The article presents an in-situ clouds-powered radioactive source detection and localization approach, namely color-depth-radiation (CDR) Mapping, using 3-D land mapping within hazardous indoor environment and incorporating sensor fusion between a RGB-D camera and a portable radiation detector. In the approach, to achieve fast and robust image registration, color images detected by the camera are initially employed to extract crucial visual features and establish pairs of matched image features between successive scanned images. Following this, matched features are incorporated with the corresponding calibrated depth information to generate 3-D keypoint cloud pairs. To remove potential noises in the acquired datasets, a novel geometric-based filtering algorithm is developed to reject incorrect keypoint pairs prior to ICP-based image registration. Most importantly, an algorithm to determine the radioactive sources’ parameters including strength and 3-D position is developed for accurate radioactive source detection and localization. With this, the radioactive sources can be accurately pinpointed in the established 3-D map for efficient contamination control and safety management. Two radiation testing experiments were performed to verify the feasibility of the approach and its detection accuracy. The simulation results indicate that the proposed approach can reach up to 95% accuracy of radiation source localization incorporated in the 3-D map.
The article presents an in-situ clouds-powered radioactive source detection and localization approach, namely color-depth-radiation (CDR) Mapping, using 3-D land mapping within hazardous indoor environment and incorporating sensor fusion between a RGB-D camera and a portable radiation detector. In the approach, to achieve fast and robust image registration, color images detected by the camera are initially employed to extract crucial visual features and establish pairs of matched image features between successive scanned images. Following this, matched features are incorporated with the corresponding calibrated depth information to generate 3-D keypoint cloud pairs. To remove potential noises in the acquired datasets, a novel geometric-based filtering algorithm is developed to reject incorrect keypoint pairs prior to ICP-based image registration. Most importantly, an algorithm to determine the radioactive sources’ parameters including strength and 3-D position is developed for accurate radioactive source detection and localization. With this, the radioactive sources can be accurately pinpointed in the established 3-D map for efficient contamination control and safety management. Two radiation testing experiments were performed to verify the feasibility of the approach and its detection accuracy. The simulation results indicate that the proposed approach can reach up to 95% accuracy of radiation source localization incorporated in the 3-D map.