为解决目标遮挡、旋转、形变等引起的目标跟踪丢失问题,提出了一种多模式融合的跟踪算法。算法综合运用模板匹配、SUSAN 角点匹配和卡尔曼滤波等三种模式:模板匹配跟踪算法应用于背景复杂、目标未发生遮挡的情况,特征点匹配则在目标部分遮挡时进行跟踪,而 Kalman 滤波则用于预测目标位置,减少匹配算法的运算量,并在目标短暂丢失的情况下继续跟踪。实验对多模跟踪算法与模板匹配算法进行了比较,证明该算法可以稳定地进行目标跟踪,在目标被遮挡时也可以有效跟踪目标。
A multimode tracking algorithm is designed in order to solve the problems of target's occlusion ,deformation and rotation .Template matching algorithm ,SUSAN corner matching algorithm and Kalman filter algorithm are integrated . Template matching algorithm is used in complex background where the target is not submerged .When part of the target ap‐pears in the background ,SUSAN corner matching algorithm is selected .The target state in the next frame can be predicted using Kalman filter which can reduce the search area of the target detection and resolve the target losing problem .The exper‐imental results indicate that the algorithm is much better than the Template matching method obviously ,and the target is tracked stably and real‐time .The target also can be tracked efficaciously ,when it is occluded .