A novel approach for space debris recognition based on the full information vectors of star points

Abstract

The recognition and detection of space debris has become one of significant research fields recently. Compared with natural images, effective information are very few contained in star images. In the past years, the gray values of star points and the continuity of sequential star images are utilized by numerous algorithms to carry out the recognition and detection through fusion of consecutive star images, which have been achieved good performance. However, with the rapid increase of star image data, those algorithms seem to be inadequate in recognition ability. In this paper, we propose one novel approach based on the full information vectors of star points to recognize moving targets with the machine learning method which is never utilized in space debris recognition field. Besides gray values, we further deeply excavate the characteristics of each star point in a single frame by the equal probability density curve of Gaussian distribution. The elliptical pattern characteristic vectors of star points can be input into the machine learning method for classification of static stars and moving targets in a single frame. Finally, trajectories of moving targets can be determined within 3 frames by the full information vectors. Therefore, traditional processing methods are abandoned and the proposed brand new approach redefines the recognition technical route of space debris. The experimental results demonstrate that moving targets can be successfully recognized in a single frame and the coverage rate of moving targets can reach 100%. Compared with other traditional methods, the proposed approach has better performance and more robustness.