一种基于算法融合的运动目标跟踪算法

视频目标跟踪的重点和难点在于如何快速、准确的匹配目标.针对复杂背景下,单一模式的跟踪算法不能准确跟踪目标的问题,提出了一种基于算法融合的运动目标跟踪方法.该算法综合运用去均值相关跟踪算法（normalized cross correlation,NCC）和均值漂移算法（MeanShift）2种基本模式对输入输入视频进行处理,结果送入Kalman滤波器进行滤波与预测,最后根据最小总均方误差准则进行自适应融合.试验证明,该算法能够较好地实现复杂场景条件下的目标跟踪,提高了跟踪的鲁棒性.

Moving targets tracking algorithm based on algorithm-fusion

The focus and difficulty of video target tracking lies in how to quickly and accurately match the target.Under complex background,a single model tracking algorithm can not accurately track the target,so we propose a model based on the fusion of target tracking algorithm.The algorithm integrates NCC and MeanShift to process the video,then uses Kalman filter to predict and estimate the result,and finally uses the minimum total mean square error criterion for adaptive fusion.Experimental results show that the algorithm can achieve better conditions of complex scenes target tracking and improve tracking robustness.