基于自适应压缩跟踪算法的视觉目标跟踪

Visual Target Tracking Based on Adaptive Compression Tracking Algorithm

针对压缩跟踪算法在目标发生遮挡、快速移动、有相似目标情况存在跟踪漂移的问题,提出了基于卡尔曼滤波的自适应学习压缩跟踪算法.该算法首先利用压缩跟踪算法对目标进行定位,然后根据跟踪结果的置信图对分类器参数自适应更新,当判定目标严重遮挡时,利用卡尔曼滤波进行预测估计.实验结果表明,该算法相比目前先进的算法有更好的跟踪精度和鲁棒性,且算法平均跟踪速度39帧/s,能够满足实时性的要求.

Aiming at the problem that the compression tracking algorithm has the effect of occlusion,fast moving and similar target,we propose the adaptive learning compressive tracking based on Kalman filter（ALCT-KF）.Firstly,the target is located by using the compressive tracking algorithm.Then,the classifier parameters are updated adaptively according to the confidence map of the tracking result.When the target is heavy occlusion,the Kalman filter is used to predict and estimate the position of object.Experimental results show that the proposed algorithm has better tracking accuracy and robustness than current advanced algorithms.And the average tracking speed of ALCT-KF is 39 frames per second,which can achieve real-time performance.