Radiation mapping system and method based on 3D magic cube structure gamma imaging detector

Abstract

The present application relates to a radiation mapping system and method based on a 3D magic cube structure gamma imaging detector, comprising a 3D magic cube structure gamma imaging subsystem, an LVI-SLAM scene perception subsystem and a data fusion and three-dimensional radiation field reconstruction subsystem, the 3D magic cube structure gamma imaging subsystem detects radiation information; the LVI-SLAM scene perception subsystem outputs the motion trajectory, pose and three-dimensional point cloud with texture color of the system; the data fusion and three-dimensional radiation field reconstruction subsystem takes the global clock and pose of the LVI-SLAM scene perception subsystem as the reference, performs space-time alignment on the detected radiation information, and maps it into the three-dimensional point cloud scene through inversion projection. It has the technical advantages of light weight, high imaging efficiency and full field of view capability, can obtain radiation intensity information at a long distance without personnel approaching the hot spot, and realizes non-contact measurement. It realizes the deep fusion of vision, laser and inertial navigation, and improves the reliability of the system.

Description

Technical Field

[0001] This invention belongs to the field of radiation detection, specifically relating to a radiation mapping system and method based on a 3D cube structure gamma imaging detector. Background Technology

[0002] Three-dimensional radiation mapping technology, by fusing radiation detection information with three-dimensional spatial scene data, enables spatial visualization and quantitative analysis of the radiation field. It can efficiently acquire multi-dimensional information about the radiation environment, providing crucial technical support for on-site planning and execution of safe and efficient operations, as well as emergency response decision-making, in nuclear facility operation, maintenance, and decommissioning. Radiation measurement, three-dimensional scene perception, and scene information fusion analysis are the core technological foundations of the three-dimensional radiation mapping system.

[0003] Currently used indoor 3D gamma radiation field mapping systems have inherent shortcomings in three aspects: radiation detection, 3D scene perception, and scene information fusion. Regarding radiation detection, non-directional detectors (such as dose rate meters and gamma spectrometers) lack directional information and require proximity to the radiation source to accurately obtain the 3D radiation source distribution, posing a risk of personnel exposure. While coded aperture gamma cameras with directional information can image at long distances, they are heavy, have a small field of view, and poor portability. Compton cameras have a large field of view, but rely on coincidence measurements of scattering and absorption events, resulting in complex electronics and imaging algorithms, low imaging efficiency, and poor portability. In terms of 3D scene perception, visual SLAM (VSLAM) based on depth cameras can simultaneously acquire distance and color information, but has a small measurement field of view, large errors, and poor robustness. Lidar-SLAM based on lidar can acquire high-precision distance point clouds, but lacks true scene color information and is prone to tracking loss in environments with weak features, such as long corridors. In terms of scene information fusion, it is necessary to combine three-dimensional scene, trajectory positioning and position and attitude information: non-directional detectors can only combine trajectory positioning and obtain radiation field distribution through three-dimensional spatial interpolation. Although directional detectors can further combine position and attitude information to achieve long-distance non-proximity observation, they are limited by the inherent defects of the above-mentioned single detectors or single sensing sensors, making it difficult to simultaneously achieve wide field of view, high angular resolution, compact structure and lightweight.

[0004] There is an urgent need for wide-field-of-view, high-angular-resolution, compact, and lightweight gamma-ray imaging equipment in current facility-based three-dimensional radiometric mapping. Therefore, there is a pressing need to develop a wide-field-of-view, high-angular-resolution, compact, and lightweight gamma-ray imaging device for use in radiometric mapping systems for gamma-ray imaging detectors. Summary of the Invention

[0005] To address the technical deficiencies in existing technologies, the present invention aims to provide a radiation mapping system and method based on a 3D cube-structure gamma imaging detector. This system and method possess technical advantages such as high integration, lightweight design, wide field of view, high angular resolution, and high imaging efficiency, enabling real-time measurement and fusion analysis of high-precision three-dimensional scene radiation distribution without physical proximity.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, the present invention discloses a radiation mapping system based on a 3D cube-structure gamma imaging detector. The system includes a 3D cube-structure gamma imaging subsystem, an LVI-SLAM scene perception subsystem, and a data fusion and three-dimensional radiation field reconstruction subsystem. The 3D cube-structure gamma imaging subsystem is used to detect radiation information using a gamma detector array. The LVI-SLAM scene perception subsystem is used to output the system's motion trajectory, pose, and textured three-dimensional point cloud. The data fusion and three-dimensional radiation field reconstruction subsystem is used to perform spatiotemporal alignment of the radiation information detected by the 3D cube-structure gamma imaging subsystem based on the global clock and pose of the LVI-SLAM scene perception subsystem, and then map it onto the three-dimensional point cloud scene through inversion projection.

[0008] Furthermore, the radiation information detected by the gamma detector array includes the direction and intensity information of radiation hotspots.

[0009] Furthermore, the 3D cube structure gamma imaging subsystem includes a shielding block and a gamma detector module. The shielding block and the gamma detector module are arranged randomly in 3D space to form a cube structure, which constitutes a gamma detector array.

[0010] Furthermore, the 3D cube structure gamma imaging subsystem is also used to obtain the location of radiation hotspots in 4π space by utilizing the inherent response mode between the gamma ray incident direction and the detector, based on the response intensity of each gamma detector module, through image reconstruction algorithms.

[0011] Furthermore, the LVI-SLAM scene perception subsystem uses LiDAR for precise ranging to acquire 3D point cloud information, uses a visual camera to acquire RGB images for feature extraction, uses an IMU to acquire pose information, and uses a deep fusion algorithm of laser-vision-inertial navigation to construct a 3D scene point cloud with high-precision geometric structure and texture color information in real time.

[0012] Furthermore, the visual camera uses multi-camera stitching to achieve a wide field of view and increases the number of feature matches between adjacent frames.

[0013] Furthermore, the data fusion and 3D radiation field reconstruction subsystem uses an IMU pre-integration algorithm to convert multiple IMU data within a frame of lidar point cloud into relative state constraints, and interpolates the results of IMU state propagation to obtain the state of each lidar point. The radiation information detected by the γ detector array is then mapped to neighboring data frames through interpolation.

[0014] Furthermore, the shielding block includes a tungsten block or a lead block.

[0015] Furthermore, the gamma detector module includes a scintillator linked to a photomultiplier tube.

[0016] Secondly, the present invention discloses a radiometric mapping method based on a 3D Rubik's Cube structure gamma imaging detector. Based on the radiometric mapping system based on a 3D Rubik's Cube structure gamma imaging detector as described in the first aspect and any optional embodiment of the present invention, the method includes the following steps:

[0017] S1. The detected radiation information, as well as the acquired 3D point cloud information, feature information, and pose information, are spatiotemporally synchronized and aligned with coordinates.

[0018] S2. Input the radiation information after spatiotemporal synchronization and coordinate alignment, as well as the acquired 3D point cloud information, feature information and pose information data, into the 3D construction and radiation field reconstruction platform;

[0019] S3. Construct a radiation observation model based on the input data;

[0020] S4. Based on the constructed radiation observation model, perform three-dimensional radiation scene visualization and fusion analysis.

[0021] The beneficial technical effects of this invention are as follows: The radiation mapping system and method based on a 3D Rubik's Cube structure gamma imaging detector disclosed in this invention have the following specific beneficial effects:

[0022] For the first time, 3D coded imaging technology based on angular response mode was applied to a portable LVI-SLAM mapping system, achieving dual innovations in the structure and algorithm of the radiometric mapping system. Through the deep fusion of vision, laser, and inertial navigation, the problem of incomplete scene construction or positioning drift caused by the performance degradation of single sensors in on-site measurements such as nuclear facility decommissioning and emergency response was solved.

[0023] Thanks to the cube detector's lightweight design, high imaging efficiency, and full field of view capability, radiation intensity information can be obtained from a distance without personnel needing to approach the hotspot, achieving non-proximity measurement.

[0024] The system is highly integrated and has efficient algorithms. Its weight can be controlled to be less than 2kg. It can realize real-time synchronous acquisition and processing of radiation data and scene data, thereby improving measurement efficiency.

[0025] LVI-SLAM technology solves the problem of data loss caused by single sensors in complex nuclear facility environments (such as long corridors and low-light conditions). The data acquisition of the cube detector modules is independent, and even if one module fails, it can still sense the direction of radiation hotspots, thus improving the reliability of the system.

[0026] Traditional methods, such as the Compton camera, rely on coincidence measurements between multiple detectors (i.e., time coincidence of scattering and absorption events to reconstruct the incident direction), which places high demands on high-speed electronics and ASIC chips. This system, based on a pre-calibrated response matrix, can invert the orientation of radiation hotspots in 4π space from the response intensity of each detector module. Each module counts independently and does not require coincidence measurements. Therefore, the electronic system is simple, the number of modules is small, and there is no need for expensive ASIC chips. This significantly reduces the manufacturing threshold of radiation detectors in high-end 3D radiometric mapping equipment and results in low cost.

[0027] By expanding the traditional 2D coded aperture into a 3D random shielding array, omnidirectional (4π) radiation sensing is achieved in a very small volume, making the system structure compact. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the radiation mapping system based on a 3D Rubik's Cube structure gamma imaging detector, as shown in Embodiment 1 of the present invention.

[0029] Figure 2 Flowchart of the multimodal data fusion algorithm for the Rubik's Cube gamma detector-laser-vision-IMU;

[0030] Figure 3 A schematic diagram illustrating the method for aligning information between a gamma detector, lidar, IMU, and camera.

[0031] Figure 4 Flowchart for reconstructing three-dimensional radiation distribution;

[0032] The components are: 1-γ detector array, 2-LiDAR, 3-Vision camera, 4-Inertial measurement unit, 5-Data acquisition and processing unit, and 6-Supporting bracket. Detailed Implementation

[0033] The present invention will now be further described with reference to the accompanying drawings and specific embodiments.

[0034] Example 1

[0035] like Figure 1As shown, this embodiment of the invention provides a radiation mapping system based on a 3D cube-structure gamma imaging detector, including a 3D cube-structure gamma imaging subsystem, an LVI-SLAM scene perception subsystem, and a data fusion and 3D radiation field reconstruction subsystem. The 3D cube-structure gamma imaging subsystem is used to detect radiation information such as the direction and intensity of radiation hotspots using a 3D cube gamma detector array. The LVI-SLAM scene perception subsystem is used to acquire the system's motion trajectory, pose, and textured 3D point cloud. The data fusion and 3D radiation field reconstruction subsystem is used to perform spatiotemporal alignment of the radiation information of the 3D cube-structure gamma imaging subsystem based on the global clock and pose of the LVI-SLAM scene perception subsystem, and then maps it onto the 3D point cloud scene through inversion projection. That is, the first two subsystems provide "radiation information" and "spatial information" respectively, and the third subsystem completes the spatiotemporal registration and fusion reconstruction of the two, ultimately realizing the 3D visualization of the radiation field.

[0036] The 3D Rubik's Cube structure gamma imaging subsystem includes shielding blocks and gamma detector modules. The shielding blocks and gamma detector modules are arranged randomly in 3D space to form a cube structure, constituting gamma detector array 1. The overall structure of the 3D Rubik's Cube structure gamma imaging subsystem is controlled within 30cm. 3 Within this range, to meet the requirements of compact structure, lightweight and portability.

[0037] The shielding block needs to be made of high atomic number and high density materials to effectively shield and modulate the direction of gamma rays, so that the different detection modules can form a direction-related differential response; the shielding block includes tungsten blocks or lead blocks, etc.

[0038] The gamma detector module needs to have high detection efficiency and energy resolution, and support independent readout for each module. The shielding block and gamma detector modules are randomly arranged within a 3×3×3 cube; their quantity and spatial position together determine the system's directional encoding capability and angular resolution. The gamma detector module includes a scintillator-coupled photomultiplier tube (SiPM). The scintillator is a GAGG (Gadolinium Aluminum Gallium Garnet), and the photomultiplier tube is a SiPM.

[0039] The 3D cube structure gamma imaging subsystem is used to obtain the location of radiation hotspots in 4π space by utilizing the inherent response mode between the gamma ray incident direction and the detector, based on the response intensity of each gamma detector module, through image reconstruction algorithms.

[0040] In this embodiment of the invention, the gamma detector array 1 adopts a 3×3×3 cubic structure, with the shielding block and gamma detector modules randomly arranged, and the overall size is <30cm. 3In fact, there are no restrictions on the specifications and size of the gamma detector array 1. The gamma detector module uses a combination of GAGG blocks and SiPM. The output signal of SiPM is sent to the analog-to-digital converter (ADC) outside the 3D cube structure gamma imaging subsystem. The signal output by the ADC is connected to the FPGA for energy spectrum acquisition. Each radiation detector module operates independently without the need for calibration. Even if a radiation detector module fails, it will not affect the detector imaging. Only the response matrix needs to be calibrated in advance.

[0041] Based on the pre-calibrated response matrix, the response intensity of each detection module can be used to determine the...

[0042] 4π spatial inversion of radiation hotspot location.

[0043] The LVI-SLAM scene perception subsystem is used for multi-sensor fusion to achieve real-time map building and localization, and to construct a 3D scene point cloud with high-precision geometric structure and texture color information.

[0044] The fused sensors include LiDAR 2, a vision camera 3, and an inertial measurement unit (IMU) 4. The LVI-SLAM scene perception subsystem uses LiDAR 2 for precise ranging to acquire 3D point cloud information, vision camera 3 to acquire RGB images for feature extraction, and inertial measurement unit 4 to acquire pose information. Through a deep fusion algorithm of LiDAR-vision-inertial navigation, it constructs a 3D scene point cloud with high-precision geometric structure and texture color information in real time.

[0045] By employing a deep fusion algorithm combining laser, vision, and inertial navigation, the system overcomes the performance limitations of single sensors in environments with weak features (such as long corridors) or drastic lighting changes, enabling real-time construction of 3D scene point clouds with high-precision geometric structure and texture color information. Multi-camera stitching is used to achieve a wide field of view, and increasing the number of feature matches between adjacent frames improves the system's robustness.

[0046] In this embodiment of the invention, the visual camera 3 uses three monocular cameras stitched together to acquire wide-field-of-view 3D scene color information. Furthermore, from the perspective of visual odometry, the wide-field-of-view scheme ensures the stability of the SLAM system during drastic rotation of the radiometric mapping system, preventing mapping interruptions, drift, or loss of positioning. In fact, there are no limitations on the type or number of cameras used in the visual camera.

[0047] The data fusion and 3D radiation field reconstruction subsystem is used to align the radiation information detected by the 3D cube structure gamma imaging subsystem with the pose information acquired by the LVI-SLAM scene perception subsystem, and uses a multimodal data fusion algorithm to achieve real-time visualization of the radiation field.

[0048] The gamma detector array 1, lidar 2, vision camera 3, and inertial measurement unit 4 are all rigidly fixed on the support bracket 6 and maintain a fixed relative pose, forming an integrated multimodal measurement head; the output of SiPM in gamma detector array 1 is used to complete energy spectrum acquisition through the ADC and FPGA of data acquisition and processing unit 5; the data from lidar 2, vision camera 3, and inertial measurement unit 4 are synchronously transmitted to processing unit 5 for spatiotemporal alignment and SLAM calculation.

[0049] like Figure 2 As shown, the data fusion and 3D radiation field reconstruction subsystem includes a vision-IMU subsystem, a lidar-IMU subsystem, and a spatiotemporal alignment module. The reason for constructing subsystems (vision-IMU subsystem and lidar-IMU subsystem) with the IMU for vision and lidar respectively is that the IMU can provide high-frequency (about 230Hz) motion observation. On the one hand, it removes motion distortion for lidar point clouds and provides inter-frame motion priors for visual feature tracking. On the other hand, the IMU pre-integration factor realizes tight coupling between lidar, vision, and inertia, thereby providing continuous and robust pose estimation in scenarios where the performance of a single sensor degrades, such as weak features (e.g., long corridors) or drastic changes in illumination, avoiding mapping interruption and positioning drift.

[0050] The lidar-IMU subsystem employs a factor graph framework. First, numerical integration is performed using IMU data to construct IMU pre-integration factors, enabling state prediction of the radiometric mapping device. Then, linear interpolation is performed on the device state to obtain the device state at each point in the lidar point cloud data at the corresponding time, thereby transforming each lidar point to the coordinate system of the current lidar point cloud's starting point, achieving motion distortion removal of the lidar point cloud. Next, an inter-frame matching algorithm for the lidar point cloud is used to construct lidar odometry factors. Finally, the factor graph framework is used to fuse the IMU pre-integration factors and the lidar odometry factors to achieve state estimation of the mapping device.

[0051] In the vision-IMU subsystem, the global map constructed by the LiDAR-IMU subsystem is first rasterized. An octree-based incremental adaptive plane extraction algorithm is used to extract planar data from the raster, and the center points of the planes are projected onto the image to obtain projection points. Then, Harris feature points (Harris corner points) are extracted from the image. LK optical flow is used to simultaneously track the projection points and feature points, and Random Sample Consensus (RANSAC) is used to remove outliers in the tracking results, ensuring the accuracy of LK optical flow tracking. Next, a sliding window-optimized feature depth estimation algorithm estimates the depth values ​​of the selected feature points, and an epipolar geometry constraint-based method is used to recover feature points that failed optical flow tracking, ensuring sufficient disparity between feature points within the sliding window, thereby improving the accuracy of feature point depth recovery. Finally, reprojection error is constructed using both feature points and projection points, and this error is used as an observation for error state iterative Kalman filtering (ESIKF) to update the state, thereby improving the state estimation accuracy of the vision-IMU subsystem.

[0052] The spatiotemporal alignment module is used to match the measurement counts, energy spectrum, and orientation data of the 3D cube gamma detector array to the nearest factor map using sensor data timestamps, and combine them with the keyframes corresponding to the SLAM measurement of the 3D scene and location. It is used to spatiotemporally correlate the radiation orientation and intensity information of the gamma imaging subsystem with the motion trajectory, attitude, and 3D point cloud acquired by LVI-SLAM.

[0053] The data fusion and 3D radiation field reconstruction subsystem combines the detector's angular response with scene depth information, device position and attitude information, and projects the radiation distribution onto the 3D point cloud scene through nonlinear optimization or inversion algorithms to achieve real-time visualization of the radiation field.

[0054] Figure 3 A schematic diagram illustrating the alignment method between the gamma detector array 1, lidar, IMU, and camera information is presented. In one example, the lidar output frame rate is 10Hz, with each frame of the point cloud containing 20k points / frame, while the IMU frequency is 230Hz, and the visual camera's frame rate is 30fps. The IMU has the highest data output frequency, and its time is global time. An IMU pre-integration algorithm transforms multiple IMU data points within a single lidar point cloud frame into relative state constraints, thereby reducing the computational load of factor graph optimization. A single lidar point cloud frame contains tens of thousands of lidar points, and the output frequency of these lidar points is much higher than the IMU output frequency. Therefore, interpolation of the IMU state propagation results is necessary to obtain the state of each lidar point, thus removing motion distortion from the lidar point cloud. The sampling frequency of the gamma detector affects its statistical characteristics; it is assumed to be one sampling point every 3 seconds, a frequency much lower than other sensors. An interpolation method is proposed to map this frequency to adjacent data frames.

[0055] In summary, the method for aligning the information of the gamma detector array 1, lidar, IMU, and camera is to use the IMU time with the highest frequency as the global clock: Multiple IMU data within a single frame of lidar point cloud are converted into relative state constraints through IMU pre-integration, and the IMU state propagation results are interpolated to obtain the pose of each lidar point at the corresponding time to remove point cloud motion distortion. Since the gamma detector array 1 has the lowest sampling frequency (approximately 3 seconds per point), it is matched and interpolated to the nearest data frame based on the timestamp. This achieves spatiotemporal alignment of the four types of information—gamma detector, lidar, IMU, and visual camera—under a unified global clock.

[0056] Example 2

[0057] like Figure 4 As shown, this embodiment of the invention provides a radiometric mapping method based on a 3D Rubik's Cube structure gamma imaging detector. The method is based on the radiometric mapping system based on a 3D Rubik's Cube structure gamma imaging detector described in Embodiment 1 of this invention and any optional embodiment thereof. The method includes the following steps:

[0058] S1. The detected radiation information, as well as the acquired 3D point cloud information, feature information, and pose information, are spatiotemporally synchronized and aligned with coordinates.

[0059] Only by aligning the timestamps of information acquired by the gamma detector array 1 with those from the vision camera, lidar, and inertial measurement unit, jointly calibrating the external parameters, and transforming the coordinates to the world coordinate system, can LVI-SLAM be calculated and a radiation observation model be constructed.

[0060] S2. Input the radiation information after spatiotemporal synchronization and coordinate alignment, as well as the acquired 3D point cloud information, feature information and pose information data, into the 3D construction and radiation field reconstruction platform;

[0061] The radiation information data detected by the γ detector array 1 is input to the count rate / energy spectrum of each detector module, as well as the γ source orientation (θ, φ) estimated using the response; the LVI-SLAM input is the system pose trajectory translation matrix T(t), rotation matrix R(t), and three-dimensional point cloud and texture color information.

[0062] It should be noted that the radiation information data, LVI-SLAM pose and point cloud data of the γ detector array 1 input in step S2 are the same batch of data that were spatiotemporally synchronized and aligned in step S1. The spatiotemporal synchronization and alignment were performed first, and then the data input was performed. This is to ensure that the data from each sensor are registered under a unified clock and world coordinate system before participating in the construction of the subsequent radiation observation model.

[0063] S3. Construct a radiation observation model based on the input data;

[0064] Using the γ source orientation and energy spectrum information acquired by γ detector array 1, and combined with the 3D scene obtained by LVI-SLAM, a detection information response function or computational model based on distance attenuation and shielding is constructed. The 3D scene point cloud is then voxelized or meshed.

[0065] S4. Based on the constructed radiation observation model, perform three-dimensional radiation scene visualization and fusion analysis.

[0066] By employing nonlinear optimization least squares / Bayes estimation, sparse regularization, and noise suppression, and through real-time iterative updates, the radiation distribution is mapped onto a 3D scene for hotspot localization, dose estimation, point cloud / mesh coloring, uncertainty labeling, etc., ultimately enabling hotspot pricing, trajectory playback, and radiation scene assessment.

[0067] As can be seen from the above embodiments, the radiation mapping system and method based on a 3D cube structure gamma imaging detector disclosed in this invention has the technical advantages of high integration, lightweight, wide field of view, and high imaging efficiency, and can realize real-time measurement and fusion analysis of high-precision three-dimensional scene radiation distribution without proximity.

[0068] The system and method described in this invention are not limited to the embodiments described in the specific implementation. Other implementation methods derived by those skilled in the art based on the technical solution of this invention also fall within the scope of technical innovation of this invention.

Claims

1. A radiometric mapping system based on a 3D Rubik's Cube structure gamma imaging detector, characterized in that: The system includes a 3D Rubik's Cube structure gamma imaging subsystem, an LVI-SLAM scene perception subsystem, and a data fusion and three-dimensional radiation field reconstruction subsystem. The 3D Rubik's Cube structure gamma imaging subsystem is used to detect radiation information using a gamma detector array; the LVI-SLAM scene perception subsystem is used to output the system's motion trajectory, pose, and textured three-dimensional point cloud. The data fusion and 3D radiation field reconstruction subsystem is used to perform spatiotemporal alignment of the radiation information detected by the 3D cube structure gamma imaging subsystem based on the global clock and pose of the LVI-SLAM scene perception subsystem, and then map it onto the 3D point cloud scene through inversion projection.

2. The radiation mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 1, characterized in that: The radiation information detected by the gamma detector array includes the direction and intensity of radiation hotspots.

3. The radiation mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 2, characterized in that: The 3D cube structure gamma imaging subsystem includes a shielding block and a gamma detector module. The shielding block and the gamma detector module are arranged randomly in 3D space to form a cube structure, which constitutes a gamma detector array.

4. The radiometric mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 3, characterized in that: The 3D cube structure gamma imaging subsystem is also used to obtain the location of radiation hotspots in 4π space by utilizing the inherent response mode between the gamma ray incident direction and the detector, based on the response intensity of each gamma detector module, through image reconstruction algorithms.

5. The radiometric mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 1, characterized in that: The LVI-SLAM scene perception subsystem uses LiDAR for precise ranging to acquire 3D point cloud information, a visual camera to acquire RGB images for feature extraction, and an IMU to acquire pose information. Through a deep fusion algorithm of LiDAR, vision, and inertial navigation, it constructs a 3D scene point cloud with high-precision geometric structure and texture color information in real time.

6. The radiation mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 5, characterized in that: The visual camera uses multi-camera stitching to achieve a wide field of view and increases the number of feature matches between adjacent frames.

7. The radiation mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 4, characterized in that: The data fusion and 3D radiation field reconstruction subsystem uses an IMU pre-integration algorithm to convert multiple IMU data within a frame of lidar point cloud into relative state constraints. The results of IMU state propagation are interpolated to obtain the state of each lidar point. The radiation information detected by the gamma detector array is to be mapped to the adjacent data frame through the interpolation method.

8. The radiation mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 3, characterized in that: The shielding block consists of a tungsten block or a lead block.

9. The radiation mapping system based on a 3D Rubik's Cube structure gamma imaging detector according to claim 3, characterized in that: The gamma detector module includes a scintillator linked to a photomultiplier tube.

10. A radiometric mapping method based on a 3D Rubik's Cube structure gamma imaging detector, and a radiometric mapping system based on a 3D Rubik's Cube structure gamma imaging detector as described in any one of claims 1-9, characterized in that, The method includes the following steps: S1. The detected radiation information, as well as the acquired 3D point cloud information, feature information, and pose information, are spatiotemporally synchronized and aligned with coordinates. S2. Input the radiation information after spatiotemporal synchronization and coordinate alignment, as well as the acquired 3D point cloud information, feature information and pose information data, into the 3D construction and radiation field reconstruction platform; S3. Construct a radiation observation model based on the input data; S4. Based on the constructed radiation observation model, perform three-dimensional radiation scene visualization and fusion analysis.

Images