An Improved Camshift Target Tracking Algorithm Based on Joint Color-Texture Histogram

Hongxia CHU,Zhongyu XIE,Kejun WANG

DOI: https://doi.org/10.7652/xjtuxb201803020

2018-01-01

Abstract:In order to solve the problem of tracking failure when the target is near the background color or the target is occluded,an improved Camshift target tracking algorithm is proposed in this paper.First,the calculation of histogram for the improved algorithm model used the probability distribution histogram with the fusion of color and texture and hence it solved the problem that using a single color model is difficult to adapt to the change of background objects caused by large range of motion and occlusion.Secondly,the weight image can be calculated from the square root of the ratio of the feature probability of target model to that of candidate target model.The calculated weight was used to further estimate the position and direction of the target.It overcomes the shortcomings of the original Camshift algorithm that only relies on the target model in the calculation of weight image and greatly reduces the influence of background features on tracking.At last,the state of moving object is estimated by particle filter to overcome the occlusion,interleaving or overlap,and then the tracking accuracy of target position is improved.The average success rate of the improved algorithm is more than 50％,and the average central position error is less than 20％.Experimental results showed that the algorithm can obviously improve the performance of target tracking,and achieve the target tracking effectively and accurately.