Radioactive Area Reconstruction With Multi-visual Information Fusion

The focus of the reconstruction of the radioactive area is to determine the type,location,and pollution boundary of radioactive materials,which can help operators intuitively and accurately analyze the nuclear radiation environment,and improve decommissioning decontamination and emergency disposal operations effectiveness.In response to the need for visualization of radioactivity distribution information in the process of nuclear facility decommissioning and nuclear emergency disposal,this paper proposes a method of radioactive area reconstruction with multi-visual information fusion.Based on visual SALM technology,the γ camera is used to detect the radioactive distribution information in the unknown environment,and the computer vision related theories are used to reconstruct the radioactive area.This article focuses on the radioactive area reconstruction technology,and the main work is as follows:The multi-visual information fusion system is constructed based on the γ camera and Kinect equipment.Aiming at the problem of low detection efficiency of traditional pinhole collimator γ camera,the imaging of γ camera based on MURA is studied.Starting from the principle of coded aperture imaging,the MURA nested code board is designed,and the algorithm of related decoding matrix is studied.Based on the Kinect camera and the visual SLAM algorithm framework,the dense point cloud of radioactive environment is constructed.The nuclear radiation scene images are filtered and then input to the front-end visual odometer to estimate the camera pose between frames.The back end uses a series of nonlinear optimization and loop detection methods to optimize the key frame pose and map points.Finally,the dense point cloud map is constructed with the help of precise key frame data.The visual equipment of the multi-visual information fusion system is calibrated.Based on the pinhole camera model,the imaging principle and calibration principle ofγ camera and Kinect V2 camera are analyzed.The combined imaging model of the two cameras is constructed and its joint calibration is completed,laying a solid foundation for subsequent fusion.In order to achieve three-dimensional reconstruction of the radioactive area,a hotspot3 D reconstruction algorithm is designed based on the Visual Hull principle.The algorithm extracts hotspot contours by analyzing the radioactive features of the γcamera image,and uses a few multi-view γ camera images to calculate the hotspot 3D minimum bounding box,and then divides the bounding box into voxels to determine whether the voxel belongs to the hotspot model.The algorithm realizes 3D reconstruction of hotspot,and realizes fusion with dense point cloud of radioactive environment based on fusion principle and the radioactive source is located.The two algorithms of δ decoding and fine sampling balanced decoding are compared and studied in reconstructing point source image quality using simulation experiments.The radioactive area reconstruction experiment is carried out in the real radioactive environment,and the point cloud map of the radiation scene including the 3D distribution model of radioactive sources is reconstructed and realized the 3D visualization of the radioactive area.The root-mean-square error of the point source localization is 0.076 m,which verifies the effectiveness of the proposed method.