Koki Fujita, Kyushu University, Naoyuki Ichimura, National Institute of Advanced Industrial Science and Technology (AIST), Toshiya Hanada, Kyushu University
Keywords: Space Debris, Trajectory Detection, Image Processing, Earth-based Observation
Abstract:
This paper presents a novel method to detect trajectories for Geosynchronous Earth Orbit (GEO) debris in image sequences. As far as Earth-based observation is concerned, detecting GEO debris is not so easy because debris often appear very faintly in image frames. A simple but effective way to detect such faint debris is decreasing a threshold value of binarization applied to an image sequence as preprocessing. However, a low threshold value of binarization leads to extracting a large number of image objects other than debris, which become obstacles to detect debris trajectories and cause high computational load. In order to detect debris from binarized image frames with a mass of obstacles, this work proposes a method that numerically evaluates the characteristics of a line segment connecting two image objects on different image frames or a candidate of debris trajectory. The proposed method adopts a cascading numerical evaluation based on the physical properties of GEO debris image. In the first stage of the cascading evaluation, the candidates of debris trajectories are selected using a displacement between two image frames for each line segment. A Monte Carlo simulation of a target breakup event derives the threshold for the displacement, and the lines with displacements exceeding the threshold are removed. The second stage evaluates a direction of a line, i.e., a direction of motion in an image sequence. The lines with motion close to the horizontal direction are rejected, because they almost correspond to stars. The continuity of trajectories is utilized in the third stage. After the points of intersection of the selected lines on all image frames are computed, it is possible to check whether image objects can be found on the points. The continuity of each line is evaluated by the count of found image objects, and the lines are finally detected as debris trajectories if the count exceeds the threshold determined in proportion to the total number of image frames. Since the order of the numerical evaluation can adjust the performance to remove obstacles, the proposed cascading numerical evaluation efficiently detects GEO debris. The effectiveness of the proposed method is demonstrated by experiments using real observation image sequences obtained from the telescopes at Lulin Observatory in Taiwan.
Date of Conference: September 15-18, 2015
Track: Poster