Image-based Targets Tracking for Multiple Remote Satellites in Complex Stellar Background

With the development of the LEO satellite constellations such as OneWeb and Starlink, the number of satellites is increasing rapidly. To improve the monitoring capability, it is necessary to develop an advanced recognition and tracking technology for multiple targets of the LEO satellite constellations. Space-based multi-targets tracking has two difficulties. First, the targets appear as dots in the image, and have no contours, textures and shape features. Second, the star would be misjudged as a target when satellite approaches a star during its motion. Aiming at the monitoring problem of satellites in a giant constellation, an GM-PHD tracking method based on target grayscale measurements is proposed. Firstly, the information of point targets is expanded by Gaussian diffusion processing to solve the problem of multi-target states description. Then, the peak value of the product of the predicted PHD and the likelihood function is used as the measurement information, and the dynamic weight threshold is calculated accordingly. Using this threshold for target screening can reduce the false judgment rate. Simulation results show that the proposed method can improve the recognition accuracy of target number and reduce the tracking error of target position.