Three-dimensional reconstruction method and system fusing RTK combined inertial navigation information
By combining RTK with inertial navigation information, the problem of accuracy and scale recovery in traditional three-dimensional reconstruction is solved, and high-precision and fast three-dimensional reconstruction effect is achieved.
Patent Information
- Application Number
- CN202510470669.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-15
- Publication Date
- 2025-07-11
AI Technical Summary
The traditional three-dimensional reconstruction method has limited reconstruction accuracy in a specific environment, and it is impossible to directly obtain the absolute scale information of the scene. The position estimation is prone to accumulated errors, which affects the iteration speed and accuracy of beam adjustment.
Fusion RTK combines inertial navigation information, obtains image data through image RTK and records inertial navigation data, performs feature matching and pose estimation, introduces position and rotation constraints, constructs an objective function for beam adjustment optimization, and restores the camera pose and point cloud three-dimensional coordinates.
The accuracy and speed of three-dimensional reconstruction is improved, the true scale of the scene is restored, error accumulation is reduced, and the accuracy of reconstruction results is enhanced.
Smart Images
Figure CN120298602A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of 3D reconstruction, and in particular to a 3D reconstruction method and system that fuse RTK combined inertial navigation information. Background Art
[0002] With the continuous development of computer vision and photogrammetry technologies, 3D reconstruction methods based on Structure from Motion (SfM) have been widely applied in fields such as surveying and mapping, UAV image modeling, and autonomous driving.
[0003] The reconstruction accuracy of traditional SfM reconstruction methods is still limited in certain specific environments. Since SfM only relies on visual features for camera pose estimation, it is impossible to directly obtain the absolute scale information of the scene. Although the scene scale can be restored by arranging control points, the process is relatively cumbersome. In addition, in the case of incorrect matching of feature points or uneven distribution of feature points, pose estimation will continuously accumulate errors as new cameras are added, thereby affecting the iteration speed and accuracy of the optimization stage of Bundle Adjustment (BA). Although some sensors can be used to obtain the camera pose during shooting, the pose estimated by the sensors still has errors. Using it directly as the camera pose for 3D reconstruction will cause the reconstruction result to overly rely on sensor data and ignore the geometric constraints between images, thereby affecting the reconstruction accuracy of the scene model. Summary of the Invention
[0004] In view of this, in order to solve the technical problem that it is difficult to obtain accurate camera pose information in real time in the existing 3D reconstruction method, which leads to low scene reconstruction accuracy, on the first aspect, the present invention proposes a 3D reconstruction method that fuses RTK combined inertial navigation information. The method includes the following steps:
[0005] Use image RTK to obtain sequence image data, and at the same time record the RTK combined inertial navigation data corresponding to each frame of image;
[0006] Perform feature matching and pose estimation according to the sequence image data;
[0007] Calculate the 3D coordinates of the point cloud according to the matching point pair information and pose estimation information;
[0008] Perform coordinate transformation according to the RTK combined inertial navigation data, and solve the camera pose and 3D coordinates of the point cloud in the local East-North-Up (ENU) coordinate system;
[0009] Introduce position constraints and rotation constraints to construct an objective function;
[0010] Perform Bundle Adjustment optimization according to the objective function to obtain the accurate camera pose and accurate 3D coordinates of the point cloud;
[0011] Perform 3D reconstruction based on the accurate camera pose and the 3D coordinates of the accurate point cloud.
[0012] In some embodiments, it further includes:
[0013] Screen the matching point pair information and eliminate some matching point pairs that do not conform to the rules.
[0014] In a second aspect, the present invention further provides a 3D reconstruction system integrating RTK combined inertial navigation information, and the system includes:
[0015] A data acquisition module, configured to obtain sequence image data by using image RTK, and record the RTK combined inertial navigation data corresponding to each frame of image at the same time;
[0016] A first calculation module, configured to perform feature matching and pose estimation according to the sequence image data, and calculate the 3D coordinates of the point cloud according to the matching point pair information and the pose estimation information;
[0017] A coordinate conversion module, configured to perform coordinate system conversion according to the RTK combined inertial navigation data, and solve the camera pose and the 3D coordinates of the point cloud in the local east-north-up coordinate system (ENU);
[0018] An optimization module, which introduces position constraints and rotation constraints to construct an objective function; performs bundle adjustment optimization according to the objective function to obtain the accurate camera pose and the accurate 3D coordinates of the point cloud;
[0019] A reconstruction module, configured to perform 3D reconstruction based on the accurate camera pose and the accurate 3D coordinates of the point cloud.
[0020] Based on the above solution, the present invention provides a 3D reconstruction method and system integrating RTK combined inertial navigation information, combines image data with RTK combined inertial navigation data, uses the RTK combined inertial navigation data to calculate the pose information of the camera during shooting, adds it as a constraint condition to the construction of the objective function of the BA algorithm, and jointly optimizes the camera pose estimation with the image data; at the same time, uses it as the initial value during the iteration of the BA algorithm to accelerate the convergence speed, and finally restores the true scale of the scene and completes the reconstruction of the 3D model. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 is a simplified step flow chart of a 3D reconstruction method integrating RTK combined inertial navigation information according to the present invention;
[0022] Figure 2 is a schematic diagram of the data processing process of a 3D reconstruction method integrating RTK combined inertial navigation information according to the present invention;
[0023] Figure 3 is a structural block diagram of a 3D reconstruction system integrating RTK combined inertial navigation information according to the present invention. Detailed implementation manners
[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.
[0025] It should be noted that, for the sake of convenience of description, only the parts related to the relevant invention are shown in the drawings. Without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other.
[0026] It should be understood that the "system", "device", "unit" and / or "module" used in the present application is a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the word can be replaced by other expressions.
[0027] As shown in the present application and the claims, unless the context clearly indicates an exception, words such as "a", "an", "one" and / or "the" are not specifically singular, but may also include plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements. The element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the element.
[0028] In the description of the embodiments of the present application, "a plurality" means two or more than two. The following terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features.
[0029] In addition, flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the previous or subsequent operations do not necessarily need to be executed precisely in sequence. On the contrary, the operations can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.
[0030] Refer to Figure 1, which is a schematic flow diagram of an optional example of the three-dimensional reconstruction method that fuses RTK combined inertial navigation information proposed by the present invention. This method can be applied to computer devices. The three-dimensional reconstruction method proposed in this embodiment may include but is not limited to the following steps:
[0031] Step S1: Take images using an imaging RTK and record the pose to obtain sequence image data and RTK combined inertial navigation data;
[0032] Step S2: Perform pose estimation and point cloud calculation based on the sequence image data to obtain the camera pose in the first coordinate system and the three-dimensional coordinates of the point cloud in the first coordinate system;
[0033] Step S3: Perform coordinate transformation based on the RTK combined inertial navigation data to obtain the camera pose in the second coordinate system;
[0034] Step S4: Restore the scene scale according to the camera pose in the second coordinate system, and solve the coordinates of all point cloud three-dimensional coordinates in the second coordinate system to obtain the three-dimensional coordinates of the point cloud in the second coordinate system;
[0035] Step S5: Construct an objective function;
[0036] Step S6: Perform bundle adjustment optimization on the camera pose in the second coordinate system and the three-dimensional coordinates of the point cloud in the second coordinate system according to the objective function to obtain the final camera pose and the final three-dimensional coordinates of the point cloud;
[0037] Step S7: Perform three-dimensional reconstruction based on the final camera pose and the final three-dimensional coordinates of the point cloud.
[0038] In some embodiments, step S1 specifically includes:
[0039] The RTK combined inertial navigation data includes the geodetic coordinates and attitude angle data of the camera.
[0040] Among them, the geodetic coordinates of the camera (latitude, longitude, and height) are denoted as (B, L, h), where B represents latitude, L represents longitude, and H represents height. The attitude angles are denoted as (θ x , θ y , θ z ), where θ x represents the roll angle Roll, θ y represents the pitch angle Pitch, and θ z represents the yaw angle Yaw.
[0041] In some feasible embodiments, step S2 specifically includes:
[0042] Step S2.1: Extract feature points from the sequence image data using the SIFT algorithm and perform feature matching using the BFMatcher algorithm to obtain matching point pair information;
[0043] Step S2.2: Select two images from the sequence image data as the initial images, calculate the essential matrix, perform SVD decomposition to estimate the camera pose, and obtain the camera pose estimation information;
[0044] Step S2.3: According to the matching point pair information and the camera pose estimation information, calculate the three-dimensional coordinates of the point cloud through triangulation;
[0045] Step S2.4: Add the next view, use the PnP algorithm to estimate the pose of the new view, find new feature matching points between the previous initial image and the new view. If the matching point does not exist in the existing point cloud, perform triangulation to calculate the new three-dimensional coordinates of the point cloud. Repeat this step until all views are processed, and finally obtain the poses of all cameras and the three-dimensional coordinates of all point clouds in the SfM coordinate system.
[0046] In some feasible embodiments, it further includes:
[0047] Use the RANSAC algorithm to screen the matching point pair information and eliminate the wrong matching point pairs.
[0048] In some feasible embodiments, step S3 specifically includes:
[0049] The RTK integrated inertial navigation system can output the longitude, latitude, altitude and attitude angles of the camera during shooting in combination with the internal parameters and calibration information of its own sensors. The longitude, latitude and altitude data are based on the geodetic coordinate system (WGS84), which is a spherical coordinate system. When directly performing three-dimensional geometric operations, it is necessary to process the ellipsoidal curvature and complex trigonometric functions, with high computational complexity and easy introduction of numerical errors. In the present invention, it is first converted to the Earth-centered, Earth-fixed coordinate system (ECEF), and then converted to the local north-east-down coordinate system (ENU). ENU is a local planar coordinate system, which helps to improve the accuracy of subsequent local reconstruction. The conversion process is as follows:
[0050] The longitude, latitude and altitude of the corresponding camera when obtaining the captured image in step 1 are (B, L, h). The geodetic coordinates (B, L, h) can be converted to the Earth-centered, Earth-fixed coordinates (X, Y, Z) through formula (1), where N is the radius of the prime vertical circle and e is the first eccentricity of the WGS84 ellipsoid:
[0051]
[0052] When converting ECEF coordinates to ENU coordinates, it is usually necessary to determine a reference point and then construct an ENU coordinate system with this as the origin. In this solution, the longitude, latitude, and altitude of the camera when taking the first photo are used as the reference point to construct the ENU coordinate system, and the ECEF coordinates are converted to ENU coordinates through Equation (2), where (X0, Y0, Z0) are the ECEF coordinates of the reference point. By performing the above conversion on the longitude, latitude, and altitude of the camera corresponding to each photo, the three-dimensional coordinates of the camera center in the ENU coordinate system can be obtained, denoted by t ENU to represent.
[0053]
[0054] From the camera pose angle information (θ x , θ y , θ z ) obtained in Step 1, it is necessary to convert it into a rotation matrix in the ENU coordinate system. The pose angles Roll, Pitch, and Yaw respectively correspond to rotations around the X-axis, Y-axis, and Z-axis of itself. The rotation matrix R ENU of the camera in the ENU coordinate system is obtained by multiplying the three Euler angle rotation matrices:
[0055]
[0056] R ENU = R z (θ z ) R y (θ y ) R x (θ x )(6)
[0057] In some feasible embodiments, Step S4 specifically includes:
[0058] Use the position information of the camera in the ENU coordinate system obtained in Step S3 to restore the scene scale information. In the process of monocular vision incremental reconstruction, since the monocular camera cannot directly restore the absolute scale, the point cloud and camera pose estimated by SfM only exist in a relative coordinate system with an unknown scale and unit, that is, the SfM coordinate system. Therefore, it is necessary to perform a similarity transformation (Sim(3)) to align the results obtained in Step S2 in the SfM coordinate system to the ENU coordinate system established in the above Step S3. The present invention uses the Umeyama method to solve Sim(3). The Umeyama algorithm is an algorithm for aligning two trajectories or point sets, mainly used to solve the coordinate system transformation between two sets of point clouds, including the calculation of rotation, translation, and scale factor. The specific solution steps are as follows:
[0059] Denote the set of camera center points obtained in Step S2 in the SfM coordinate system as The set of camera center points in the ENU coordinate system in Step 7 is denoted as Calculate the means of the two sets respectively:
[0060]
[0061] Eliminate the translational effect and calculate the centered coordinates:
[0062] X i = C SFMi - C SFM (9)
[0063]
[0064] Construct the covariance matrix and perform singular value decomposition (SVD):
[0065]
[0066] H = U∑V T (12)
[0067] Solve for the rotation matrix R as in Equation (13), where D is the alignment matrix to ensure that the rotation matrix R is orthogonal and has a determinant of 1.
[0068] R = VDU T (13)
[0069]
[0070] Calculate the scale factor s, where tr(·) represents the trace of a matrix:
[0071]
[0072] Calculate the translation vector t based on the rotation matrix R and scale factor s obtained above:
[0073] t = C ENU - sRC SFM (16)
[0074] Perform coordinate system alignment and transform the three-dimensional coordinates P of the point cloud in the SfM coordinate system SFM to the ENU coordinate system to obtain the three-dimensional coordinates P of the point cloud in the ENU coordinate system ENU , and the calculation formula is as follows:
[0075] P ENU = s·R·P SFM + t(19)
[0076] In some feasible embodiments, step S5 specifically includes:
[0077] Global optimization is performed using Bundle Adjustment (BA). When constructing the objective function, camera position constraints and rotation constraints are added to jointly optimize the camera pose and 3D point cloud. In the BA stage of traditional SfM 3D reconstruction methods, the camera pose and the positions of 3D points are optimized by minimizing the reprojection error of pixel coordinates, and its form is:
[0078]
[0079] where p ij represents the observed point of the 3D point X j on the i-th image, R i , t i correspond to the rotation matrix and translation vector of the camera, K is the camera internal parameter, and π(K, R i , t i , X j ) is the camera projection model, and ρ(·) is the Cauchy kernel function, which can avoid the interference of outliers. On this basis, the present invention constructs a position constraint term E pos and a rotation constraint term E rot , and their forms are as follows:
[0080]
[0081]
[0082] where t ENU,i and R ENU,i are the camera translation vector and rotation matrix solved according to RTK combined inertial navigation information in step S3. log(·) calculates the Lie algebra logarithmic mapping to obtain the rotation error vector ω i =(ω x , ω y , ω z ), W rot =diag(ω x , ω y , ω z ) is the weight matrix of the rotation constraint, and ω x , ω y , ω z correspond to the weights of the roll angle, pitch angle, and yaw angle errors respectively. The weights of each attitude angle are determined by formula (23), where σ is the standard deviation of each attitude angle. According to the measurement accuracy of the IMU in different directions, different rotation weights are adaptively assigned.
[0083]
[0084] Add the above position constraint terms and rotation constraint terms into the final BA algorithm optimization. Fully consider the reprojection error and the pose constraints provided by RTK integrated inertial navigation, and jointly construct the final optimization objective function, whose form is as follows:
[0085] E total = E proj + λ pos E pos + λ rot E rot (24)
[0086] Among them, λ pos and λ rot are weight factors used to adjust the weights of each error term. Their role is to balance the contributions of different error terms and prevent a certain error term from being ignored or dominating the optimization process. Weights are assigned according to the different characteristics of the errors caused by RTK and IMU, and the weights are adjusted to improve the accuracy and robustness of 3D reconstruction in different scenarios. Finally, Ceres Solver is used to solve the objective function constructed in Equation (23). The camera coordinates t ENU in the ENU coordinate system and the rotation matrix R ENU calculated in step S3 are used as the initial values of the translation vector t i and the rotation matrix R i of the camera in the objective function. Through continuous iterative calculation, the objective function is minimized to solve the optimal camera pose and the 3D coordinates of the point cloud. Finally, a 3D model of the scene is reconstructed based on the optimal camera pose and the 3D coordinates of the point cloud.
[0087] The present invention integrates RTK integrated inertial navigation data into the traditional SfM 3D reconstruction method, uses RTK integrated inertial navigation data to solve the camera pose parameters, aligns the coordinate systems with the camera pose estimation of the traditional SfM, so as to restore the absolute scale information of the scene; during BA optimization, the position constraints and rotation constraints provided by RTK integrated inertial navigation are introduced, combined with the reprojection error constraints of the traditional SfM, to jointly construct an objective function to optimize the camera pose and point cloud information. At the same time, the pose provided by RTK integrated inertial navigation is used as the initial value during the iteration of the BA algorithm to improve the algorithm convergence speed; fully consider the data errors of different sensors of RTK and IMU, and add weight factors to balance the influence of different constraint terms. According to the error characteristics of IMU in different directions, different weights are applied to the rotation errors, further subdividing the rotation constraints and reducing the influence of yaw angle errors on the optimization results.
[0088] In summary, the schematic diagram of the specific data flow of the present invention is referred to Figure 2 .
[0089] As Figure 3 shown, a 3D reconstruction system integrating RTK integrated inertial navigation information includes:
[0090] A data acquisition module for performing step S1;
[0091] A first calculation module for performing step S2;
[0092] A coordinate transformation module for performing step S3 and step S4;
[0093] An optimization module for performing step S5 and step S6;
[0094] A reconstruction module for performing step S7.
[0095] The content in the above method embodiments is applicable to the system embodiments of the present invention. The functions specifically implemented by the system embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.
[0096] A three-dimensional reconstruction device integrating RTK and inertial navigation information:
[0097] At least one processor;
[0098] At least one memory for storing at least one program;
[0099] When the at least one program is executed by the at least one processor, the at least one processor implements the three-dimensional reconstruction method integrating RTK and inertial navigation information as described above.
[0100] The content in the above method embodiments is applicable to the device embodiments of the present invention. The functions specifically implemented by the device embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.
[0101] A storage medium storing processor-executable instructions, which are used to implement the three-dimensional reconstruction method integrating RTK and inertial navigation information as described above when executed by a processor.
[0102] The content in the above method embodiments is applicable to the storage medium embodiments of the present invention. The functions specifically implemented by the storage medium embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.
[0103] The above is a specific description of the preferred embodiments of the present invention. However, the present invention is not limited to the described embodiments. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included in the scope defined by the claims of this application.
Claims
1. A three-dimensional reconstruction method that fuses RTK and inertial navigation information, characterized in that, Including the following steps: Based on the imaging RTK, take images and record the pose to obtain sequence image data and RTK integrated inertial navigation data; Perform pose estimation and point cloud calculation according to the sequence image data to obtain the camera pose in the first coordinate system and the three-dimensional coordinates of the point cloud in the first coordinate system; Perform coordinate transformation according to the RTK integrated inertial navigation data to obtain the camera pose in the second coordinate system; Restore the scene scale according to the camera pose in the second coordinate system, and solve the coordinates of all the three-dimensional coordinates of the point cloud in the second coordinate system to obtain the three-dimensional coordinates of the point cloud in the second coordinate system; Construct an objective function; According to the objective function, perform bundle adjustment optimization on the camera pose in the second coordinate system and the three-dimensional coordinates of the point cloud in the second coordinate system to obtain the final camera pose and the final three-dimensional coordinates of the point cloud; Perform three-dimensional reconstruction according to the final camera pose and the final three-dimensional coordinates of the point cloud.
2. The three-dimensional reconstruction method for fusing RTK combined inertial navigation information according to claim 1, wherein It also includes: Screen the matching point pair information.
3. The three-dimensional reconstruction method for fusing RTK and inertial navigation information according to claim 1, characterized in that The step of performing pose estimation and point cloud calculation according to the sequence image data to obtain the camera pose in the first coordinate system and the three-dimensional coordinates of the point cloud in the first coordinate system specifically includes: Perform feature matching on the sequence image data to obtain matching point pair information; Select two images from the sequence image data as the initial images, and decompose and estimate the pose of the camera to obtain camera pose estimation information; According to the matching point pair information and the camera pose estimation information, calculate the three-dimensional coordinates of the point cloud through triangulation; Select a new image from the sequence image data and estimate the corresponding pose, calculate the new three-dimensional coordinates of the point cloud, and loop this step until all the images in the sequence image data are traversed to obtain the camera pose in the first coordinate system and the three-dimensional coordinates of the point cloud in the first coordinate system corresponding to all the images.
4. The three-dimensional reconstruction method for fusing RTK and inertial navigation information according to claim 1, characterized in that, The step of constructing the objective function specifically includes: Construct position constraints and rotation constraints based on the second coordinate system; Introduce the position constraints and the rotation constraints to construct an objective function.
5. The three-dimensional reconstruction method for fusing RTK combined inertial navigation information according to claim 4, wherein, The rotation matrix of the camera in the second coordinate system is expressed as follows: R ENU = R z (θ z )R y (θ y )R x (θ x ) Among them, R ENU represents the rotation matrix of the camera in the second coordinate system, and θ x , θ y , θ z respectively represent the attitude angle information of the corresponding coordinate axes of the camera.
6. The three-dimensional reconstruction method for fusing RTK and inertial navigation information according to claim 5, wherein The calculation formula for the three-dimensional coordinates of the point cloud in the second coordinate system is expressed as follows: P ENU = s·R·P SFM + t Among them, P ENU represents the three-dimensional coordinates of the point cloud in the second coordinate system, s represents the scale factor, R represents the rotation matrix, and t represents the translation vector.
7. A three-dimensional reconstruction method for fusing RTK and inertial navigation information according to claim 4, characterized in that The formula for the objective function is expressed as follows: E total = E proj + λ pos E pos + λ rot E rot where p ij represents the observation point of the three-dimensional point X j on the i-th image; R i , t i represent the rotation matrix and translation vector of the camera corresponding to the i-th image; K represents the camera intrinsic parameter; π(K, R i , t i , X j ) represents the camera projection model; ρ represents the Cauchy kernel function; t ENU,i and R ENU,i represent the camera translation vector and rotation matrix corresponding to the i-th image solved from the RTK combined inertial navigation information; λ pos and λ rot represent the corresponding weight factors; W rot represents the weight matrix of the rotation constraint.
8. A three-dimensional reconstruction system that fuses RTK and integrated inertial navigation information, characterized in that Including: A data acquisition module that takes images based on the imaging RTK and records the pose to obtain sequence image data and RTK integrated inertial navigation data; A first calculation module for performing pose estimation and point cloud calculation according to the sequence image data to obtain the camera pose in the first coordinate system and the three-dimensional coordinates of the point cloud in the first coordinate system; A coordinate transformation module for performing coordinate transformation according to the RTK integrated inertial navigation data to obtain the camera pose in the second coordinate system; For restoring the scene scale according to the camera pose in the second coordinate system and solving the coordinates of all the three-dimensional coordinates of the point cloud in the second coordinate system to obtain the three-dimensional coordinates of the point cloud in the second coordinate system; An optimization module for constructing an objective function and performing bundle adjustment optimization on the camera pose in the second coordinate system and the three-dimensional coordinates of the point cloud in the second coordinate system according to the objective function to obtain the final camera pose and the final three-dimensional coordinates of the point cloud; A reconstruction module for performing 3D reconstruction based on the final camera pose and the final 3D coordinates of the point cloud.
9. A three-dimensional reconstruction device that fuses RTK and integrated inertial navigation information, characterized in that, Comprising: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a 3D reconstruction method for fusing RTK and inertial navigation information as described in any one of claims 1-7.