Space Target Spin Motion Recognition Based on 3D Convolutional Networks
Yanjin Zhang, Xuejian Feng, Xiao Wei, Chenxi Zhu, Chaoying Huo, Hongcheng Yin · 2023
As aerospace technology continues to progress, an increasing number of space targets are being launched into orbit, putting pressure on limited space resources. Consequently, detecting and identifying the anomalous motion of these space targets has become a pivotal necessity in the field of space situational awareness. This paper aims to recognize spin motion patterns exhibited by anomalous space targets using three-dimensional Convolutional Networks (3D ConvNets). Leveraging the Convolutional 3D (C3D) network and Two-Stream Inflated 3D ConvNet (I3D), spatio-temporal information is extracted from sequences of Inverse Synthetic Aperture Radar (ISAR) image and optical flow features. This facilitates the effective recognition of spin motion performed by space targets. The promising results from our experiments validate the feasibility of applying 3D ConvNets for the task of recognizing motion patterns in space targets.