一种基于稀疏表达和光流的目标跟踪方法

提出一种将稀疏表达技术融入到传统光流算法的目标跟踪方法.首先使用FAST和Harris算法在视频序列的每一帧中为光流算法采集运动目标的特征点,之后光流算法基于后向跟踪-形心配准的方法对跟踪目标完成粗略定位.在当前帧的粗略定位处应用仿射变换产生N个候选区域.最后应用稀疏表达技术判断出与原始目标匹配率最高的仿射变换区域做为最终目标跟踪区域.实验结果表明,该算法既能较好地适应目标的外观变化,又具有较强的抗遮挡能力,鲁棒性强.

Tracking Object Based on Sparse Representation and Optical Flow

A novel tracking algorithm that merges sparse representation with the optical flow is presentsed Fast and Harris algorithms are used to pick up the corner points of motion targets in every video image frame for the optical flow method.Based on backward tracking and object cantroid registration,the approximately tracking result is achieved.Affine transformation is employed draw N candidates around the approximately tracking result in the current frame,and a sparse representation method is developed to compute the highest similarity between the candidates and the original target.Experimental results are presented to demonstrate that the algorithm can be adaptive to the appearance changes as well as occlusions with strong robust.