Deep Kalman-based Trajectory Estimation of Moving Target from Satellite Images

Trajectory estimation of moving target from satellite images facilitates analysis of target characteristics. Aiming at this purpose, this paper proposes a deep Kalman-based trajectory estimation method, which combining the learning ability of deep learning method and the noise filtering ability of Kalman filter method. Through the analysis of target motion, a state space model of target motion is established. Then the deep learning method are used to extract the features of the target from satellite images, and the features are introduced in Kalman filter method to estimate the trajectory. Finally, simulation results show that the proposed trajectory estimation method is superior in terms of estimation continuity and accuracy.