Method and system for constructing a nuclear environment radiation map based on three-dimensional laser slam

Abstract

The application discloses a nuclear environment radiation map construction method and system based on three-dimensional laser SLAM, and the method comprises the following steps: estimating the pose information of a robot according to point cloud data and inertial data, constructing a three-dimensional point cloud map by using an incremental k-d tree, and then constructing a three-dimensional environment map layer; mapping radiation measurement data to the coordinate system of the three-dimensional environment map layer through coordinate transformation to obtain a radiation measurement point set; performing spatial interpolation and continuous modeling on the radiation measurement point set by using a Gaussian process regression, predicting the radiation intensity value of any position in the three-dimensional space, and generating a three-dimensional radiation field distribution layer; fusing the three-dimensional environment map layer and the three-dimensional radiation field distribution layer to generate a three-dimensional radiation map, and realizing the layered fusion and visualization of the environment map and the radiation information. The problems in the prior art that a two-dimensional radiation map cannot express a three-dimensional structure, three-dimensional radiation mapping has poor real-time performance, and the fusion precision of sparse radiation data and a point cloud map is low are solved.

Description

Technical Field

[0001] This invention belongs to the field of robot environmental perception and nuclear radiation monitoring technology, specifically involving a method and system for constructing nuclear environmental radiation maps based on three-dimensional laser SLAM. Background Technology

[0002] With the rapid development of the nuclear industry, the risk of nuclear accidents has increased accordingly. In nuclear accident emergency response, using mobile robots to enter high-radiation environments for detection has become an important means. However, in the complex three-dimensional nuclear environment, mobile robots not only need to accurately perceive the environmental geometry (such as stairs, equipment, pipes, etc.), but also need to acquire and visualize the spatial distribution of the radiation field in real time to provide decision support for operators.

[0003] Traditional radiation mapping methods primarily focus on two-dimensional planes, fusing 2D occupancy grid maps constructed using laser SLAM with radiation field information. These methods fail to represent environmental height variations and three-dimensional structures. In scenes with multi-layered structures or spatial occlusion, this leads to inaccurate registration between radiation information and geometric structures, resulting in poor visualization. Furthermore, existing 3D radiation mapping methods often face challenges such as high computational cost, poor real-time performance, and difficulty in effectively fusing sparse radiation measurement data with dense point cloud maps. For example, some methods employ graph optimization frameworks for pose estimation, resulting in high latency and failing to meet real-time mapping requirements; others suffer from insufficient interpolation accuracy in radiation data processing, leading to discontinuous or distorted reconstructed radiation field distributions. Therefore, there is an urgent need for a radiation mapping method capable of efficiently and accurately fusing environmental geometry and radiation field information in complex 3D kernel environments. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, the nuclear environment radiation map construction method and system based on three-dimensional laser SLAM provided by this invention solves the problems of existing technologies, such as the inability of two-dimensional radiation maps to represent three-dimensional structures, poor real-time performance of three-dimensional radiation mapping, and low accuracy of fusion between sparse radiation data and point cloud maps.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is: a method for constructing a nuclear environment radiation map based on three-dimensional laser SLAM, comprising the following steps: S1. Collect point cloud data, inertial data, and radiation measurement data, and synchronize the collected data in time. S2. Based on point cloud data and inertial data, a tightly coupled iterative Kalman filter is used to estimate the robot's pose information. An incremental kd-tree is used to construct a 3D point cloud map. The 3D point cloud map is then subjected to voxelization filtering to construct a 3D environment map layer. S3. Based on the robot's pose information, the radiation measurement data is mapped to the coordinate system of the three-dimensional environment map layer through coordinate transformation to obtain a set of radiation measurement points with spatial location information. S4. Gaussian process regression is used to perform spatial interpolation and continuous modeling of the radiation measurement point set to predict the radiation intensity value at any location in three-dimensional space and generate a three-dimensional radiation field distribution layer. S5. Fuse the three-dimensional environment map layer with the three-dimensional radiation field distribution layer to generate a three-dimensional radiation map that integrates the environmental geometry and radiation intensity distribution.

[0006] Furthermore: In S2, the specific method for constructing the octree-structured 3D environment map layer is as follows: The space occupied by the 3D point cloud map is recursively divided into multiple cubic voxels, each voxel corresponding to a node in an octree, and each voxel is assigned an occupancy probability value. The specific process is as follows: Let the side length of the cube space covered by the root node be... The maximum depth of an octree is Then at a depth of The corresponding voxel side length The specific expression is: In the formula, For depth indexing, ; Suppose at a given depth Above, the world coordinates of any point are defined as follows: In the formula It is the transpose symbol. , and for Given the three-dimensional world coordinates, the coordinates of the origin of the root node are: In the formula , and for The three-dimensional world coordinates of the point at depth voxel integer index Represented as: In the formula, Indicates the floor operation, depth upper voxel The following conditions must be met between its index and its parent node index: Let the voxel corresponding to the parent node be Its child nodes correspond to voxels as , For child node indexes, the log-likelihood ratio of the parent node The specific expression is: In the formula, Let be the log-likelihood ratio function.

[0007] Furthermore, in S3, the method for mapping radiometric measurement data to the coordinate system of the 3D environment map layer through coordinate transformation is as follows: Assume the radiation measurement point is in the world coordinate system. homogeneous coordinates In the formula , and for The three-dimensional world coordinates are then mapped to the map coordinate system of the three-dimensional environment map layer through coordinate transformation. coordinates below The specific expression is: In the formula, This is the rigid body transformation matrix from the world coordinate system to the map coordinate system. Indicates the rotational relationship between coordinate systems. Represents the space of real numbers. This represents the translation vector.

[0008] Furthermore, in S4, the method of using Gaussian process regression to perform spatial interpolation and continuity modeling of the radiation measurement point set is as follows: Set a set of radiation measurement points And the first The observed values ​​corresponding to each measurement point are The set of radiation measurements Define the covariance function, and establish the observed values ​​of radiation measurements based on the set of radiation measurement points. and predicted value joint prior distribution The specific expression is: In the formula, It is a mean function. for A symmetric positive definite covariance matrix of order 1. The standard deviation of noise. for An identity matrix of order 1. It follows a Gaussian distribution. For test points With radiation measurement point set Between The covariance matrix of order, , For test points Its own covariance, covariance function for: In the formula, and Let be the input vector of the covariance function. The characteristic length scale matrix, , and is the hyperparameter of the covariance function; Predicted value posterior probability distribution The specific expression is: In the formula, It follows a posterior Gaussian distribution. The mean of the radiation field. Let V be the variance of the radiation field.

[0009] Furthermore: S5 specifically refers to: The 3D environment map layer is published as a static map topic, and the 3D radiation field distribution layer is published as a radiation information topic. After time synchronization and spatial alignment, the radiation intensity information is superimposed on the idle voxels of the environment map layer in the form of color coding to generate a 3D radiation map.

[0010] A nuclear environment radiation map construction system based on 3D laser SLAM includes: The mobile robot platform is equipped with a 3D LiDAR, IMU, and radiation sensor; The data acquisition and synchronization module acquires and synchronizes data from the 3D lidar, IMU, and radiation sensor. The 3D SLAM mapping module, based on point cloud data and inertial data, uses a tightly coupled iterative Kalman filter algorithm to estimate the robot pose in real time and construct a 3D point cloud environment map, thereby generating a 3D environment map layer. The radiation field reconstruction module maps radiation measurement data to the environmental map coordinate system and uses the Gaussian process regression algorithm to reconstruct the three-dimensional radiation field distribution layer. The map fusion module merges the 3D environment map layer with the 3D radiation field distribution layer to generate a 3D radiation map. The visualization module displays the 3D radiation map in the RVIZ tool of the robot operating system.

[0011] The beneficial effects of this invention are as follows: (1) High-precision three-dimensional environment perception: The FAST-LIO2 tightly coupled laser-inertial SLAM algorithm is adopted, combined with ikd-Tree efficient map management, to realize high-precision, low-latency pose estimation and dense point cloud map construction in complex three-dimensional environment, providing a reliable geometric benchmark for accurate mapping of radiation information.

[0012] (2) Efficient radiation field reconstruction: Gaussian process regression is used to interpolate and model the sparse radiation measurement data, which can effectively handle data sparsity and noise, and generate a smooth and physically reliable three-dimensional radiation field distribution.

[0013] (3) Deep information fusion: Through the hierarchical fusion of the octree structured environment map layer and the GPR-based radiation field distribution layer, radiation information is fused only in the free voxel space, avoiding invalid calculations in obstacles or unknown areas, and improving the accuracy and storage efficiency of the map.

[0014] (4) Intuitive visualization: The generated three-dimensional radiation map can fully present the three-dimensional structure of the environment (such as stairs and platforms) and the gradual attenuation characteristics of the radiation field in space, providing more comprehensive information support for nuclear emergency decision-making and robot path planning.

[0015] (5) Simulation and actual experimental results show that the method can effectively construct radiation maps with spatial consistency and three-dimensional gradient characteristics under both single-source and multi-source conditions, verifying its good feasibility and robustness in complex structural environments. Attached Figure Description

[0016] Figure 1 This is a flowchart of the nuclear environment radiation map construction method based on three-dimensional laser SLAM of the present invention; Figure 2 This is a schematic diagram of the nuclear environment radiation map construction system based on three-dimensional laser SLAM of the present invention; Figure 3 This is a comparison chart of the mapping effects of three 3D laser SLAM algorithms in this embodiment of the invention; Figure 4 This is a schematic diagram of the construction result of a three-dimensional blank environment map layer in an embodiment of the present invention; Figure 5 This is a schematic diagram of the single-source three-dimensional radiation map construction result in an embodiment of the present invention; Figure 6 This is a schematic diagram of the multi-source three-dimensional radiation map construction result in an embodiment of the present invention; Figure 7 This is a schematic diagram of a physical experimental scene of a three-dimensional radiation map in an embodiment of the present invention; Figure 8This is a schematic diagram of the construction result of a three-dimensional blank environment map layer in a physical experiment according to an embodiment of the present invention; Figure 9 This is a schematic diagram of the construction result of a three-dimensional radiation map from a physical experiment in an embodiment of the present invention. Detailed Implementation

[0017] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0018] Example 1: like Figure 1 As shown, in one embodiment of the present invention, the method for constructing a nuclear environment radiation map based on three-dimensional laser SLAM includes the following steps: S1. Collect point cloud data, inertial data, and radiation measurement data, and synchronize the collected data in time. S2. Based on point cloud data and inertial data, the robot's pose information is estimated using tightly coupled iterative Kalman filter (FAST-LIO2). An incremental kd-tree (ikd-Tree) is used to construct a 3D point cloud map. The 3D point cloud map is then subjected to voxelization filtering to construct an octree-structured 3D environment map layer. S3. Based on the robot's pose information, the radiation measurement data is mapped to the coordinate system of the three-dimensional environment map layer through coordinate transformation to obtain a set of radiation measurement points with spatial location information. S4. Gaussian process regression (GPR) is used to perform spatial interpolation and continuous modeling on the radiation measurement point set to predict the radiation intensity value at any location in three-dimensional space and generate a three-dimensional radiation field distribution layer. S5. Fuse the three-dimensional environment map layer with the three-dimensional radiation field distribution layer to generate a three-dimensional radiation map that integrates the environmental geometry and radiation intensity distribution.

[0019] In S1, real-time point cloud data from the 3D LiDAR on the mobile robot, inertial data from the inertial measurement unit, and radiation measurement data from the radiation sensor are acquired, and the data are synchronized in time.

[0020] In this embodiment, a quadcopter is used as the mobile robot platform, equipped with a Livox Mid-360 3D lidar, an integrated IMU, and a WNB-8301 scintillator gamma radiation sensor. After the robot begins its task, each sensor starts collecting data. The lidar outputs 3D point cloud data at a frequency of 10Hz, the IMU outputs acceleration and angular velocity data at a frequency of 200Hz, and the radiation sensor outputs timestamped radiation dose rate values ​​at a frequency of 1Hz. Synchronization is achieved using a nearest neighbor matching method based on the timestamps of each sensor, synchronizing various types of data from the same or adjacent times to prepare for subsequent processing.

[0021] In S2, tightly coupled iterative Kalman filtering is a tightly coupled lidar inertial odometry method, characterized by: State estimation: The IMU state (position, velocity, attitude, zero bias, etc.) and the LiDAR pose are jointly estimated within an iterative Kalman filter framework. When a new point cloud frame is received, state prediction is first performed using IMU pre-integration. Then, the registration error between the point cloud and the current map is used as an observation for iterative updates, achieving high-precision real-time pose estimation.

[0022] Map Building: The map is managed using an incremental kd-tree (ikd-tree). The ikd-tree supports dynamic insertion, deletion, and rebalancing, enabling efficient processing of large-scale point cloud data. The algorithm transforms the current frame's point cloud to the world coordinate system based on the estimated pose, matches it with existing map point clouds, and updates the ikd-tree structure simultaneously.

[0023] After a quadcopter-mounted Livox Mid-360 lidar and IMU sensor perform mobile measurements in an unknown environment, spatial perception of the environment can be achieved through a 3D SLAM algorithm. This paper selects FAST-LIO2 as the 3D SLAM algorithm, and achieves high-frequency, high-precision 3D pose estimation by tightly coupling and fusing point cloud and inertial data, and reconstructs a continuous 3D point cloud environment map.

[0024] After obtaining the 3D point cloud map, considering that distant and invalid point clouds can adversely affect the efficiency and stability of 3D octree map construction, this invention first filters the original point cloud in the 3D point cloud map. That is, spatial constraint filtering is performed on the point cloud data before map construction. Subsequently, the preprocessed point cloud is used to construct an octree-structured 3D environment map layer. As the bottom-level state storage structure, the 3D environment map layer has the main advantages of significantly reducing storage capacity and querying the raster index of coordinate points at a constant time cost. In addition, the spatial representation of the octree map is similar to that of the 2D occupied raster map, both of which discretize the space; the difference is that the octree divides the continuous space into multiple cubic voxels and assigns an occupancy probability value in the range of 0 to 1 to distinguish between occupied, idle, and unknown states. Therefore, similar to the construction method of the 2D radial map, in the construction of the 3D octree map, only idle voxels are updated, while the other two states (occupied voxels and unknown voxels) are not processed, thereby improving processing efficiency and accuracy.

[0025] In S2, the specific method for constructing the octree structured 3D environment map layer is as follows: The space occupied by the 3D point cloud map is recursively divided into multiple cubic voxels, each voxel corresponding to a node in an octree, and each voxel is assigned an occupancy probability value. The specific process is as follows: Let the side length of the cube space covered by the root node be... The maximum depth of an octree is Then at a depth of The corresponding voxel side length The specific expression is: In the formula, For depth indexing, ; Suppose at a given depth Above, the world coordinates of any point are defined as follows: In the formula It is the transpose symbol. , and for Given the three-dimensional world coordinates, the coordinates of the origin of the root node are: In the formula , and for The three-dimensional world coordinates of the point at depth voxel integer index Represented as: In the formula, Indicates the floor operation, depth upper voxel The following conditions must be met between its index and its parent node index: That is, each parent node corresponds to its eight child voxels. This allows for the recursive partitioning of the three-dimensional space. During probabilistic map updates, the log-likelihood ratio of child nodes can be used to perform bottom-up fusion of the parent node's state. Let the voxel corresponding to the parent node be... Its child nodes correspond to voxels as , For child node indexes, the log-likelihood ratio of the parent node The specific expression is: In the formula, Let be the log-likelihood ratio function.

[0026] Based on the above formula, a three-dimensional environment map layer with an octree structure can be constructed in the kernel environment, so that each voxel represents the occupied, idle and unknown state of space with the probability of occupancy, and the hierarchical structure of the octree can adjust the voxel resolution according to the spatial complexity.

[0027] In S3, the method for mapping radiometric measurement data to the coordinate system of the 3D environment map layer through coordinate transformation is as follows: After aligning the radiometric measurement data with the robot pose data, it is necessary to accurately map the radiometric measurement points onto the 3D environment map layer. Therefore, the coordinates of the radiometric measurement points need to be transformed. Let's assume the radiometric measurement points are in the world coordinate system. homogeneous coordinates In the formula , and for The three-dimensional world coordinates are then mapped to the map coordinate system of the three-dimensional environment map layer through coordinate transformation. coordinates below The specific expression is: In the formula, This is the rigid body transformation matrix from the world coordinate system to the map coordinate system. Indicates the rotational relationship between coordinate systems. Represents the space of real numbers. This represents the translation vector.

[0028] Background correction and numerical conversion are performed on radiation measurements to eliminate interference from environmental background radiation. in, This indicates the corrected radiation value. Indicates the first Radiation measurements at each measurement point It represents the background radiation in the natural environment.

[0029] In S4, the method of spatial interpolation and continuous modeling of the radiation measurement point set using Gaussian process regression is as follows: Set a set of radiation measurement points And the first The observed values ​​corresponding to each measurement point are The set of radiation measurements This paper assumes that the distribution of the radiation field throughout the region follows a Gaussian process. Form a set of random variables, where For the input vector, Let be the potential radiation intensity function, and this variable follows a joint Gaussian distribution, where Therefore, the regression problem model can be expressed as: In the formula, the input vector Typically, it is a two-dimensional or three-dimensional coordinate system. These are observations affected by white noise, i.e., radiation measurements. It is a noise-free radiation intensity function. To observe the noise, It obeys Gaussian white noise, where Let be the noise standard deviation. Choosing an appropriate covariance function is crucial for performing nonlinear fitting on these data points. The covariance function defined in this paper... for: In the formula, and Let be the input vector of the covariance function. The characteristic length scale matrix, , and The hyperparameters of the covariance function are used to represent the features at the width and length scales, respectively; the aforementioned hyperparameters... , and noise standard deviation This can be learned from the observed data by maximizing the marginal log-likelihood function, or pre-defined based on prior environmental knowledge. Based on this, observed values ​​of radiation measurements are established from the set of radiation measurement points. and predicted value joint prior distribution The specific expression is: In the formula, It is a mean function. for A symmetric positive definite covariance matrix of order 1. The standard deviation of noise. for An identity matrix of order 1. It follows a Gaussian distribution. For test points (Location to be predicted) and set of radiation measurement points (The set of measured radiation points, i.e., the set of three-dimensional coordinates of radiation detection points actually measured) The covariance matrix of order, , For test points Its own covariance; Predicted value posterior probability distribution The specific expression is: In the formula, It follows a posterior Gaussian distribution. The mean of the radiation field. Let V be the variance of the radiation field.

[0030] After obtaining the radiation field distribution in the region, this paper analyzes and processes the data again to accurately locate the position of the unknown radioactive source.

[0031] S5 specifically refers to: The 3D environment map layer is published as a static map topic, and the 3D radiation field distribution layer is published as a radiation information topic. After time synchronization and spatial alignment, the radiation intensity information is superimposed on the idle voxels of the environment map layer in the form of color coding to generate a 3D radiation map.

[0032] like Figure 2 As shown, a nuclear environment radiation map construction system based on 3D laser SLAM includes: The mobile robot platform is equipped with a 3D LiDAR, IMU, and radiation sensor; The data acquisition and synchronization module acquires and synchronizes data from the 3D lidar, IMU, and radiation sensor. The 3D SLAM mapping module, based on point cloud data and inertial data, uses a tightly coupled iterative Kalman filter algorithm to estimate the robot pose in real time and construct a 3D point cloud environment map, thereby generating an octree structured 3D environment map layer. The radiation field reconstruction module maps radiation measurement data to the environmental map coordinate system and uses the Gaussian process regression algorithm to reconstruct the three-dimensional radiation field distribution layer. The map fusion module merges the 3D environment map layer with the 3D radiation field distribution layer to generate a 3D radiation map. The visualization module displays the 3D radiation map in the RVIZ tool of the robot operating system.

[0033] In this embodiment, the 3D SLAM mapping module adopts the FAST-LIO2 algorithm, which tightly couples LiDAR point cloud data and IMU pre-integration data through an iterative Kalman filter framework to achieve high-precision, low-latency pose estimation and map construction.

[0034] Example 2: This embodiment verifies the method of the present invention through simulation experiments, such as... Figure 3 As shown, a simulation environment was built in the Gazebo system to test the mapping performance of three 3D laser SLAM algorithms: LEGO-LOAM, LIO-SAM, and FAST-LIO2. Figure 3 (a) is the dataset environment. Figure 3 (b) The result of LEGO-LOAM mapping. Figure 3 (c) The result of LIO-SAM mapping. Figure 3 (d) shows the mapping results of FAST-LIO2. Experimental results show that FAST-LIO2 performs better in terms of real-time performance, continuity, and environmental adaptability, and can ensure the accurate fusion of sparse radiometric data in three-dimensional space.

[0035] like Figure 4 As shown, a three-dimensional simulation scene with multi-layered structure and varying height was constructed in Gazebo, and experiments were conducted using a quadcopter equipped with a Mid-360 lidar. Figure 4 (a) is the original point cloud environment map. Figure 4 (b) is a three-dimensional octree environment map. Based on this, experiments were conducted to construct three-dimensional radiation maps under single-source and multi-source conditions. Figure 5 The result of constructing a single-source 3D radiation map. Figure 6 The results of constructing a multi-source 3D radiation map validated the visualization effect of the designed method under complex spatial structures, different height source term distributions, and multi-source overlay conditions.

[0036] Example 3: This embodiment verifies the method of the present invention through physical experiments, such as... Figure 7 As shown, the experimental setting was a staircase environment with significant height changes and spatial hierarchy. Two [unclear - possibly referring to equipment or instruments] were used in the experiment. 60Co gamma-ray sources, with radioactivity levels of [missing information], are respectively [missing information]. and The radiation sources were fixedly placed on the surface of the stools on the stair platform.

[0037] During the environmental mapping phase, a quadcopter equipped with a Mid-360 lidar autonomously scanned the experimental area and acquired 3D geometric information of the environment. Simultaneously, an onboard scintillator detector continuously collected gamma radiation measurement data around two radiation sources during flight. The FAST-LIO2 algorithm was used to complete the 3D point cloud map of the experimental environment, resulting in the original point cloud environmental map, as shown below. Figure 8 As shown in (a). Further, the original point cloud data is filtered and voxelized, and an octree structure is used to construct a 3D environment map. The voxel resolution is set to 0.05m, and the final 3D octree environment map is shown below. Figure 8 As shown in (b).

[0038] After constructing the 3D blank environment map layer, the environment map data topic was published to the ROS system, and the octree environment map voxels were uniformly set to white. The 3D radial map construction result is as follows: Figure 9 As shown, the results of constructing a 3D radiation map from four different observation directions—right, left, backward, and forward—are presented. It can be seen that the radiation layer reconstructed based on the GPR method is successfully integrated into the 3D environment map, and the radiation layer and the environment layer maintain spatial consistency.

[0039] The system and method of this invention can efficiently and accurately construct a comprehensive map that combines environmental structure and radiation field information in a complex three-dimensional nuclear environment, which is of great significance for nuclear emergency response, radiation monitoring and autonomous robot navigation.

[0040] In the description of this invention, the above are merely preferred embodiments and are not intended to limit the scope of protection of this invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for constructing a nuclear environment radiation map based on three-dimensional laser SLAM, characterized in that, Includes the following steps: S1. Collect point cloud data, inertial data, and radiation measurement data, and synchronize the collected data in time. S2. Based on point cloud data and inertial data, a tightly coupled iterative Kalman filter is used to estimate the robot's pose information. An incremental kd-tree is used to construct a 3D point cloud map. The 3D point cloud map is then subjected to voxelization filtering to construct a 3D environment map layer. S3. Based on the robot's pose information, the radiation measurement data is mapped to the coordinate system of the three-dimensional environment map layer through coordinate transformation to obtain a set of radiation measurement points with spatial location information. S4. Gaussian process regression is used to perform spatial interpolation and continuous modeling of the radiation measurement point set to predict the radiation intensity value at any location in three-dimensional space and generate a three-dimensional radiation field distribution layer. S5. Fuse the three-dimensional environment map layer with the three-dimensional radiation field distribution layer to generate a three-dimensional radiation map that integrates the environmental geometry and radiation intensity distribution.

2. The method for constructing a nuclear environment radiation map based on three-dimensional laser SLAM according to claim 1, characterized in that, In S2, the specific method for constructing the octree structured 3D environment map layer is as follows: The space occupied by the 3D point cloud map is recursively divided into multiple cubic voxels, each voxel corresponding to a node in an octree, and each voxel is assigned an occupancy probability value. The specific process is as follows: Let the side length of the cube space covered by the root node be... The maximum depth of an octree is Then at a depth of The corresponding voxel side length The specific expression is: In the formula, For depth indexing, ; Suppose at a given depth Above, the world coordinates of any point are defined as follows: In the formula It is the transpose symbol. , and for Given the three-dimensional world coordinates, the coordinates of the origin of the root node are: In the formula , and for The three-dimensional world coordinates of the point at depth voxel integer index Represented as: In the formula, Indicates the floor operation, depth upper voxel The following conditions must be met between its index and its parent node index: Let the voxel corresponding to the parent node be Its child nodes correspond to voxels as , For child node indexes, the log-likelihood ratio of the parent node The specific expression is: In the formula, Let be the log-likelihood ratio function.

3. The method for constructing a nuclear environment radiation map based on three-dimensional laser SLAM according to claim 1, characterized in that, In S3, the method for mapping radiometric measurement data to the coordinate system of the 3D environment map layer through coordinate transformation is as follows: Assume the radiation measurement point is in the world coordinate system. homogeneous coordinates In the formula , and for The three-dimensional world coordinates are then mapped to the map coordinate system of the three-dimensional environment map layer through coordinate transformation. coordinates below The specific expression is: In the formula, This is the rigid body transformation matrix from the world coordinate system to the map coordinate system. Indicates the rotational relationship between coordinate systems. Represents the space of real numbers. This represents the translation vector.

4. The method for constructing a nuclear environment radiation map based on three-dimensional laser SLAM according to claim 3, characterized in that, In S4, the method of spatial interpolation and continuous modeling of the radiation measurement point set using Gaussian process regression is as follows: Let the radiation measurement point set be set. And the first The observed values ​​corresponding to each measurement point are The set of radiation measurements Define the covariance function, and establish the observed values ​​of radiation measurements based on the set of radiation measurement points. and predicted value joint prior distribution The specific expression is: In the formula, It is a mean function. for A symmetric positive definite covariance matrix of order 1. The standard deviation of noise. for An identity matrix of order 1. It follows a Gaussian distribution. For test points With radiation measurement point set Between The covariance matrix of order, , For test points Its own covariance, covariance function for: In the formula, and Let be the input vector of the covariance function. The characteristic length scale matrix, , and is the hyperparameter of the covariance function; Predicted value posterior probability distribution The specific expression is: In the formula, It follows a posterior Gaussian distribution. The mean of the radiation field. Let V be the variance of the radiation field.

5. The method for constructing a nuclear environment radiation map based on three-dimensional laser SLAM according to claim 1, characterized in that, S5 specifically refers to: The 3D environment map layer is published as a static map topic, and the 3D radiation field distribution layer is published as a radiation information topic. After time synchronization and spatial alignment, the radiation intensity information is superimposed on the idle voxels of the environment map layer in the form of color coding to generate a 3D radiation map.

6. A nuclear environment radiation map construction system based on three-dimensional laser SLAM, applied to the nuclear environment radiation map construction method based on three-dimensional laser SLAM as described in any one of claims 1 to 5, characterized in that, include: The mobile robot platform is equipped with a 3D LiDAR, IMU, and radiation sensor; The data acquisition and synchronization module acquires and synchronizes data from the 3D lidar, IMU, and radiation sensor. The 3D SLAM mapping module, based on point cloud data and inertial data, uses a tightly coupled iterative Kalman filter algorithm to estimate the robot pose in real time and construct a 3D point cloud environment map, thereby generating a 3D environment map layer. The radiation field reconstruction module maps radiation measurement data to the environmental map coordinate system and uses the Gaussian process regression algorithm to reconstruct the three-dimensional radiation field distribution layer. The map fusion module merges the 3D environment map layer with the 3D radiation field distribution layer to generate a 3D radiation map. The visualization module displays the 3D radiation map in the RVIZ tool of the robot operating system.

Images