Two-Stage Attitude Estimation of Satellite from ISAR Images

As more satellites are launched into space, space security has become an important issue, and the attitude estimation of space targets has been widely studied. Many recent works extract typical features of targets from ISAR images and utilize projection relationship to estimate the attitude. However, it is challenging to extract the features of typical components due to factors like occlusion. And some of these methods are based on conditions that are difficult to meet in reality. In this paper, we propose a two-stage attitude estimation approach for satellite with complex rotation conditions. The first step is to construct a 2D-3D correspondence relationship through convolutional neural network, and the second step is to obtain the satellite attitude according to the projection relationship and 2D-3D correspondence relationship. In order to verify the effectiveness of the proposed approach, a satellite with complex rotation conditions is simulated under different view angles. And the experimental results demonstrate the effectiveness of the proposed method.