In the satellite-to-satellite optical communications, two optical terminals need to keep precise tracking for the beacon each other. However, in the tracking stage, the beacon recognition can be influenced by the star background light, which will result in the decline of tracking precision or the loss of the beacon. To solve these problems, we propose the centroid coordinate sequence fast Fourier transform algorithm, which distinguishes the beacon and the star background light by the frequency feature. Furthermore, for improving the tracking precision, the prediction method for the beacon is also introduced. To study the tracking precision of the algorithm, we simulated the satellite platform vibration. Finally, an experiment was performed to verify the practicability of the algorithm. Compared with the tradition method, the results showed that the prediction accuracy was improved more obviously. When the vibration frequencies are 50 and 80 Hz, the tracking precisions were improved by 58% and 33%, respectively. The new algorithm will be beneficial for improving the tracking precision and maintaining the steady laser link.