Real time motion compensation technology based on least square support vector machine prediction

When cameras are used in aerial photography, satellite imaging or other scenes, the motion of the observational target causes image blur. The corresponding motion compensation systems often present some response delay. Thus, we introduce effective and fast motion prediction for the target to achieve steady real-time motion compensation. We first analyze the target motion states to propose a fast and robust prediction method based on the least square support vector machine algorithm. Then, we evaluate the performance between ours and other state-of-the-art methods through experiments. Experimental results confirm that the proposed method provides a fast and robust prediction for target motion. At last, we embed our method with dual-resolution camera system to perform high-quality motion compensation effect in real time.