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.%针对背景中存在颜色相近目标或目标被遮挡时Camshift算法跟踪失败的问题,提出了一种改进的Camshift目标跟踪算法.首先,改进算法模型直方图的计算选用颜色和纹理相融合的直方图概率分布,解决了Camshift算法只使用单一的颜色模型、很难适应物体大范围运动造成的背景变化或遮挡的不足;其次,图像权值采用目标模型与目标候选模型特征概率之比的平方根来计算,并用权值进一步估计目标的位置和方向,克服了原始Camshift算法中图像权值仅依靠目标模型计算的不足,大大减少了背景特征对跟踪的影响;最后,利用粒子滤波对运动目标状态进行估计,以克服目标运动引起的遮挡、交错或重叠,进而提高目标位置跟踪精度.实验结果表明,改进算法的平均每帧成功率达到50％以上,平均中心位置误差低于20％.改进算法能有效改善目标跟踪性能,从而实现目标跟踪的有效性、准确性.
展开▼
