A multi-source heterogeneous data fusion method and system based on a three-dimensional scene for a UAV
By combining topological matching degree and joint probability density model, the problem of insufficient fusion accuracy of UAV 2D surveillance data and 3D scene data is solved, realizing high-precision data fusion and visualization in complex 3D scenes.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING ZHIWANG YILIAN TECH CO LTD
- Filing Date
- 2025-11-06
- Publication Date
- 2026-06-02
AI Technical Summary
Existing technologies cannot effectively integrate two-dimensional surveillance data from drones with three-dimensional scene data, resulting in insufficient fusion accuracy in complex three-dimensional scenes and an inability to accurately determine whether the drone's flight path is below the mountain's elevation and poses a collision risk.
A multi-source heterogeneous data fusion method based on 3D scene UAVs is adopted. Spatially consistent data are screened by topological matching degree, and the credibility is quantified by joint probability density model. The fusion weight is calculated and visualized.
It improves data reliability and fusion accuracy, reduces the impact of sensor errors and 3D scene modeling deviations, adapts to the high-precision needs of different industries, and supports multi-scenario applications.
Smart Images
Figure CN121366254B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of multi-source heterogeneous data fusion technology, and more specifically to a method and system for multi-source heterogeneous data fusion from unmanned aerial vehicles (UAVs) based on a three-dimensional scene. Background Technology
[0002] Currently, existing patents only process two-dimensional surveillance data of UAVs (such as two-dimensional tracks, equipment status, and planar coordinate data), and cannot be compatible with three-dimensional scene data in UAV operation scenarios (such as three-dimensional LiDAR point clouds, oblique photogrammetry three-dimensional images, and digital elevation model (DEM) terrain data). This results in the fusion results lacking the three-dimensional dimension of two-dimensional surveillance + three-dimensional space, and cannot support the precise operation and monitoring of UAVs in complex three-dimensional scenarios (such as between urban high-rise buildings and in mountainous areas).
[0003] Although the joint probability density association model can achieve association at the two-dimensional data level (such as probabilistic matching between flight tracks and planar position data), it does not take into account the spatial topological relationships of the three-dimensional scene (such as the elevation difference between the UAV flight track and the three-dimensional terrain, and the spatial overlap between the three-dimensional point cloud and the UAV monitoring target). This results in insufficient fusion accuracy in the three-dimensional scene. For example, it is impossible to accurately determine whether the UAV flight track is below the mountain elevation and there is a risk of collision.
[0004] Therefore, designing a method and system for fusing multi-source heterogeneous data from UAVs based on 3D scenes, and realizing the multi-source heterogeneous fusion of UAV 2D surveillance data and 3D scene data, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] In view of this, the present invention provides a method and system for fusing multi-source heterogeneous data from UAVs based on a 3D scene, which can effectively improve data credibility. It filters spatially consistent data through topological matching degree and quantifies credibility by combining a joint probability density model, thus avoiding the problem of UAV sensor error and 3D scene modeling deviation.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for fusing multi-source heterogeneous data from UAVs based on a three-dimensional scene, comprising:
[0007] Step 1: Acquire drone flight data and 3D scene data;
[0008] Step 2: Preprocess the UAV flight data and 3D scene data respectively;
[0009] Step 3: Calculate the topology matching degree based on the preprocessed UAV flight data and 3D scene data;
[0010] Step 4: Calculate the fusion weights of UAV flight data and 3D scene data based on the topology matching degree and the joint probability density association model;
[0011] Step 5: Fuse the UAV flight data and 3D scene data according to the fusion weights, and visualize the fusion results.
[0012] Preferably, the UAV flight data includes GPS latitude and longitude, altitude, flight speed, and sampling timestamp;
[0013] The preprocessing of UAV flight data includes: data format standardization, handling of missing and outlier values, coordinate system unification, data resampling, and noise filtering.
[0014] Preferably, the three-dimensional scene data includes: terrain model, scene mesh, and texture map;
[0015] The preprocessing of the 3D scene data includes: data lightweighting and redundancy removal, geometric error correction, coordinate and resolution unification, and invalid region masking.
[0016] The preprocessed UAV flight data and the preprocessed 3D scene data are subjected to consistency verification.
[0017] Preferably, step 3, which calculates the topology matching degree, includes: calculating the elevation difference and spatial overlap based on the preprocessed UAV flight data and 3D scene data, integrating the elevation difference and spatial overlap according to weights, and obtaining the final topology matching degree.
[0018] Preferably, the process of calculating the elevation difference includes:
[0019] Spatial alignment: Using the coordinate system of the 3D scene data as a reference, the UAV flight data is mapped to the same coordinate system through coordinate transformation to ensure that the spatial positions of the two correspond.
[0020] Sampling point selection: Select N sampling points at fixed intervals within the overlapping area of the fused scene of UAV flight data and 3D scene data to cover the entire data area;
[0021] Calculate elevation difference: Calculate the elevation difference of each point, and use the average elevation difference or the maximum elevation difference to characterize the overall vertical matching degree.
[0022] Preferably, the process of calculating spatial overlap includes:
[0023] Extracting the region boundary: Extracting the minimum bounding rectangle S of the flight trajectory from the UAV flight data. 无人机 Extract the corresponding effective scene boundary S from the 3D scene data 场景 ;
[0024] Calculate the overlapping area and the total area: Use spatial geometry algorithms to calculate the intersection area S between the drone region and the 3D scene region. 重叠 Calculate the area S of the union of the two sets. 并集 =S 无人机 +S 场景 -S 重叠 ;
[0025] Calculate spatial overlap: S 空间重叠度 =S 重叠 / S 并集 ;
[0026] The topology matching degree is obtained by weighting the elevation difference and spatial overlap.
[0027] Preferably, the calculation process for the fusion weights includes:
[0028] The elevation difference and spatial overlap in the topological matching degree are extracted and used as constraints input into the joint probability density association model.
[0029] Based on the mean, variance, and joint covariance of the observation errors of the preprocessed UAV flight data and 3D scene data, a joint probability density function of a two-dimensional normal distribution is constructed, and the initial probability density value of each spatial sampling point is calculated.
[0030] The initial probability density value is weighted and corrected using the topological matching degree to enhance the probability contribution of high-matching data.
[0031] The corrected probability density values are normalized to obtain the fusion weights corresponding to each sampling point. The fusion weights range from [0, 1].
[0032] Preferably, the expression for the joint probability density association model is:
[0033] ;
[0034] Where x represents the preprocessed UAV flight data observation at the target sampling point, y represents the preprocessed 3D scene data observation at the same target sampling point, f(x,y) is the joint probability density value of the two types of data at that sampling point; μ1 is the mean elevation error of the UAV flight data; σ1 2 σ² represents the elevation error variance of the UAV flight data; μ² represents the mean elevation error of the 3D scene data; σ² represents the mean elevation error of the UAV flight data. 2 ρ represents the elevation error variance of the 3D scene data; ρ represents the joint error covariance, which is determined by analyzing the error correlation between the two types of data in historical matching samples.
[0035] Preferably, the formula for weighted correction of the initial probability density value using the topological matching degree is as follows:
[0036] F'(x,y)=f(x,y)×[0.5+0.5×M];
[0037] Where 0.5 is the basic weight coefficient, and 0.5×M is the contribution weight of topology matching degree; M is the topology matching degree.
[0038] Preferably, a UAV multi-source heterogeneous data fusion system based on a 3D scene includes:
[0039] Data acquisition module: used to acquire UAV flight data and 3D scene data;
[0040] Preprocessing module: used to preprocess the UAV flight data and 3D scene data respectively;
[0041] Topology matching degree calculation module: Calculates the topology matching degree based on the preprocessed UAV flight data and 3D scene data;
[0042] Model calculation module: used to calculate the fusion weights of UAV flight data and 3D scene data based on the topological matching degree and the joint probability density association model;
[0043] Visualization module: Used to fuse UAV flight data and 3D scene data according to fusion weights, and visualize the fusion results.
[0044] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for fusing multi-source heterogeneous data from UAVs based on three-dimensional scenes, with the following specific beneficial effects:
[0045] 1. Improve data credibility and reduce the risk of errors from a single data source: By calculating the topology matching degree, basic data with high spatial consistency is first screened, and then the data credibility is quantified by combining the joint probability density association model. This avoids the impact of drone sensor errors (such as elevation fluctuations caused by airflow interference) or 3D scene modeling deviations (such as splicing seam errors) on the fusion results. Compared with using drone or 3D scene data alone, the data reliability is significantly improved.
[0046] 2. Improve fusion accuracy and achieve precise spatial information matching: Through a progressive design of coordinate unification, topology matching and probability weighting, it ensures that UAV data and 3D scene data are highly consistent in spatial position (horizontal overlap) and elevation information (vertical deviation), and the fusion weight is strongly correlated with data quality (high matching data has higher weight), which meets the needs of high-precision scenarios such as surveying and inspection, and is superior to the accuracy performance of traditional simple average fusion methods.
[0047] 3. Enhanced data adaptability to meet the needs of multiple application scenarios: The method of this invention supports the processing of different types of UAV data (such as GPS trajectory and elevation sampling values) and 3D scene data (such as DEM terrain model and oblique photogrammetry 3D mesh), and the fusion severity can be flexibly adjusted through weight thresholds. For example, in emergency rescue scenarios, the threshold can be relaxed to quickly obtain fusion results; in precision surveying scenarios, the threshold can be increased to ensure data accuracy, adapting to the differentiated needs of different industries (such as geographic surveying, power line inspection, and digital twins). Attached Figure Description
[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0049] Figure 1 The method flowchart provided by the present invention. Detailed Implementation
[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0051] like Figure 1 As shown, this embodiment of the invention discloses a method for fusing multi-source heterogeneous data from UAVs based on a three-dimensional scene, including:
[0052] Step 1: Acquire drone flight data and 3D scene data;
[0053] Step 2: Preprocess the UAV flight data and 3D scene data respectively;
[0054] Step 3: Calculate the topology matching degree based on the preprocessed UAV flight data and 3D scene data;
[0055] Step 4: Weighting the fusion of UAV flight data and 3D scene data;
[0056] Step 5: Fuse the UAV flight data and 3D scene data according to the fusion weights, and visualize the fusion results.
[0057] Specifically, the UAV flight data includes GPS latitude and longitude, altitude, flight speed, and sampling timestamp;
[0058] The preprocessing of UAV flight data includes: data format standardization, handling of missing and outlier values, coordinate system unification, data resampling, and noise filtering.
[0059] In a specific embodiment of the present invention, the handling of missing values and outliers includes:
[0060] Missing values are caused by GPS signal loss (such as obstruction or tall buildings), resulting in some data point fields being empty (such as missing elevation).
[0061] Minor missing values (<5%): Linear interpolation method to complete (e.g., using the average elevation of two normal points before and after to fill the missing values);
[0062] Large number of missing data (>10%): Delete the data segment directly (to avoid the accumulation of interpolation errors).
[0063] Outliers are caused by sensor errors or sudden interference that cause data jumps (such as the elevation suddenly changing from 100m to 500m, or the latitude and longitude exceeding the actual flight area).
[0064] Use the 3σ principle to identify outliers: Calculate the mean (μ) and standard deviation (σ) of a field (such as elevation), and remove outliers with values > μ + 3σ or < μ - 3σ.
[0065] Spatial rationality verification: Combine the flight area boundaries (such as the pre-set latitude 30~30.5°N, longitude 120°~120.5°E) to delete abnormal data points that exceed the boundaries.
[0066] The coordinate system is uniformly based on the coordinate system of the 3D scene data, and the UAV data is transformed using professional tools (such as ArcGIS and QGIS).
[0067] WGS84 latitude and longitude → UTM plane coordinates (e.g., UTM Zone 50N);
[0068] Geodetic height → Normal height (If the 3D scene uses elevation, the "geoid difference" of the area needs to be subtracted to ensure that the elevation benchmark is consistent).
[0069] Furthermore, data resampling includes spatial resampling and temporal resampling;
[0070] Spatial resampling is based on the pixel grid of the 3D scene. For all UAV data points in the grid, the average elevation / position of one grid is calculated using the mean method or the median method to ensure that one scene grid corresponds to one UAV data point.
[0071] Temporal resampling (if temporal matching is required): If the 3D scene is temporal data (such as terrain changes at different times), extract the average value of the data within the corresponding time period of the drone according to the scene time interval (such as 1 hour / frame) to achieve temporal-spatial dual alignment.
[0072] Specifically, the 3D scene data includes: terrain model, scene mesh, and texture map;
[0073] The preprocessing of the 3D scene data includes: data lightweighting and redundancy removal, geometric error correction, coordinate and resolution unification, and invalid region masking.
[0074] The preprocessed UAV flight data and the preprocessed 3D scene data are subjected to consistency verification.
[0075] Specifically, step 3, which calculates the topology matching degree, includes: calculating the elevation difference and spatial overlap based on the preprocessed UAV flight data and 3D scene data, integrating the elevation difference and spatial overlap according to weights, and obtaining the final topology matching degree.
[0076] In a specific embodiment of the present invention, after completing the preprocessing of two types of data, a consistency check is performed to ensure that the requirements for subsequent calculations are met. Specifically, this includes:
[0077] Spatial range verification: Check whether the coverage area of the UAV data falls completely within the coverage area of the 3D scene data (±5% edge deviation is allowed, and the excess part needs to be cropped).
[0078] Accuracy verification: Randomly select 100 sampling points and check the "coordinate deviation" (e.g., UTM coordinate X / Y deviation < 0.5 meters) and "elevation datum deviation" (e.g., root mean square error RMSE < 0.3 meters) of the two types of data.
[0079] Format validation: Confirm that both types of data are in a "readable, universal format" (e.g., GeoJSON for drones and LAS / OBJ for 3D scenes), and that field / attribute information is complete and without missing information.
[0080] Once the verification is successful, the data can be used to calculate the topology matching degree.
[0081] Specifically, the process of calculating the elevation difference includes:
[0082] Spatial alignment: Using the coordinate system of the 3D scene data as a reference, the UAV flight data is mapped to the same coordinate system through coordinate transformation to ensure that the spatial positions of the two correspond.
[0083] Sampling point selection: Select N sampling points at fixed intervals within the overlapping area of the fused scene of UAV flight data and 3D scene data to cover the entire data area;
[0084] Calculate elevation difference: Calculate the elevation difference of each point, and use the average elevation difference or the maximum elevation difference to characterize the overall vertical matching degree.
[0085] Specifically, the process of calculating spatial overlap includes:
[0086] Extracting the region boundary: Extracting the minimum bounding rectangle S of the flight trajectory from the UAV flight data. 无人机 Extract the corresponding effective scene boundary S from the 3D scene data 场景 ;
[0087] Calculate the overlapping area and the total area: Use spatial geometry algorithms to calculate the intersection area S between the drone region and the 3D scene region. 重叠 Calculate the area S of the union of the two sets. 并集 =S 无人机 +S 场景 -S 重叠 ;
[0088] Calculate spatial overlap: S 空间重叠度 =S 重叠 / S 并集 ;
[0089] The topology matching degree is obtained by weighting the elevation difference and spatial overlap.
[0090] Specifically, the calculation process for the fusion weights includes:
[0091] The elevation difference and spatial overlap in the topological matching degree are extracted and used as constraints input into the joint probability density association model.
[0092] Based on the mean, variance, and joint covariance of the observation errors of the preprocessed UAV flight data and 3D scene data, a joint probability density function of a two-dimensional normal distribution is constructed, and the initial probability density value of each spatial sampling point is calculated.
[0093] The initial probability density value is weighted and corrected using the topological matching degree to enhance the probability contribution of high-matching data.
[0094] The corrected probability density values are normalized to obtain the fusion weights corresponding to each sampling point. The fusion weights range from [0, 1].
[0095] Specifically, the expression for the joint probability density association model is:
[0096] ;
[0097] Where x represents the preprocessed UAV flight data observation at the target sampling point, y represents the preprocessed 3D scene data observation at the same target sampling point, f(x,y) is the joint probability density value of the two types of data at that sampling point; μ1 is the mean elevation error of the UAV flight data; σ1 2 σ² represents the elevation error variance of the UAV flight data; μ² represents the mean elevation error of the 3D scene data; σ² represents the mean elevation error of the UAV flight data. 2 ρ represents the elevation error variance of the 3D scene data; ρ represents the joint error covariance, which is determined by analyzing the error correlation between the two types of data in historical matching samples.
[0098] Specifically, the formula for weighting and correcting the initial probability density value using the topological matching degree is as follows:
[0099] F'(x,y)=f(x,y)×[0.5+0.5×M];
[0100] Where 0.5 is the basic weight coefficient, and 0.5×M is the contribution weight of topology matching degree; M is the topology matching degree.
[0101] In a specific embodiment of the present invention, this embodiment takes the preliminary surveying and mapping scenario of the renovation of an old residential area in a certain city as the background. It is necessary to integrate the on-site surveying and mapping data of UAVs with the existing three-dimensional planning model data of the residential area to provide accurate spatial reference for the renovation plan design. The specific implementation process is as follows:
[0102] Sampling point selection: In the overlapping areas of the community, 1,000 sampling points were evenly selected in a 2m x 2m grid to cover key areas such as building tops, road surfaces, and green areas to ensure that the sampling points are representative.
[0103] Elevation difference calculation: For each sampling point, extract the elevation value (H1) of the UAV DSM and the elevation value (H2) of the 3D scene elevation raster map, and calculate the elevation difference ΔH=|H1-H2|.
[0104] Spatial overlap calculation: Using ArcGIS software, extract the minimum bounding rectangle of the UAV-mapped area (area S1 = 82,000 square meters) and the minimum bounding rectangle of the area covered by the 3D scene model (area S2 = 85,000 square meters), and calculate the intersection area S. 重叠 =79,000 square meters, spatial overlap = S 重叠 / (S1+S2-S 重叠 =7.9 / (8.2+8.5-7.9)≈0.89.
[0105] Comprehensive topology matching degree calculation: Setting the weights for elevation difference and spatial overlap at 0.5, substituting into the formula: Topology matching degree M = (1 - average elevation difference) × 0.5 + spatial overlap × 0.5 = (1 - 0.23) × 0.5 + 0.96 × 0.5 ≈ 0.865. 0.865 > 0.6, therefore it can be proven that elevation difference and spatial overlap are effectively fused. Furthermore, based on the preprocessed data, statistics show that the mean elevation error of the UAV DSM data is μ1 = 0.15 meters, and the error variance σ1... 2 =0.03; The mean elevation error of the 3D scene data is μ2 = 0.10 meters, and the variance of the error is σ2. 2 =0.02; based on historical data from similar projects, the joint error covariance ρ of the two types of data is 0.25.
[0106] With the topological matching degree M=0.865 as a constraint, for each sampling point, substitute the formula for the joint probability density of the two-dimensional normal distribution:
[0107]
[0108] When x=0.2 and y=0.15, the final f(x,y)≈6.21, indicating that the probability of this sampling point belonging to the same associated target as both types of data is extremely high at the statistical level. The final fusion weight will also reach 0.865 due to the high probability density. 0.5 + 0.95 0.5 = 0.9075; (assuming the probability density is normalized to 0.95), which is far beyond the effective fusion threshold (such as the 0.6 set above), directly supporting the decision that this sampling point can participate in effective fusion.
[0109] Finally, QGIS software was used for visualization.
[0110] Specifically, a UAV multi-source heterogeneous data fusion system based on a 3D scene includes:
[0111] Data acquisition module: Acquires drone flight data and 3D scene data;
[0112] Preprocessing module: performs preprocessing on the UAV flight data and 3D scene data respectively;
[0113] Topology matching degree calculation module: Calculates the topology matching degree based on the preprocessed UAV flight data and 3D scene data;
[0114] Model calculation module: used to calculate the fusion weights of UAV flight data and 3D scene data based on the topological matching degree and the joint probability density association model;
[0115] Visualization module: Used to fuse UAV flight data and 3D scene data according to fusion weights, and visualize the fusion results.
[0116] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0117] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for fusing multi-source heterogeneous data from UAVs based on 3D scenes, characterized in that, include: Step 1: Acquire drone flight data and 3D scene data; Step 2: Preprocess the UAV flight data and 3D scene data respectively; Step 3: Calculate the topology matching degree based on the preprocessed UAV flight data and 3D scene data; Step 4: Calculate the fusion weights of UAV flight data and 3D scene data based on the topology matching degree and the joint probability density association model; Step 5: Fuse the UAV flight data and 3D scene data according to the fusion weights, and visualize the fusion results; The calculation process for the fusion weights includes: The elevation difference and spatial overlap in the topological matching degree are extracted and used as constraints input into the joint probability density association model. Based on the mean, variance, and joint covariance of the observation errors of the preprocessed UAV flight data and 3D scene data, a joint probability density function of a two-dimensional normal distribution is constructed, and the initial probability density value of each spatial sampling point is calculated. The initial probability density value is weighted and corrected using the topological matching degree to enhance the probability contribution of high-matching data. The corrected probability density values are normalized to obtain the fusion weights corresponding to each sampling point. The fusion weights take values in the range of [0, 1]. The expression for the joint probability density correlation model is: ； Where x represents the preprocessed UAV flight data observation at the target sampling point, y represents the preprocessed 3D scene data observation at the same target sampling point, f(x,y) is the joint probability density value of the two types of data at that sampling point; μ1 is the mean elevation error of the UAV flight data; σ1 2 σ² represents the elevation error variance of the UAV flight data; μ² represents the mean elevation error of the 3D scene data; σ² represents the mean elevation error of the UAV flight data. 2 ρ is the elevation error variance of the 3D scene data; ρ is the joint error covariance, which is determined by analyzing the error correlation between the two types of data in historical matching samples. The formula for weighting and correcting the initial probability density value using the topological matching degree is as follows: F'(x,y)=f(x,y)×[0.5+0.5×M]; Where 0.5 is the basic weight coefficient, and 0.5×M is the contribution weight of topology matching degree; M is the topology matching degree.
2. The method for fusing multi-source heterogeneous data from UAVs based on a three-dimensional scene according to claim 1, characterized in that, The UAV flight data includes GPS latitude and longitude, altitude, flight speed, and sampling timestamp; The preprocessing of UAV flight data includes: data format standardization, handling of missing and outlier values, coordinate system unification, data resampling, and noise filtering.
3. The method for fusing multi-source heterogeneous data from UAVs based on a three-dimensional scene according to claim 2, characterized in that, The 3D scene data includes: terrain model, scene mesh, and texture map; The preprocessing of the 3D scene data includes: data lightweighting and redundancy removal, geometric error correction, coordinate and resolution unification, and invalid region masking. The preprocessed UAV flight data and the preprocessed 3D scene data are subjected to consistency verification.
4. The method for fusing multi-source heterogeneous data from UAVs based on a three-dimensional scene according to claim 3, characterized in that, The process of calculating the topology matching degree in step 3 includes: calculating the elevation difference and spatial overlap based on the preprocessed UAV flight data and 3D scene data, integrating the elevation difference and spatial overlap according to weights, and obtaining the final topology matching degree. Topology matching degree calculation: Set the weight of elevation difference to 0.5 and the weight of spatial overlap to 0.5, and substitute them into the formula: Topology matching degree M = (1 - average elevation difference) × 0.5 + spatial overlap × 0.
5.
5. The method for fusing multi-source heterogeneous data from UAVs based on a three-dimensional scene according to claim 4, characterized in that, The process of calculating elevation difference includes: Spatial alignment: Using the coordinate system of the 3D scene data as a reference, the UAV flight data is mapped to the same coordinate system through coordinate transformation to ensure that the spatial positions of the two correspond. Sampling point selection: Select N sampling points at fixed intervals within the overlapping area of the fused scene of UAV flight data and 3D scene data to cover the entire data area; Calculate elevation difference: Calculate the elevation difference of each point, and use the average elevation difference or the maximum elevation difference to characterize the overall vertical matching degree.
6. The method for fusing multi-source heterogeneous data from UAVs based on a three-dimensional scene according to claim 4, characterized in that, The process of calculating spatial overlap includes: Extracting the region boundary: Extracting the minimum bounding rectangle S of the flight trajectory from the UAV flight data. 无人机 Extract the corresponding effective scene boundary S from the 3D scene data 场景 ; Calculate the overlapping area and the total area: Use spatial geometry algorithms to calculate the intersection area S between the drone region and the 3D scene region. 重叠 Calculate the area S of the union of the two sets. 并集 =S 无人机 +S 场景 -S 重叠 ; Calculate spatial overlap: S 空间重叠度 =S 重叠 / S 并集 ; The topology matching degree is obtained by weighting the elevation difference and spatial overlap.
7. A UAV multi-source heterogeneous data fusion system based on a 3D scene, characterized in that, include: Data acquisition module: Acquires drone flight data and 3D scene data; Preprocessing module: performs preprocessing on the UAV flight data and 3D scene data respectively; Topology matching degree calculation module: Calculates the topology matching degree based on the preprocessed UAV flight data and 3D scene data; Model calculation module: used to calculate the fusion weights of UAV flight data and 3D scene data based on the topological matching degree and the joint probability density association model; The calculation process for the fusion weights includes: The elevation difference and spatial overlap in the topological matching degree are extracted and used as constraints input into the joint probability density association model. Based on the mean, variance, and joint covariance of the observation errors of the preprocessed UAV flight data and 3D scene data, a joint probability density function of a two-dimensional normal distribution is constructed, and the initial probability density value of each spatial sampling point is calculated. The initial probability density value is weighted and corrected using the topological matching degree to enhance the probability contribution of high-matching data. The corrected probability density values are normalized to obtain the fusion weights corresponding to each sampling point. The fusion weights take values in the range of [0, 1]. The expression for the joint probability density correlation model is: ； Where x represents the preprocessed UAV flight data observation at the target sampling point, y represents the preprocessed 3D scene data observation at the same target sampling point, f(x,y) is the joint probability density value of the two types of data at that sampling point; μ1 is the mean elevation error of the UAV flight data; σ1 2 σ² represents the elevation error variance of the UAV flight data; μ² represents the mean elevation error of the 3D scene data; σ² represents the mean elevation error of the UAV flight data. 2 ρ is the elevation error variance of the 3D scene data; ρ is the joint error covariance, which is determined by analyzing the error correlation between the two types of data in historical matching samples. The formula for weighting and correcting the initial probability density value using the topological matching degree is as follows: F'(x,y)=f(x,y)×[0.5+0.5×M]; Where 0.5 is the basic weight coefficient, and 0.5×M is the contribution weight of topology matching degree; M is the topology matching degree. Visualization module: Used to fuse UAV flight data and 3D scene data according to fusion weights, and visualize the fusion results.