一种改进的Camshift视频目标跟踪算法

鉴于连续自适应均值漂移(Camshift)算法在光照变化,相似背景颜色干扰及目标遮挡时鲁棒性不高,易造成跟踪错误等问题,提出了一种联合多特征和最大类间方差法的视频运动目标跟踪算法。该算法将色度直方图、梯度方向直方图和LBP纹理特征进行巧妙的融合,构建了一种高效的联合直方图目标外观特征模型,并在Camshift算法中嵌入最大类间方差法,增强目标和背景的区分度。不同场景的视频跟踪结果表明,改进算法有效克服了传统Camshift算法应对光照变化、颜色干扰和目标遮挡的缺点,与同类算法相比,具有更高的准确度和鲁棒性。

Video target tracking algorithm based on improved Camshift

Under illumination variation,similar background interference and target occlusion,continuous adaptive mean shift(Camshift)algorithm has low robustness and is easy to track astray.For solving this problem,a video target tracking algorithm based on multi-feature and Otsu algorithm is presented.This algorithm fuses color histogram,histogram of oriented gradient and LBP texture feature.An efficient target appearance characteristic model based on joint histograms was built.Meanwhile,Otsu algorithm was embedded to Camshift algorithm,which can enhance distinction degree between target and the background.Video target tracking results of different scenes show that the improved algorithm can effectively overcome the disadvantages of the traditional Camshift algorithm,such as illumination variation,similar background interference and target occlusion.Compared with the other algorithms,the improved algorithm has higher robustness.