一种基于PSO的改进光流算法

An Efficient Optical Flow Method Based on PSO Algorithm

针对变分光流法无法有效检测由间断、遮挡等因素造成的错误光流分量的缺陷,提出一种基于PSO(Particle Swarm Optimization)的光流算法。该方法在Classic+NL算法模型的基础上计算出光流后,引入前向光流和后向光流的运动一致性理论来判断遮挡区域,并通过基于PSO的修补法来实现对遮挡区域错误光流的有效修补,同时,利用邻域光流修补法实现了再次修补。实验结果表明,该方法能有效克服由间断、遮挡等因素造成的错误光流分量的缺陷,更准确地刻画出光流,提高光流的计算精度。

For variable optical flow methods cannot effectively detect wrong optical flow caused by discontinues and occlusions. In this case, this paper proposed an efficient optical flow method based on PSO（Particle Swarm Optimization） algorithm. Firstly, we calculated optical flow using Classic＋ NL model and then introduced the motion consistency of the forward and backward flow vectors which have motion uniqueness to determine occlusion area. In order to effectively repair the wrong optical flow in occlusion area,we used a PSO-based repair method. Furthermore, we completed doubble repair of wrong optical flow in occlusion area by using neighborhood repair method. The experimental results proved that our approach can effectively overcome the impact made by wrong optical flow in occlusion areas, more accurately represent the optical flow and improve the calculation precision of optical flow.