Faint Debris Detection by Particle based Track-before-detect Method

This study aims at developing a particle method to detect faint debris, which is hardly seen in single image, from an image sequence based on a track-before-detect (TBD) method. TBD methods try to track targets without explicitly detecting the signals followed by evaluation of goodness of each track and obtaining detection results. This study proposes a particle based TBD (P-TBD) method consisting of tracking of target state and heuristic search of initial state. The target tracking is implemented by a particle filter (PF). PF is an optimal filtering technique that can estimate the target state once an initial distribution of target state is given as a prior knowledge. An evolutional algorithm (EA) is utilized to search the initial distribution. The EA iteratively evaluates goodness of initial particles for the same image sequences and resulting set of particles is used as an initial distribution of PF. This paper designs the P-TBD method and verifies it. The P-TBD method is applied to image sequences acquired during observation campaigns dedicated to GEO breakup fragments, which would contain a sufficient number of faint debris images. The results indicate the feasibility of tracking faint debris by the proposed method.