A deep learning-based multi-source data high-precision fusion method

Abstract

The application discloses a kind of multi-source data high-precision fusion methods based on deep learning.The step of epipolar line calculation in the application can ensure the consistency of point cloud data in geometry.Epipolar line is the straight line connecting a point in three-dimensional space and its corresponding projection point on the image, therefore, the correct calculation of epipolar line provides accurate geometric constraints for subsequent point cloud processing.The calculation of epipolar line parameters provides directional information for point cloud denoising and smoothing processing.In the denoising process, filtering can be performed along the epipolar line direction, which can more effectively identify and remove outliers while retaining useful structural information.By pre-computing epipolar lines, the computational load in subsequent processing steps can be reduced.In the point cloud registration and fusion process, epipolar line information can be used to quickly locate corresponding points, thereby speeding up the convergence speed of the algorithm.Provide more accurate basic data for subsequent three-dimensional reconstruction.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of multi-source data fusion, and specifically relates to a multi-source data high-precision fusion method based on deep learning. BACKGROUND

[0002] With the rapid development of computer vision, machine learning and deep learning, three-dimensional modeling technology has been widely used in virtual reality (VR), augmented reality (AR) and game development. Traditional three-dimensional modeling methods usually rely on manual creation or scanning devices to obtain point cloud data, and then manual processing is performed to construct three-dimensional models. However, this method has problems such as low efficiency, high cost and difficulty in large-scale scene modeling. In recent years, researchers have begun to explore the use of multi-source data for automated three-dimensional modeling technology. These multi-source data include photo sequences, laser radar point clouds and other sensor data from different perspectives. By effectively fusing these data sources, the accuracy and efficiency of three-dimensional modeling can be significantly improved.

[0003] However, some existing technologies only consider a single data source, such as using only photo sequences or only laser radar point clouds for modeling. This single-source data processing method often fails to fully utilize the information carried by other types of data, resulting in less accurate three-dimensional models. SUMMARY

[0004] The purpose of the present application is to solve the above-mentioned problems, and to provide a multi-source data high-precision fusion method based on deep learning.

[0005] The technical solution adopted by the present application is as follows: a multi-source data high-precision fusion method based on deep learning, the method comprising the following steps:

[0006] S1: input a sequence of images, collect and process the original data of multiple sensors, including laser point clouds and camera images, to form a series of image data;

[0007] S2: solve relative orientation elements and feature points, for the data obtained by each sensor, respectively solve the relative orientation elements and extract the feature points to establish a preliminary spatial geometric relationship;

[0008] S3: perform epipolar line calculation, calculate the parameters of the epipolar line according to the relative orientation elements, for subsequent point cloud denoising and smoothing processing;

[0009] S4: perform point cloud denoising and smoothing, denoise and smooth the laser point cloud data to improve the accuracy of the data;

[0010] S5: Perform point cloud smoothing, further smooth the point cloud data, reduce noise interference;

[0011] S6: Coordinate transformation is performed on the data of different sensors to make them in the same coordinate system, facilitating subsequent fusion processing;

[0012] S7: Combine the point cloud fusion processing algorithm of improved SAC-IA algorithm and improved ICP algorithm, use the improved random sample consensus iterative closest point algorithm and the improved iterative closest point algorithm to fuse the point cloud data, to further improve the accuracy and integrity of the data;

[0013] S8: Use R-tree structure to index and organize point cloud data for quick search and fusion of adjacent point cloud data;

[0014] S9: Through geometric modeling of the fused point cloud data, a three-dimensional model of the object is constructed

[0015] S10: Use the texture information in the camera image to texture map the three-dimensional model, making the model more realistic;

[0016] S11: After all data processing and fusion is completed, output the final complete and high-precision three-dimensional model.

[0017] In a preferred embodiment, in step S1, the unmanned aerial vehicle is equipped with a high-precision laser radar and a high-resolution RGB camera; the input sequence image data includes point cloud data obtained by the laser radar scanning at a rate of 100,000 points per second, and 1920x1080 resolution JPEG images captured by the camera at a rate of 30 frames per second; each data file contains timestamp information, ensuring time synchronization; the laser radar data is stored in LAS format, and the camera image is stored in JPEG format containing EXIF metadata.

[0018] In a preferred embodiment, in step S2, the ORB-SLAM3 algorithm is used to solve the relative orientation elements of the camera; the specific parameters include: the FAST threshold of the ORB algorithm used for feature point extraction is set to 20, the pyramid level is 8 layers, and the size of the feature point descriptor is 32 bytes; when solving the relative orientation elements, the five-point algorithm is used to estimate the camera pose, and the RANSAC iteration number is set to 1000 times, and the inlier threshold is set to 0.005.

[0019] In a preferred embodiment, in step S3, the calculation process includes:

[0020] S3-1: Camera parameter acquisition: First, the intrinsic matrix of the camera needs to be obtained, which includes focal length, principal point coordinates and possible distortion coefficients;

[0021] S3-2 Image pair registration: Next, the relative orientation elements between two or more camera images need to be determined, which is achieved through feature matching and geometric transformation to obtain a rotation matrix and a translation vector;

[0022] S3-3: Calculate epipolar lines: Using the above parameters, calculate the epipolar lines for each camera image;

[0023] The collinear equation is as follows:

[0024] Where f is the focal length; (c_x, c_y) is the principal point coordinate; R = [r_1, r_2,..., r_9] is the rotation matrix; t = [t_x, t_y, t_z] is the translation vector; X, Y, Z are the coordinates of a point in space; x, y are the pixel coordinates on the image.

[0025] In a preferred embodiment, in step S4, the StatisticalOutlierRemoval filter in the PCL library is used for denoising; the specific parameter settings are: set the number of neighbor points to 50 and the standard deviation multiple to 1.0; then, the VoxelGrid filter is used for downsampling and smoothing, and the voxel size is set to 0.01 meters.

[0026] In a preferred embodiment, in step S5, the MovingLeastSquares filter in the PCL library is used for point cloud smoothing; the specific parameters include: set the search radius to 0.03 meters to determine the neighborhood size that affects each point; the order of polynomial fitting is set to 2 to ensure that the smoothing process does not excessively change the original shape of the point cloud.

[0027] In a preferred embodiment, in step S6, the calculation process includes:

[0028] S6-1: Obtain sensor data: First, collect data from different sensors, such as point cloud data from a laser radar and image data from a camera;

[0029] S6-2: Determine the sensor coordinate system: Identify the original coordinate system of each sensor data;

[0030] S6-3: Sensor calibration: Obtain the intrinsic and extrinsic parameters of each sensor through the calibration process; the intrinsic parameters include focal length, principal point coordinate, and distortion coefficient, and the extrinsic parameters include rotation matrix and translation vector;

[0031] S6-14 Coordinate transformation: Use the extrinsic parameters to convert the data of each sensor to a unified reference coordinate system;

[0032] The coordinate transformation formula is:

[0033] Wherein: (X_{ref},Y_{ref},Z_{ref}) is the point coordinate converted to the reference coordinate system; (Xsensor,Ysensor,Zsensor) is the point coordinate in the original sensor coordinate system; R is the rotation matrix, which is used to describe the rotation from the sensor coordinate system to the reference coordinate system; t is the translation vector, which is used to describe the translation from the sensor coordinate system to the reference coordinate system.

[0034] In a preferred embodiment, in the step S7, the improved SAC-IA algorithm is used for initial registration; the parameter settings include: setting the sampling ratio to 0.05 and the distance threshold to 0.1 meters; then, the improved ICP algorithm is applied for fine registration, and the parameters include setting the maximum number of iterations to 100 and the convergence threshold to 0.001 meters;

[0035] In the step S8, the R-tree is used to index the point cloud data; the specific parameters include: setting the maximum number of entries of each R-tree node to 10 and the minimum number of entries of each node to 5; this ensures the effective organization of the point cloud data in space and facilitates fast retrieval.

[0036] In a preferred embodiment, in the step S9, the Poisson reconstruction algorithm is used for geometric modeling; the parameter settings include: setting the grid resolution to 0.005 meters to ensure that the model has high accuracy; and setting the neighborhood size used for normal estimation to 0.01 meters.

[0037] In a preferred embodiment, in the step S10, the MVS texture mapping technology is used; the specific parameters include: setting the minimum parallax angle for texture mapping to 5 degrees and the maximum parallax angle to 75 degrees to ensure the quality of texture mapping; and setting the resolution of the texture image to 1024x1024 pixels.

[0038] In the step S11, the three-dimensional model is exported in OBJ format; the parameter settings include: setting the compression ratio of the model grid to 95% to reduce the file size; using JPEG as the compression format for texture mapping and setting the compression quality to 90%; the exported model contains vertex, texture coordinate and face information, ensuring the integrity and usability of the model in subsequent applications.

[0039] In summary, due to the adoption of the above technical solutions, the present application has the following advantages:

[0040] 1、In the present application, the step of epipolar line calculation can ensure the consistency of point cloud data in geometry. The epipolar line is a straight line connecting a point in three-dimensional space and its corresponding projection point on the image. Therefore, the correct calculation of the epipolar line provides accurate geometric constraints for subsequent point cloud processing. The calculation of epipolar line parameters provides directional information for point cloud denoising and smoothing. In the denoising process, filtering can be performed along the epipolar line direction, which can more effectively identify and remove outliers while preserving useful structural information. By pre-calculating the epipolar line, the computational load in subsequent processing steps can be reduced. In the process of point cloud registration and fusion, the epipolar line information can be used to quickly locate the corresponding points, thereby speeding up the convergence speed of the algorithm. Epipolar line calculation helps to improve the overall quality of the data. By analyzing and processing data in the epipolar line direction, the internal structure and characteristics of the data can be better understood, thereby providing more accurate basic data for subsequent three-dimensional reconstruction. Epipolar line calculation provides additional geometric constraints for the algorithm, which helps to improve the robustness of the entire fusion method. Even in the presence of noise and errors, the algorithm can effectively correct and optimize the data through epipolar line information.

[0041] 2、In the present application, by performing coordinate transformation on data from different sensors, all sensor data is placed in the same coordinate system, which is a prerequisite for effective data fusion. The unified coordinate system ensures spatial alignment between different data sources, providing a basis for subsequent fusion processing. The accuracy of coordinate transformation directly affects the precision of the fusion result. Through accurate coordinate transformation, errors caused by inconsistent coordinate systems can be minimized, thereby improving the accuracy of the final three-dimensional model. Coordinate transformation enables data from different sensors to be interoperable in the same framework, enhancing the flexibility of the data processing flow. This interoperability is particularly important for complex data fusion tasks. Once all data is in the same coordinate system, subsequent data processing procedures will be greatly simplified. Registration, fusion and reconstruction steps can apply unified methods and algorithms without the need for individual adjustments for each sensor. Coordinate transformation promotes the integration of data from different sensors, allowing the strengths of each data source to be complementary. Laser radar data can provide accurate geometric information, while camera images can provide rich texture information, and the effective combination of the two relies on accurate coordinate transformation. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0043] Embodiment:

[0044] A multi-source data high-precision fusion method based on deep learning, the method comprising the following steps:

[0045] S1: Perform sequence image input, collect and process raw data of multiple sensors, including laser point cloud, camera image, etc., to form a series of image data;

[0046] S2: Perform relative orientation element and feature point calculation. For the data obtained by each sensor, the relative orientation element is calculated and the feature points are extracted to establish the preliminary spatial geometric relationship;

[0047] S3: Perform core line calculation. According to the relative orientation element, the parameters of the core line are calculated for subsequent point cloud denoising and smoothing processing;

[0048] S4: Perform point cloud denoising and smoothing. Denoising and smoothing processing is performed on the laser point cloud data to improve the accuracy of the data;

[0049] S5: Perform point cloud smoothing. Further smoothing processing is performed on the point cloud data to reduce noise interference;

[0050] S6: Coordinate conversion is performed on the data of different sensors to make them in the same coordinate system, which is convenient for subsequent fusion processing;

[0051] S7: Combine the improved SAC-IA algorithm and the improved ICP algorithm for point cloud fusion processing. The improved random sample consensus (SAC) iterative closest point (IA) algorithm and the improved iterative closest point (ICP) algorithm are used to fuse the point cloud data to further improve the accuracy and completeness of the data;

[0052] S8: Use R-tree structure to index and organize point cloud data for quick search and fusion of adjacent point cloud data;

[0053] S9: Through geometric modeling of the fused point cloud data, a three-dimensional model of the object is constructed

[0054] S10: Use the texture information in the camera image to perform texture mapping on the three-dimensional model to make the model more realistic;

[0055] S11: After all data processing and fusion are completed, output the final complete and high-precision three-dimensional model.

[0056] In step S1, the UAV is equipped with a high-precision laser radar and a high-resolution RGB camera. The input sequence image data includes point cloud data scanned by the laser radar at a rate of 100,000 points per second, and 1920x1080 resolution JPEG images captured by the camera at a rate of 30 frames per second. Each data file contains timestamp information to ensure time synchronization. Laser radar data is stored in LAS format, and camera images are stored in JPEG format containing EXIF metadata.

[0057] In step S2, the ORB-SLAM3 algorithm is used to solve the relative orientation elements of the camera. The specific parameters include: the FAST threshold of the ORB algorithm used for feature point extraction is set to 20, the pyramid level is 8 layers, and the size of the feature point descriptor is 32 bytes. When solving the relative orientation elements, the five-point algorithm is used to estimate the camera pose, and the RANSAC iteration number is set to 1000 times, and the inlier threshold is set to 0.005.

[0058] In step S3, the calculation process includes:

[0059] S3-1: Camera parameter acquisition: First, the intrinsic matrix of the camera needs to be obtained, which usually includes the focal length (f), the principal point coordinates (c_x, c_y), and possibly the distortion coefficients.

[0060] S3-2 Image pair registration: Then, the relative orientation elements between two or more camera images need to be determined, which is usually obtained by feature matching and geometric transformation to get the rotation matrix (R) and translation vector (t).

[0061] S3-3: Calculate epipolar line: using the above parameters, the epipolar line for each camera image can be calculated;

[0062] The collinear equation is as follows:

[0063] Where f is the focal length; (c_x, c_y) is the principal point coordinate; R = [r_1, r_2,..., r_9] is the rotation matrix; t = [t_x, t_y, t_z] is the translation vector; X, Y, Z are the coordinates of the space point; x, y are the pixel coordinates on the image;

[0064] In step S4, the StatisticalOutlierRemoval filter in the PCL library is used for denoising. The specific parameter settings are: set the number of neighbor points to 50, and the standard deviation multiple to 1.0. Then, the VoxelGrid filter is used for downsampling and smoothing, and the voxel size is set to 0.01 meters.

[0065] In step S5, point cloud smoothing is performed using the Moving Least Squares filter in the PCL library. The specific parameters include: the search radius is set to 0.03 meters to determine the neighborhood size that affects each point; the order of polynomial fitting is set to 2 to ensure that the smoothing process does not excessively change the original shape of the point cloud.

[0066] In step S6, the calculation process includes:

[0067] S6-1: Obtain sensor data: First, collect data from different sensors, such as point cloud data from a lidar and image data from a camera.

[0068] S6-2: Determine the sensor coordinate system: Identify the original coordinate system of each sensor data.

[0069] S6-3: Sensor calibration: Obtain the intrinsic and extrinsic parameters of each sensor through the calibration process. The intrinsic parameters include focal length, principal point coordinates, and distortion coefficients, and the extrinsic parameters include rotation matrix (R) and translation vector (t).

[0070] S6-14 Coordinate transformation: Use the extrinsic parameters to convert the data of each sensor to a unified reference coordinate system;

[0071] The coordinate transformation formula is:

[0072] Where (X_{ref},Y_{ref},Z_{ref}) is the point coordinate converted to the reference coordinate system; (Xsensor,Ysensor,Zsensor) is the point coordinate in the original sensor coordinate system. R is the rotation matrix, which describes the rotation from the sensor coordinate system to the reference coordinate system. t is the translation vector, which describes the translation from the sensor coordinate system to the reference coordinate system.

[0073] In step S7, the improved SAC-IA algorithm is used for initial registration. The parameter settings include: the sampling ratio is set to 0.05, and the distance threshold is set to 0.1 meters. Then, the improved ICP algorithm is applied for fine registration, and the parameters include: the maximum number of iterations is set to 100, and the convergence threshold is set to 0.001 meters;

[0074] In step S8, R-tree is used to index point cloud data. The specific parameters include: the maximum number of entries in each R-tree node is set to 10, and the minimum number of entries in each node is set to 5. This ensures effective organization of point cloud data in space, facilitating fast retrieval

[0075] In step S9, the Poisson reconstruction algorithm is used for geometric modeling. The parameter settings include: the grid resolution is set to 0.005 meters to ensure that the model has high accuracy; the neighborhood size used for normal estimation is set to 0.01 meters.

[0076] In step S10, the MVS texture mapping technique is used. The specific parameters include: setting the minimum parallax angle of texture mapping to 5 degrees and the maximum parallax angle to 75 degrees to ensure the quality of texture mapping; the resolution of the texture image is set to 1024x1024 pixels.

[0077] In step S11, the three-dimensional model is exported in OBJ format. The parameter settings include: the compression ratio of the model grid is set to 95% to reduce the file size; the compression format of the texture mapping uses JPEG, and the compression quality is set to 90%. The exported model contains vertex, texture coordinate and face information, ensuring the integrity and usability of the model in subsequent applications.

[0078] In the present application, the step of performing epipolar line calculation can ensure the geometric consistency of the point cloud data. The epipolar line is a straight line connecting a point in three-dimensional space and its corresponding projection point on the image, so the correct calculation of the epipolar line provides accurate geometric constraints for subsequent point cloud processing. The calculation of the epipolar line parameters provides directional information for point cloud denoising and smoothing. In the denoising process, filtering can be performed along the epipolar line direction, which can more effectively identify and remove outliers while preserving useful structural information. By pre-calculating the epipolar line, the amount of calculation in subsequent processing steps can be reduced. In the process of point cloud registration and fusion, the epipolar line information can be used to quickly locate the corresponding points, thereby speeding up the convergence speed of the algorithm. Epipolar line calculation helps to improve the overall quality of the data. By performing data analysis and processing in the epipolar line direction, the internal structure and characteristics of the data can be better understood, thereby providing more accurate basic data for subsequent three-dimensional reconstruction. Epipolar line calculation provides additional geometric constraints for the algorithm, which helps to improve the robustness of the entire fusion method. Even in the presence of noise and errors, the algorithm can effectively correct and optimize the data through epipolar line information.

[0079] In the present application, by coordinate transformation of data of different sensors, all sensor data are placed in the same coordinate system, which is the premise of effective data fusion. The unified coordinate system ensures the spatial alignment between different data sources, providing a basis for subsequent fusion processing. The accuracy of coordinate transformation directly affects the accuracy of the fusion result. Through accurate coordinate transformation, errors caused by inconsistent coordinate systems can be minimized, thereby improving the accuracy of the final three-dimensional model. Coordinate transformation enables data from different sensors to be interoperable in the same framework, enhancing the flexibility of the data processing flow. This interoperability is particularly important for complex data fusion tasks. Once all data are in the same coordinate system, subsequent data processing procedures will be greatly simplified. Registration, fusion and reconstruction steps can apply uniform methods and algorithms without the need for individual adjustments for each sensor. Coordinate transformation facilitates the integration of different sensor data, enabling the complementary advantages of each data source. Laser radar data can provide accurate geometric information, while camera images can provide rich texture information, and the effective combination of the two relies on accurate coordinate transformation.

[0080] It should be noted that, in this text, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply there is any such actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or device including the element.

[0081] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A deep learning-based multi-source data high-precision fusion method, characterized in that: The method comprises the following steps: S1: Perform sequence image input, collect and process raw data of multiple sensors, including laser point cloud and camera image, to form a series of image data; S2: Perform relative orientation element and feature point calculation, calculate the relative orientation element and extract the feature points for the data obtained by each sensor to establish a preliminary spatial geometric relationship; S3: Perform epipolar line calculation, calculate the parameters of the epipolar line according to the relative orientation element, which is used for subsequent point cloud denoising and smoothing processing; in the denoising process, filtering can be performed along the direction of the epipolar line, which can more effectively identify and remove abnormal points while retaining useful structural information; by pre-calculating the epipolar line, the corresponding points can be quickly located by using the epipolar line information in the point cloud registration and fusion process, thereby accelerating the convergence speed of the algorithm; S4: Perform point cloud denoising and smoothing, perform denoising and smoothing processing on the laser point cloud data to improve the accuracy of the data; S5: Perform point cloud smoothing, further perform smoothing processing on the point cloud data to reduce noise interference; S6: Perform coordinate conversion on the data of different sensors to make them in the same coordinate system, which is convenient for subsequent fusion processing; S7: Combine the point cloud fusion processing algorithm of the improved SAC-IA algorithm and the improved ICP algorithm, use the improved random sample consensus iterative closest point algorithm and the improved iterative closest point algorithm to perform fusion processing on the point cloud data to further improve the accuracy and integrity of the data; S8: Use the R-tree structure to index and organize the point cloud data to facilitate quick search and fusion of adjacent point cloud data; S9: Perform geometric modeling on the fused point cloud data to construct a three-dimensional model of the object S10: Use the texture information in the camera image to perform texture mapping on the three-dimensional model to make the model more realistic; S11: After completing all data processing and fusion, output the final complete and high-precision three-dimensional model.

2. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In the step S1, the unmanned aerial vehicle is equipped with a high-precision laser radar and a high-resolution RGB camera; the input sequence image data includes point cloud data scanned by the laser radar at a rate of 100,000 points per second and 1920x1080 resolution JPEG images captured by the camera at a rate of 30 frames per second; each data file contains timestamp information to ensure time synchronization; the laser radar data is stored in LAS format, and the camera image is stored in JPEG format containing EXIF metadata.

3. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In the step S2, the ORB-SLAM3 algorithm is used to calculate the relative orientation element of the camera; the specific parameters include: the FAST threshold of the ORB algorithm used for feature point extraction is set to 20, the pyramid level is 8 layers, and the size of the feature point descriptor is 32 bytes; when calculating the relative orientation element, the five-point algorithm is used to estimate the camera pose, and the RANSAC iteration number is set to 1000 times and the inlier threshold is set to 0.

005.

4. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In the step S3, the calculation process comprises: S3-1: Camera parameter acquisition: First, the intrinsic matrix of the camera needs to be obtained, which includes the focal length, principal point coordinates, and possibly distortion coefficients; S3-2 Image registration: Then, the relative orientation elements between two or more camera images need to be determined, which is a rotation matrix and a translation vector obtained through feature matching and geometric transformation; S3-3: Calculate the epipolar line: using the above parameters, calculate the epipolar line for each camera image; The collinearity equation is as follows: ; where f is the focal length; (c_x, c_y) is the principal point coordinate; R = [r_1, r_2,..., r_9] is the rotation matrix; t = [t_x, t_y, t_z] is the translation vector; X, Y, Z are the coordinates of the space point; x, y are the pixel coordinates on the image.

5. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In step S4, the StatisticalOutlierRemoval filter in the PCL library is used for denoising; the specific parameter settings are: set the number of neighbor points to 50 and the standard deviation multiple to 1.0; then, the VoxelGrid filter is used for downsampling and smoothing, and the voxel size is set to 0.01 meters.

6. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In step S5, the MovingLeastSquares filter in the PCL library is used for point cloud smoothing; the specific parameters include: set the search radius to 0.03 meters to determine the neighborhood size that affects each point; the order of polynomial fitting is set to 2 to ensure that the smoothing process does not excessively change the original shape of the point cloud.

7. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In step S6, the calculation process includes: S6-1: Obtain sensor data: First, collect data from different sensors, such as point cloud data from a laser radar and image data from a camera; S6-2: Determine the sensor coordinate system: Identify the original coordinate system of each sensor data; S6-3: Sensor calibration: Obtain the intrinsic and extrinsic parameters of each sensor through the calibration process; the intrinsic parameters include focal length, principal point coordinates, and distortion coefficients, and the extrinsic parameters include rotation matrix and translation vector; S6-14 Coordinate transformation: Use the extrinsic parameters to convert the data of each sensor to a unified reference coordinate system; The coordinate conversion formula is: ; where: (X_{ref}, Y_{ref}, Z_{ref}) is the point coordinate converted to the reference coordinate system; (Xsensor, Ysensor, Zsensor) is the point coordinate in the original sensor coordinate system; R is the rotation matrix, which describes the rotation from the sensor coordinate system to the reference coordinate system; t is the translation vector, which describes the translation from the sensor coordinate system to the reference coordinate system.

8. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In step S7, the improved SAC-IA algorithm is used for initial registration; The parameter settings include: set the sampling ratio to 0.05 and the distance threshold to 0.1 meters; then, apply the improved ICP algorithm for fine registration, with the maximum number of iterations set to 100 and the convergence threshold set to 0.001 meters; In step S8, use R-tree to index point cloud data; the specific parameters include: set the maximum number of entries for each R-tree node to 10 and the minimum number of entries for each node to 5; this ensures effective organization of point cloud data in space, facilitating fast retrieval.

9. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In the step S9, a Poisson reconstruction algorithm is used for geometry modeling; the parameter settings include: the grid resolution is set to 0.005 meters to ensure that the model has high precision; the neighborhood size used in normal estimation is set to 0.01 meters.

10. The deep learning-based multi-source data high-precision fusion method of claim 1, wherein: In the step S10, an MVS texture mapping technique is used; the specific parameters include: the minimum parallax angle for texture mapping is set to 5 degrees, and the maximum parallax angle is set to 75 degrees to ensure the quality of texture mapping; the resolution of the texture image is set to 1024x1024 pixels; In the step S11, the three-dimensional model is exported in OBJ format; the parameter settings include: the compression ratio of the model grid is set to 95% to reduce the file size; the compression format of the texture mapping uses JPEG, and the compression quality is set to 90%; the exported model contains vertex, texture coordinate and face information, ensuring the integrity and availability of the model in subsequent applications.

Images