The moving planets have a few of pixels which can lead to lacking enough image characteristics when tracking them. So we propose the motion planets detection and tracking algorithm based on gestalt principle. We structure Gaussian mixture model to detect the motion area from the visual cognition perspective, and use astronomy images graphics characteristics to confirm the planet position. Then we propose space-time fusion model for tracking planet. All the experiments use 1000 16-bitframe images of wide view CCD camera which contain 5684 motion planets in total. The results show that our algorithm reaches an accuracy of 94% and has robustness.