中文摘要：

针对压缩跟踪(CT)算法不能解决跟踪目标形变、被遮挡、光照变化等问题,提出改进的压缩跟踪(ICT)算法.采用卡尔曼预测下一帧中的目标状态,从而减小搜索域,并在目标被遮挡时估计运动轨迹;采用定向二进制特征(ORB)匹配算法,跟踪形变目标和判断目标是否被遮挡;采用ORB匹配跟踪、CT检测和贝叶斯学习相结合的方法,融合ORB匹配跟踪和CT检测,输出最优结果,减弱光照变化的影响,提高跟踪帧率的同时增强鲁棒性.实验结果表明:ICT算法能准确地跟踪形变及被遮挡目标,跟踪效果在多种数据集上表现出更高的鲁棒性和精确性,平均帧率达到74.137Hz,具有良好的实时性.

英文摘要：

To solve the problems that CT (compressive tracking) algorithm does not consider deform‐ation ,occlusion ,illumination changes and other issues ,an ICT (improved compressive tracking ) al‐gorithm was proposed .Using Kalman ,the target state in the next frame was predicted in order to re‐duce the search field and the target trajectory was estimated when the target was occlusive .ORB (ori‐ented binary robust independent elementary features ) matching algorithm was applied to track the de‐formation target and determine the occlusion of target .The ORB match tracking ,CT detection and Bayesian learning were combined ,so that the optimal result could be output through fusing ORB match tracking and CT detection .Thus ,the influence of illumination changes was weakened ,tracking frame rate was improved ,and robustness was enhanced .Experimental results show that the ICT algo‐rithm can accurately track the target under deformation and occlusion ,tracking results of w hich in a variety of tracking data sets exhibit a higher robustness and accuracy ,and the ICT algorithm has a good real‐time with the average frame rate achieving 74 .137 Hz .