direction dir;

[0184] f d(A) is a directional function used to adjust the visible range in the direction dir according to the slope aspect A. This function can be a function based on the angle difference between the slope aspect and the direction dir;

[0185] θ dA is the angle between the slope aspect A and the direction dir, and the direction function f d (A) It reaches the maximum value 1 when the slope aspect A is consistent with the direction dir, and reaches the minimum value 0 when they are perpendicular.

[0186] Composite viewshed analysis is used to determine the common visible range between multiple observation points, so as to accurately display the composite viewshed in 3D GIS applications and conduct further spatial analysis;

[0187] Spatial statistics: Introduce spatial autocorrelation analysis to study the spatial aggregation and distribution patterns of geographical phenomena.

[0188]

[0189] Among them, P i is Moran's I index, indicating spatial autocorrelation, N is the number of observations, and w ij is the spatial weight matrix, h i and h j are the i-th and j-th observations, respectively. is the average of the observations of all observation points, and W0 is the sum of the spatial weight matrices.

[0190] Moran's I index calculates the differences between all observations and combines them with the spatial weight matrix to assess whether these differences are spatially clustered or dispersed;

[0191] When the Moran's I index is positive, it indicates that the observations in adjacent areas are relatively similar and there is a spatial aggregation phenomenon;

[0192] When the Moran's I index is negative, it means that the observations in adjacent areas are quite different and there is a spatial dispersion phenomenon;

[0193] When the Moran's I index is close to zero, it indicates that the observations are randomly distributed in space;

[0194] In the air pollution data displayed by this system, the PM2.5 value recorded at each observation point is the observed value. By analyzing the spatial autocorrelation between these observed values, it can be determined whether pollutants are concentrated in certain areas.

[0195] Spatial weight matrix w ij The calculation method is

[0196]

[0197] Among them, d ij is the distance between the i-th observation point and the j-th observation point, τ is the parameter that controls the exponential decay, λ is the parameter that controls the polynomial decay, and φ is the ratio of the two decay modes. When φ approaches 1, the weight matrix is ​​mainly determined by the exponential decay, and when φ approaches 0, the weight matrix is ​​mainly determined by the polynomial decay.

[0198] 5. 3D data display

[0199] 3D Modeling:

[0200] Convert terrain data into 3D surface models.

[0201] 3D GIS software:

[0202] Use 3D GIS software for data display.

[0203] Attribute Mapping:

[0204] The results of Moran's I index and weight matrix are mapped to the attributes of 3D models to enhance the readability and intuitiveness of the data.

[0205] The combined three-dimensional data after mapping is

[0206] D combined ={(x,y,z D ,P i ,w ij )}

[0207] Moran's I Index P i Using Color Mapping

[0208]

[0209] Weight matrix w ij Using transparency mapping

[0210]

[0211] Color represents the calculated color value, and Opacity represents the calculated transparency;

[0212] Attribute is the attribute value of the current data point. min and Attribute max are the minimum and maximum values ​​of the attribute;

[0213] Color min and Color max are the maximum and minimum values ​​in the color space;

[0214] Opacity min and Opacity max are the minimum and maximum values ​​of transparency;

[0215] The role of color mapping and transparency mapping is to standardize the attribute values, and then interpolate in the color space and transparency range according to the standardized ratio, so as to dynamically change the color and transparency of the data points and improve the intuitiveness of the visualized data.

[0216] D combined is the Moran's I index P i , weight matrix w ij and the three-dimensional data point set D fused The combined three-dimensional data is used to achieve richer three-dimensional data display.

[0217] like Figure 2 As shown, using the combined three-dimensional data D combined Generate a 3D mesh and render it to get a visual 3D model, allowing users to intuitively explore and analyze spatial data.

[0218] 6. Experimental Data

[0219] 6.1 Accuracy evaluation experiment of building height: This experiment focuses on verifying the accuracy of the 3D building model generated by the system and the actual building height. To this end, three areas with different building heights were selected, namely 50 meters, 75 meters and 100 meters. In each area, the building height generated by the system was compared with the building height measured on the spot, and the root mean square error (RMSE) was calculated. The experimental results are as follows:

[0220] like Figure 3 As shown, in area 1, the actual building height is 50 meters, while the height generated by the system is 49.5 meters, and the RMSE value is 0.5 meters, indicating that the generated model has high accuracy.

[0221] In area 2, the actual building height is 75 meters, the generated height is 74.8 meters, and the RMSE is only 0.2 meters, which further verifies the reliability of the system.

[0222] In region 3, the actual building height is 100 meters, the generated height is 99.7 meters, and the RMSE is 0.3 meters.

[0223]

[0224] This experiment verified that the system has very high accuracy in generating three-dimensional building models through measurements in multiple height areas, with errors controlled within a very small range, proving the practical application value of the system.

[0225] 6.2 Spatial Statistics and Autocorrelation Analysis Experiment: In order to better understand the performance of the system in geospatial analysis, the experiment analyzed the spatial autocorrelation of areas with different building densities. The Moran’s I index, as a measure of spatial autocorrelation, was applied to three different areas with increasing building density.

[0226] like Figure 4 Shown: The building density in area 1 is 50 units / m 2 , Moran's I index is 0.85, indicating that the building distribution in this area shows strong spatial autocorrelation.

[0227] The building density in area 2 is 75 units / m 2 , Moran's I index is 0.65, and the spatial autocorrelation begins to weaken.

[0228] The building density in area 3 is 100 units / m 2 , Moran's I index is 0.40, and the autocorrelation is further reduced.

[0229]

[0230] Through analysis, it can be seen that with the increase of building density, Moran's I index gradually decreases, which indicates that building density has a significant impact on spatial distribution patterns. The experimental results show that the system can successfully capture and analyze spatial autocorrelation and is suitable for spatial analysis scenarios such as building planning and land use.

[0231] 6.3 Rendering performance analysis experiment: To evaluate the performance of the system when processing data sets of different sizes, a rendering performance test was conducted. The experiment involved different numbers of data points and recorded the rendering time and memory usage.

[0232] like Figure 5 As shown, in scene 1, the number of data points is 10,000, the rendering time is 0.8 seconds, and the memory usage is 200MB, indicating that the system is able to quickly render smaller-scale data.

[0233] In scenario 2, the number of data points increases to 50,000, the rendering time is 1.5 seconds, and the memory usage is 400MB. The system still maintains efficient performance under medium-sized data.

[0234] In scene 3, the number of data points reaches 100,000, the rendering time is 2.3 seconds, and the memory usage is 600MB. Although the data scale is large, the rendering speed and memory usage are still within an acceptable range.

[0235]

[0236] This experiment proves that the system still has good responsiveness when processing large-scale data, can complete three-dimensional rendering within a reasonable time, and has the potential to be used in actual projects.

[0237] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A working system for three-dimensional data display based on GIS spatial data, characterized in that: include: Data acquisition module: acquires spatial data, including remote sensing image data, digital elevation model and building data; Data processing module: converts the acquired spatial data into the same coordinate system through the projection transformation matrix, determines the boundary of the spatial data, and performs interpolation processing on missing values ​​and outliers; Feature extraction module: extracts data features from cropped remote sensing images, detects buildings and reconstructs 3D models; Data fusion module: fuses the 3D building model generated by deep learning with terrain data; Spatial analysis module: calculate and analyze spatial characteristics, including spatial autocorrelation analysis and composite viewshed analysis; 3D data display module: Generate 3D meshes and render them using 3D data, and map spatial features to corresponding 3D data; The data fusion module comprises: Data Fusion Fusion of deep learning generated 3D building models with terrain data D fused {(x,y,z D )|z D =α(x,y)·z DL +β(x,y)·z DEM } Among them, D fused represents the fused spatial data point set, z D The height of the building after the height is fused with the terrain data, z DL is the building height generated by the deep learning model, z DEM is the height of terrain data generated by the deep learning model; Among them, α(x,y) and β(x,y) are the dynamic learning rates of the building height and terrain data height at (x,y), respectively, and σ target is the target uncertainty coefficient, which indicates the expected uncertainty level of the system; κ is the adjustment coefficient, which is used to adjust the impact of spatial features on the learning rate. σ DL (x,y) represents the uncertainty of the building height generated by deep learning at the location (x,y); σ DEM (x,y) represents the uncertainty of the height of the terrain data generated by deep learning at the position (x,y); G DL (x,y) represents the spatial features of building height generated by deep learning at the location (x,y); G DEM (x,y) represents the spatial characteristics of the height of terrain data generated by deep learning at the position (x,y); and The uncertainty of building height generated by deep learning and the uncertainty of terrain data height generated by deep learning respectively reflect the impact on the learning rate. In places with low uncertainty, the learning rate is increased, and in places with high uncertainty, the learning rate is reduced to reduce the negative impact on the model. and It reflects the impact of the spatial characteristics of building heights generated by deep learning and the spatial characteristics of terrain data heights generated by deep learning on the learning rate. In places where the terrain changes dramatically, the learning rate is reduced to reduce the instability of the model.

2. The working system for displaying three-dimensional data based on GIS spatial data according to claim 1 is characterized in that: The data acquisition module comprises: Remote sensing image data is obtained through high-resolution satellite images, digital elevation models are generated through LiDAR scanning technology, and building data is obtained from existing geographic databases.

3. The working system for displaying three-dimensional data based on GIS spatial data according to claim 1 is characterized in that: The data processing module comprises: Process the obtained spatial data; Projection transformation: convert all spatial data into the same coordinate system; Among them, (x, y, z) is the converted spatial data coordinates, (x′, y′, z′) is the original spatial data coordinates, and Q is the projection transformation matrix; Apply the above transformation to each spatial dataset, including remote sensing image data, digital elevation model and building data, to ensure that all data are in the same coordinate system; For missing values ​​and outliers in spatial data, a mixed Kriging interpolation formula is used where Z(S0) is the estimated value at the point S0(x0,y0,z0) where spatial data is missing, and λ u is a known spatial data point S u The weight of Z(S u ) represents the data point S in the known space u The observed value, n represents n known spatial data points, the mixed Kriging interpolation formula uses n known spatial data points to estimate the missing value of the spatial data missing point S0, λ u The calculation formula is as follows: Among them, γ(S u ,S v ) is the variogram, which represents the known spatial data point S u and S v spatial correlation of The variation function γ(S0,S v ) represents the spatial data missing point S0 and the known spatial data point S v spatial correlation of μ is the Lagrange multiplier; By S u , S v The relationship between , can be listed Kriging equation: λ1γ(S1,S1)+λ2γ(S1,S2)+…+λ n γ(S1,S n )+μ=γ(S0,S1) λ1γ(S2,S1)+λ2γ(S2,S2)+…+λ n γ(S2,S n )+μ=γ(S0,S2) λ1γ(S n ,S1)+λ2γ(S n ,S2)+…+λ n γ(S n ,S n )+μ=γ(S0,S n ) λ1+λ2+…+λ n =1 Solve the system of equations to get the weight λ u and the value of the Lagrange multiplier μ, which is used to calculate the estimated value Z(S0) of the missing point S0; Determine the bounding box B of the spatial data d : B d ={(x min ,y min ),(x max ,y max )} in, (x min ,y min ) and (x max ,y max ) are the coordinates of the bounding box; If (x,y)∈B d When, retain the information; like Delete the information when 4. The working system for displaying three-dimensional data based on GIS spatial data according to claim 1 is characterized in that: The feature extraction module comprises: Image feature extraction: Extracting data features from clipped remote sensing impacts; Object Detection: B detected ={(x c ,y c ,w,h)|Score(x c ,y c ,w,h)>Threshold} Among them, (x c ,y c ) is the center coordinate of the building, w and h are the width and height of the building, Score is the detection score, and Threshold is the threshold; 3D reconstruction of buildings: B 3D =Reconstruct(F detected ) B 3D is the reconstructed 3D building model, Reconstruct is the reconstruction function, and F detected is the detected building feature map.

5. The working system for displaying three-dimensional data based on GIS spatial data according to claim 1 is characterized in that: The spatial analysis module includes: Slope calculation and aspect calculation: Among them, S is the slope, A is the slope direction, and are the rates of change of height in the x and y directions, respectively; Composite viewshed analysis: f d (A)=cos 2 (i dA ) Among them, V composite Represents the joint visual range, V base,dir is the basic visible range in the direction dir, ρ is the attenuation coefficient; f d (A) is a directional function used to adjust the visible range in the direction dir according to the slope aspect A. This function can be a function based on the angle difference between the slope aspect and the direction dir; θ dA is the angle between the slope aspect A and the direction dir, and the direction function f d (A) It reaches the maximum value 1 when the slope aspect A is consistent with the direction dir, and reaches the minimum value 0 when they are perpendicular.

6. The working system for displaying three-dimensional data based on GIS spatial data according to claim 1 is characterized in that: The spatial analysis module includes: Spatial statistics: Introduce spatial autocorrelation analysis to study the spatial aggregation and distribution patterns of geographical phenomena; Among them, P i is Moran's I index, indicating spatial autocorrelation, N is the number of observations, and w ij is the spatial weight matrix, h i and h j are the i-th and j-th observations, respectively. is the average value of all observation points, and W0 is the sum of the spatial weight matrix; Spatial weight matrix w ij The calculation method is Among them, d ij is the distance between the i-th observation point and the j-th observation point, τ is the parameter that controls the exponential decay, λ is the parameter that controls the polynomial decay, and φ is the ratio of the two decay modes. When φ approaches 1, the weight matrix is ​​mainly determined by the exponential decay, and when φ approaches 0, the weight matrix is ​​mainly determined by the polynomial decay.

7. The working system for displaying three-dimensional data based on GIS spatial data according to claim 6 is characterized in that: The three-dimensional data display module includes: 3D Modeling: Convert terrain data into three-dimensional surface models; 3D GIS software: Use 3D GIS software for data display; Attribute Mapping: Map the results of Moran's I index and weight matrix to the attributes of the 3D model to enhance the readability and intuitiveness of the data; The combined three-dimensional data after mapping is: D combined ={(x,y,z D ,P i ,w ij )} Moran's I Index P i Use a color map: Spatial weight matrix w ij Use transparency mapping: Color represents the calculated color value, and Opacity represents the calculated transparency; Attribute is the attribute value of the current data point. min and Attribute max are the minimum and maximum values ​​of the attribute; Color min and Color max are the minimum and maximum values ​​in the color space; Opacity min and Opacity max are the minimum and maximum values ​​of transparency; D combined is the Moran's I index P i , spatial weight matrix w ij and the three-dimensional data point set D fused The combined three-dimensional data is used to achieve richer three-dimensional data display; Combined 3D data D combined Generate a 3D mesh and render it to get a visual 3D model, allowing users to intuitively explore and analyze spatial data.

Images