光学技术, 2015, 41 (6): 528, 网络出版: 2016-01-25
团块建模与粒子滤波相结合的行人目标视觉跟踪
Person tracking based on combination of blob modeling and particle filter
摘要
针对粒子滤波算法在遮挡情况下导致视觉跟踪不稳定甚至丢失目标的问题,提出了一种基于团块建模与粒子滤波相结合的目标跟踪算法。首先通过图像分割的方法得到视频帧中的初始目标,并构建目标团块模型;然后基于多团块目标信息并结合粒子滤波算法进行分块跟踪;最后利用高斯加权的方式,得到最终的目标预测位置。实验结果表明,该算法具有较强的鲁棒性,尤其是在遮挡的情况下能够实现目标的稳定跟踪。
Abstract
The particle filter algorithm may increase the prediction error due to the unstable tracking,and lose the object under occlusion. A new local feature based method is proposed,which combines the blob modeling and particle filter together. The initial target of video frame is obtained by the image segmentation,and is denoted as the blob model. The tracking is based on the multiblob target information. Finally,it uses the approach of Gaussianweighted and integrates the tracking results in order to obtain the predicted position of the target. Experimental results indicate it is robust and has good performances under occlusion.
孔鹏, 丁辉, 尚媛园, 王琳, 周修庄, 付小雁. 团块建模与粒子滤波相结合的行人目标视觉跟踪[J]. 光学技术, 2015, 41(6): 528. KONG Peng, DING Hui, SHANG Yuanyuan, WANG Lin, ZHOU Xiuzhuang, FU Xiaoyan. Person tracking based on combination of blob modeling and particle filter[J]. Optical Technique, 2015, 41(6): 528.