Attitude and Size Estimation of Satellite Targets Based on Imaging Geometry Analysis for Space Observation

Attitude and size estimation of satellite targets is crucial for space surveillance. Existing methods that utilize radar or optical images from ground-based sensors typically rely on the basic imaging projection model. These methods require long-time or multistation observation to provide sufficient views and struggle to achieve accurate estimations under conditions of small view differences. To mitigate reliance on extensive views, we propose a method based on imaging geometry analysis for space observation. In our work, five crucial properties about the imaging geometry of rectangular solar panels are explored. Compared with the basic projection model, these properties impose additional constraints on the estimation process, thereby decreasing estimation errors by approximately 50% under limited observation conditions. Then, guided by these properties, we establish an over-determined system of equations to estimate two side vectors of the rectangle from their projections in images. To solve this complex system, the coarse estimation with a weighting strategy is conducted to decouple the system into linear subsystems, and these subsystems are solved by an Alternating Iterative Reweighted Least Square algorithm. The accuracy, real-time performance, and noise robustness of our method are demonstrated by experiments under various observation conditions.