Effective Multi-Frame Optical Detection Algorithm for GEO Space Objects

Abstract

The limited resource of Geostationary Earth Orbit (GEO) is precious and most telecommunication, weather and navigational satellites are placed in this orbit. In order to guarantee the safety and health of active satellites, advanced surveillance and warning of unknown space targets such as space debris are crucial. However, space object detection still remains a very challenging problem because of the weak target characteristics and complex star background. To solve this problem, we conduct a deep-learning-based framework called PP-YOLOv2 for single-frame object detection and design a post-processing algorithm named CFS for further candidate filtration and supplement. First, we transform the label information and generate the according bounding boxes to train the PP-YOLOv2 detector to extract candidate coordinates for each frame. Then, the CFS technique is designed as an effective post-processing procedure to obtain the eventual prediction results. Experiments were conducted over a dataset from the Kelvins SpotGEO challenge, which demonstrate the effectiveness and the comparable detection performance of our proposed pipeline. Finally, the deployment results on NVIDIA Jetson Nano show that the proposed method has a competitive application prospect for a space target monitoring system.

A Robust Space Target Extraction Algorithm Based on Standardized Correlation Space Construction

“…The effectiveness will decrease a lot when detecting space targets in actual scenes. Although some methods trained the model with actual observed datasets like SpotGEO [26], [27], [54], the limitation of data will result in domain shift and influence the effectiveness in other actual observed datasets. To improve the generalization of the three mentioned deep learning methods, we used 6500 simulated and actual observed images to train networks.…”

“…Other researchers tried to design an end-to-end CNN for source extraction or target detection. Representative object detection networks like YOLO [26], [27] and Fast-RCNN [28] are used to detect objects when the morphology of surrounding stars is different. For example, Jia et al used the Faster-RCNN to detect and classify space targets in wide-field small aperture telescopes [28].…”

“…For example, Jia et al used the Faster-RCNN to detect and classify space targets in wide-field small aperture telescopes [28]. Dai and Jiang et al tried to use a YOLO-based network for geostationary (GEO) target detection [26], [27]. There are only a few pixels for some faint targets.…”