Assessment of radiation dose received by nuclear plant personnel through a video-based surveillance system

Abstract

This work aims at developing a video-based surveillance system for safety purposes in nuclear plants. The objective is to assess the radiation dose received by nuclear plant personnel, while they execute daily tasks, by means of computer vision methods. The system is conceived to provide some redundancy to the radioprotection means currently in use, being independent and complementary to them. After evaluating some methods from the literature for automatic target detection and tracking, a novel system is developed to correctly detect, track and identify people, so that the radiation dose received by each person is reliably computed. The video data are supplied by cameras installed in the nuclear plant room. Radiation dose rate mapping is combined with the tracking results to account for the received doses. We provide experimental results from a research reactor room, which show that the proposed system achieves radiation dose estimates that are in general similar to the ones of the ground truth. The database developed in this work for performance evaluation has been made publicly available for the research community.