3D radiation field reconstruction for multiple unknown radioactive sources based on limited measurements

In recent years, nuclear energy has played an important role in terms of energy structure optimization and energy security. In order to reduce the radiation exposure of occupational technicians and obtain radiation intensity distribution in the environment, it is essential to reconstruct the three-dimensional (3D) radiation field. However, in some scenes, especially those with multiple radioactive sources, how to accurately reconstruct the 3D radiation field using limited measurements remains a major challenge. This paper explores a novel 3D radiation field reconstruction method based on back-propagation neural network and genetic algorithm to accurately reconstruct the 3D radiation field of multiple radioactive sources with limited measurements. First, the volume of interest is represented as an octree map. Then, the radiation dose distribution of radioactive sources in the octree map is obtained by Monte Carlo (MC) simulation method, and multiple sets of radiation data are collected at a low sampling rate by the random sampling method as the radiation dataset. Further, the radiation dataset is fed into the designed network architecture optimized by genetic algorithm to fit the missing dose rates in the octree map. The feasibility of the proposed method is demonstrated through three representative cases. The experimental results show that in open indoor scenes, the average relative error of the proposed method is less than 2.73% using only 1.625% of measurement data, which is reduced by 29.27% compared with the traditional Gaussian process regression (GPR) method; in indoor scenes with obstacle shielding, the average relative error of the proposed method is less than 3.01%, which is reduced by 30.65% compared to the GPR method. The experimental results reveal the important practicality of our proposed method for 3D radiation field reconstruction tasks with multiple radioactive sources.