Building group three-dimensional reconstruction and radiation dose field visualization method based on aerial images

Abstract

The invention discloses a building group three-dimensional reconstruction and radiation dose field visualization method based on aerial images, relates to the technical field of image processing, enables the timeliness of a new visualization method to be higher, and solves the core problems of low manual modeling efficiency and poor geometric transformation universality in the prior art. Comprising the following steps: S1, collecting a multi-view image of a building group at a preset site through an unmanned aerial vehicle; s2, sparse point cloud reconstruction is carried out on the image obtained in the step S1, and a model is optimized; s3, format conversion is carried out on the sparse point cloud model and the camera pose data obtained in the step S2, dense point cloud reconstruction, grid model reconstruction and texture mapping are carried out in sequence, and a high-precision building group three-dimensional model is generated; s4, converting the three-dimensional model obtained in the step S3 into a CAD geometric model, and eliminating geometric overlapping defects in the three-dimensional model; s5, the CAD model is converted into a Monte Carlo program through a BRep-to-CSG conversion engine; and S6, configuring radiation transport parameters, executing Monte Carlo calculation, and outputting three-dimensional distribution of the radiation dose field.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to a method for three-dimensional reconstruction of building complexes and visualization of radiation dose fields based on aerial images. Background Technology

[0002] Drone aerial imagery 3D reconstruction technology can accurately reconstruct the 3D structure of building complexes by analyzing the geometry, texture, and spatial relationships in multi-view images, providing an efficient solution for the digitization of physical scenes. This technology has become one of the important technical approaches for 3D modeling of building complexes.

[0003] The Monte Carlo method, as a numerical simulation technique based on random sampling, has irreplaceable advantages in calculating nuclear radiation dose fields. By simulating the transport process of particles (neutrons / photons, etc.) in a medium, it can accurately obtain the radiation dose distribution, which is a core technical support for nuclear accident emergency decision-making.

[0004] Studies have shown (e.g., Terada H et al., Journal of Environmental Radioactivity 2012) that buildings have a significant blocking and guiding effect on the spread of radioactive materials. The structural damage to the plant caused by the hydrogen explosion in the Fukushima nuclear accident further demonstrates that the real-time nature of the building geometry at the accident site directly determines the accuracy of dose field calculations. Current technology faces three major bottlenecks: 1. Timeliness deficiency: When relying on Geographic Information Systems (GIS) or satellite imagery to reconstruct building complexes, the data update cycle is long (usually several hours to several days), making it impossible to capture real-time structural changes at the accident site (such as building damage, new barriers, etc.).

[0005] 2. Compatibility Barriers: There is an inherent gap between existing 3D models and Monte Carlo geometric input formats (such as CSG, surface description): BRep models generated by GIS contain a large number of curved surfaces and non-manifold topologies; manually writing Monte Carlo input cards takes several weeks and is prone to errors; geometric defects such as overlapping surfaces and gaps in model conversion lead to a calculation failure rate of >60% (according to SuperMC test data). 3. Automated chain break: In the process from image reconstruction → CAD model → Monte Carlo geometry, manual intervention is required in key steps such as geometric repair and parameter adjustment, which leads to delays in emergency response (>3 hours). Summary of the Invention

[0006] I. Technical problems to be solved This invention addresses the shortcomings of existing technologies by proposing a method for 3D reconstruction of building complexes and visualization of radiation dose fields based on aerial images. It achieves fully automated processing from UAV images to radiation dose fields, making the established visualization method more timely. At the same time, it solves the core problems of low efficiency of manual modeling and poor universality of geometric transformation in existing technologies.

[0007] II. Specific Technical Solutions A method for 3D reconstruction of building complexes and visualization of radiation dose fields based on aerial images includes the following steps: S1, acquiring a group of multi-view oblique photographic images of the building complex using a UAV at a preset spherical coordinate system location, and storing the obtained building images in a set file format; S2, reconstructing sparse point clouds from the image groups acquired in step S1 and optimizing the model scale, outputting the point cloud model and camera position and pose data for each building image group, and then storing the image model and pose data in a set format; S3, sequentially converting the sparse point cloud model and camera pose data obtained in step S2 into different formats, and then sequentially performing dense point cloud reconstruction, mesh model reconstruction, and texture mapping to generate a high-precision 3D model of the building complex, and saving the 3D model in a set format; S4, converting the 3D model obtained in step S3 into a CAD geometric model and eliminating geometric overlap defects within the 3D model; S5, converting the CAD model into a Monte Carlo program's solid geometry input card using the BRep-to-CSG conversion engine; S6, configuring radiation transport parameters and performing Monte Carlo calculations to output the 3D distribution of the radiation dose field of the building complex.

[0008] Implementation principle and working principle: In step S1, this scheme uses a spherical coordinate system for image acquisition, which is beneficial for constructing an optimal image dataset, reducing the failure rate and image redundancy in subsequent modeling, and eliminating reliance on human experience. Step S2 generates a spatially accurate framework with three non-collinear control points inputting their real coordinates, such as GPS markers. Least squares optimization binds the relative coordinate system to the absolute physical coordinate system, effectively reducing model size errors. Step S3 constructs a lightweight model that is easy to process in engineering, solving the pain point that point cloud data cannot be directly used for CAD conversion. In step S4, CAD conversion and geometric repair significantly reduce the failure rate of CSG conversion caused by geometric defects. Through BRep-to-CSG conversion, the time taken is reduced from several weeks to several hours, making it much faster. Finally, through S4 and S5, the nuclear accident dose field analysis cycle is shortened from several days in traditional methods to several hours, and real-time reconstruction of damaged buildings is supported.

[0009] Preferably, the drone location in step S1 adopts a spherical coordinate system layout, with the center of the building complex as the origin and a radius set. R satisfyR > max (建筑物到原点距离) Pitch angle φ Take {20°, 40°, 60°, 80°} as the azimuth angle. θ The images cover 0°~360° at 20° intervals, with a total of 60~90 images. The advantage of this preferred method is that, while ensuring the accuracy of the 3D reconstruction of the building, the number of images is compressed to the range of optimal computational efficiency by optimizing the distribution of shooting positions, thus avoiding redundant data that would cause a surge in reconstruction time.

[0010] As a preferred option, step S2 uses COLMAP to achieve feature matching and sparse reconstruction, and optimizes the model scale using the real coordinates of at least three non-collinear control points. The advantage of this preferred option is that it can eliminate the scale uncertainty of UAV image reconstruction and ensure that the subsequent dose field calculation has physically quantifiable spatial accuracy.

[0011] Preferably, step S3 uses OpenMVS to perform five-stage processing on the acquired coefficient point cloud, model, and camera pose data, including sparse point cloud format conversion, dense point cloud reconstruction, Delaunay triangulation to generate a coarse mesh, mesh optimization to generate a watertight model, and texture mapping based on the original image. The beneficial effect of this preferred method is that it constructs a lightweight mesh model that preserves the details of the building surface, providing topologically complete input for geometric transformation.

[0012] Preferably, step S4 uses SuperMC to repair geometric defects in the acquired CAD geometric model, eliminating overlapping and gaps in the CAD geometric model. The beneficial effect of this preferred method is to solve the problem of failure in BRep-to-CSG conversion caused by geometric defects and improve the conversion success rate of complex building complex models.

[0013] As a preferred option, step S5 uses the McCAD tool to perform a four-layer transformation, including: inputting and preprocessing the BRep model, performing geometric decomposition and convexification, constructing a CSG Boolean operation tree, and finally generating a geometric descriptor that can be parsed by the Monte Carlo program. The advantage of this preferred option is that it breaks through the technical barrier of traditionally manually writing Monte Carlo geometric input cards and realizes the automatic parametric expression of complex geometry of building complexes.

[0014] Preferably, the radiation transport parameters in step S6 include: source term energy spectrum, which is Maxwell energy spectrum and nuclide fission energy spectrum; source term location, i.e., spatial distribution of radioactive sources; boundary, particle number; and dose field calculation using Monte Carlo method. The beneficial effect of this preferred option is that, through a standardized parameter configuration framework, the dose field calculation results are ensured to have physical comparability. Attached Figure Description

[0015] Figure 1This is a flowchart illustrating the method for 3D reconstruction of building complexes and visualization of radiation dose fields based on aerial images, as presented in this invention.

[0016] Figure 2 This is a partial schematic diagram of the method for 3D reconstruction of building complexes and visualization of radiation dose field based on aerial images according to the present invention.

[0017] Figure 3 The 3D model of the building complex in .ply format generated for an embodiment of the present invention.

[0018] Figure 4 This is a 3D CAD geometric model generated in an embodiment of the present invention.

[0019] Figure 5 - Figure 16 This is a schematic diagram illustrating the computer calculation and analysis performed in step S6. Detailed Implementation

[0020] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0021] like Figure 1-16 As shown: A method for 3D reconstruction of building complexes and visualization of radiation dose fields based on aerial images includes the following steps: S1. Collect multi-view oblique photographic images of the building complex using a drone at a preset spherical coordinate system location, and store the obtained building images in a set file format. In step S1, a spherical coordinate system is first constructed with the center of the building complex as the origin, and the radius is set. R satisfy R > max (建筑物到原点距离) , specifically R = max (建筑物到原点距离)+10m Furthermore, this radius must ensure unobstructed views and a suitable pitch angle. φ Take {20°, 40°, 60°, 80°} as the azimuth angle. θ The system covers the area from 0° to 360° at 20° intervals, capturing a total of 72 images, which are stored in .png or .jpg format. This step uses a spherical coordinate system for image acquisition, which helps to build an optimal image dataset, reduce the failure rate and image redundancy in subsequent modeling, and eliminate the reliance on human experience. At the same time, while ensuring the accuracy of the 3D reconstruction of the building, the number of images can be compressed to the optimal range of computational efficiency by optimizing the distribution of shooting points, avoiding the surge in reconstruction time caused by redundant data.

[0022] In this specific implementation, first create a main folder for this embodiment within the COLMAP program folder directory, and then create images and sparse subfolders within the main folder. Use a drone to take multi-view photos of the center of the building complex using the set point selection method, and save the images in .jpg format in the images folder.

[0023] S2. The image groups obtained in step S1 are reconstructed into sparse point clouds and the model scale is optimized. Specifically, the open-source vision software COLMAP is used to perform feature matching and sparse reconstruction. The point cloud model and the camera position and pose data of each building image group are output. Then, the point cloud model and pose data of the image group are stored in a set .bin format file. The specific process of COLMAP is feature extraction, feature matching and sparse reconstruction. COLMAP can generate a spatially accurate framework. The real coordinates are represented by three non-collinear control points in the camera pose data, such as GPS markers. The relative coordinate system is then bound to the absolute physical coordinate system by least squares optimization, which can effectively reduce the model size error. It can eliminate the scale uncertainty of UAV image reconstruction and ensure that the subsequent dose field calculation has physically quantifiable spatial accuracy.

[0024] In this step, run COLMAP.bat. Select File→New Project, then select New to create a .db format project file in the main folder. Then, select Select to save the images folder of the image group in this example. Select Processing→Feature extraction to perform feature extraction in PINHOLE mode with default parameters. Next, select Processing→Feature matching to perform feature matching with default parameters. Finally, select Reconstruction→Start reconstruction to start sparse point cloud reconstruction with default parameters. After sparse reconstruction is complete, open the mode view, right-click on a feature point and select Add Control Point. Enter the coordinates of the point in the real world, add at least three control points, and then click Run Bundle Adjustment and check Use ControlPoints. After optimization is complete, select File→Export model, index the sparse folder, and export the camera pose and sparse point cloud data to the sparse folder.

[0025] S3. The sparse point cloud model and camera pose data obtained in step S2 are converted in sequence, and then dense point cloud reconstruction, mesh model reconstruction and texture mapping are performed in sequence to generate a high-precision 3D model of the building complex. The 3D model is then saved in the set format, i.e. .ply format. Step S3 uses OpenMVS to perform five-stage processing on the acquired coefficient point cloud, model, and camera pose data, including sparse point cloud format conversion, dense point cloud reconstruction, Delaunay triangulation to generate a coarse mesh, mesh optimization to generate a watertight model, and texture mapping based on the original image. The beneficial effect of this preferred method is that it constructs a lightweight mesh model that preserves the details of the building surface, providing topologically complete input for geometric transformation. In practice, use the command prompt to enter the format as follows: Figure 3 The command is as follows: -w specifies the address of the program running the specific step of format conversion, dense point cloud reconstruction, coarse mesh model reconstruction, high-precision mesh model reconstruction, or texture mapping; -w specifies the execution address of this command; -i specifies the address of the input file for this command; and -o specifies the name and address of the output file for this command.

[0026] S4. Convert the 3D model obtained in step S3 into a 3D CAD geometric model in .step format, and eliminate geometric overlap defects in the 3D model. Step S4 specifically involves using SuperMC to eliminate geometric overlaps or repair geometric defects in the acquired 3D CAD geometric model, removing overlapping surfaces and gaps. Specifically, this involves importing the .ply file into FreeCAD software, using the Part Workbench module to convert the mesh into a geometric model, selecting Convert to Solid to generate a solid, checking the solid's validity, and finally exporting it as a .step file. This addresses the failure issue in BRep-to-CSG conversion caused by geometric defects, thereby improving the conversion success rate of complex building complex models.

[0027] S5. The CAD model is converted into a Monte Carlo program's constructive solid geometry input card through the BRep-to-CSG conversion engine. Step S5 uses the McCAD tool to perform a four-layer conversion, including: inputting and preprocessing the BRep model, performing geometric decomposition and convexization, constructing a CSG Boolean operation tree, and finally generating a Monte Carlo program-resolvable geometric descriptor. This breaks through the technical barrier of traditionally manually writing Monte Carlo geometry input cards and realizes the automatic parametric expression of complex geometry of building complexes. In step S5, create a new folder anywhere to serve as the working directory. In the command line window containing the solid model's STP file, enter the command "McCAD". This will generate a file named "McCADInputConfig.i", which can be opened directly with Notepad. The file contains decomposition and transformation parameter commands that McCAD can read. By modifying the commands in the file, such as... Input section: units = cm, and the units of the model can be changed by modifying cm.

[0028] inputFileName = <material name> - <material density>.stp (The material name is abbreviated, and the material density is converted to scientific notation (in grams per cubic centimeter) according to the numerical value of each material.) Conversion section: void Generation = false: Disables the void element generation function to simplify the model.

[0029] compoundsSingleCell = false: Decomposes the complex in the STEP file into independent sub-entity units to avoid geometric conflicts in Monte Carlo simulations caused by forced merging.

[0030] minVoidVolume = 1.0 [cm³]: Sets the minimum threshold for void volume, filtering out tiny voids smaller than this value, reducing computational redundancy and improving simulation efficiency.

[0031] maxSolidsPerVoidCell = 20: Limits the maximum number of entities contained in a single void cell, balancing code quality and computational accuracy.

[0032] Output format configuration: MCcode = OpenMC: Specifies the target Monte Carlo code as OpenMC, ensuring the generated CSG syntax (using OpenMC as an example).

[0033] Finally, run the "McCAD RUN" command. The system will perform geometric analysis, decomposition, and transformation operations in sequence, ultimately generating five core files: STEP parsing file, decomposed model file, Monte Carlo input file, volume statistics file, and void mapping file, thus realizing the conversion from BRep to CSG.

[0034] S6. Configure radiation transport parameters and perform Monte Carlo calculations to output the three-dimensional distribution of radiation dose field of the building complex; specifically, set the relevant radiation transport calculation parameters of the Monte Carlo method for the input card that OpenMC can calculate in step 5, and then use Monte Carlo transport calculations to obtain the radiation dose field under complex building complex conditions.

[0035] The radiation transport parameters in step S6 include: source term energy spectrum, which is the Maxwell energy spectrum and the nuclide fission energy spectrum; source term location, i.e., the spatial distribution of the radioactive source; boundary and particle number; and dose field calculation using the Monte Carlo method. A standardized parameter configuration framework is used to ensure that the dose field calculation results are physically comparable.

[0036] In step S6, the source term uses the 300k Maxwell energy spectrum distribution and a common nuclide fission spectrum distribution. The source is located 100 meters above the ground along the z-axis. The cuboid calculation boundary is approximately -100 meters to +100 meters along the x-axis, -100 meters to +100 meters along the y-axis, and -1 meter to +200 meters along the z-axis. The factory building height is approximately 20 meters. The source is set as a point source, with each particle having the same emission probability. In this example, the number of particles is calculated to be 500 million. The calculation method uses neutron-photon combined transport, and the detectors must be set to neutron and photon types separately. Finally, the simulation results are obtained for the neutron and photon radiation dose distribution along the z-axis, inside the factory building, and at Z=0 under the conditions of this building complex.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims.

Claims

1. A method for 3D reconstruction of building complexes and visualization of radiation dose fields based on aerial images, characterized in that, Including the following steps: S1. Collect multi-view oblique photographic images of the building complex using a drone at a preset spherical coordinate system location, and store the obtained building images in a set file format. S2. The image groups obtained in step S1 are reconstructed into sparse point clouds and the model scale is optimized. The point cloud model and the camera position pose data of each building image group are output. The image model and pose data are then stored in the set format. S3. The sparse point cloud model and camera pose data obtained in step S2 are converted in order, and then dense point cloud reconstruction, mesh model reconstruction and texture mapping are performed in sequence to generate a high-precision 3D model of the building complex. The 3D model is then saved in the set format. S4. Convert the 3D model obtained in step S3 into a CAD geometric model and eliminate geometric overlap defects in the 3D model. S5. Convert the CAD model into a solid geometry input card for the Monte Carlo program using the BRep-to-CSG conversion engine; S6. Configure radiation transport parameters and perform Monte Carlo calculations to output the three-dimensional distribution of radiation dose field of the building complex.

2. The method for three-dimensional reconstruction of building complexes and visualization of radiation dose field based on aerial images according to claim 1, characterized in that: In step S1, the UAV location adopts a spherical coordinate system layout, with the center of the building complex as the origin. The radius R is set to satisfy R>max (distance from the building to the origin). The pitch angle φ is {20°, 40°, 60°, 80°}, and the azimuth angle θ covers 0°~360° in 20° intervals. The total number of images is 60~90.

3. The method for three-dimensional reconstruction of building complexes and visualization of radiation dose field based on aerial images according to claim 1, characterized in that: Step S2 uses COLMAP to achieve feature matching and sparse reconstruction, and optimizes the model scale using the real coordinates of at least three non-collinear control points.

4. The method for three-dimensional reconstruction of building complexes and visualization of radiation dose field based on aerial images according to claim 1, characterized in that: Step S3 uses OpenMVS to perform five-stage processing on the acquired coefficient point cloud, model, and camera pose data, including sparse point cloud format conversion, dense point cloud reconstruction, Delaunay triangulation to generate a coarse mesh, mesh optimization to generate a watertight model, and texture mapping based on the original image.

5. The method for three-dimensional reconstruction of building complexes and visualization of radiation dose field based on aerial images according to claim 1, characterized in that: Step S4 uses SuperMC to repair geometric defects in the acquired CAD geometric model, eliminating overlapping and gaps in the CAD geometric model.

6. The method for three-dimensional reconstruction of building complexes and visualization of radiation dose field based on aerial images according to claim 1, characterized in that: Step S5 uses the McCAD tool to perform a four-layer transformation, including: inputting and preprocessing the BRep model, performing geometric decomposition and convexization, constructing a CSG Boolean operation tree, and finally generating a geometric descriptor that can be parsed by the Monte Carlo program.

7. The method for three-dimensional reconstruction of building complexes and visualization of radiation dose field based on aerial images according to claim 1, characterized in that: The radiation transport parameters in step S6 include: source term energy spectrum, which is the Maxwell energy spectrum and the nuclide fission energy spectrum; source term location, i.e., the spatial distribution of the radioactive source; boundary, particle number; and dose field calculation using the Monte Carlo method.

Images