一种改进的IVT目标跟踪算法

An Improved IVT Algorithm for Object Tracking

针对增量视觉跟踪(IVT)算法无法对受遮挡目标进行有效跟踪的问题,提出了一种改进的IVT目标跟踪算法。该算法对IVT算法中目标外观模型表示单一的问题进行了改进,对目标外观采用混合表示方法。若目标未被遮挡则使用增量主成分分析与高斯观测噪声进行表示,反之则使用连续均匀概率分布进行表示,对混合模型进行能量最小化求解来实现对目标的跟踪。实验结果表明,该算法在跟踪过程中具有较好的抗遮挡性能,同时能够实现对目标的实时跟踪。

Aiming at the problem that the occlusion interference of unable tracking object effectively in incremental visual tracking （IVT） algorithm, an improved IVT target tracking algorithm is proposed. The problem of a single target appearance model in the IVT algorithm is solved and a hybrid representation method is adopted to represent the target appearance. If the target is not blocked, using the incremental principal component analysis and Gaussian observing noise to represent, otherwise using the continuous uniform probability distribution to represent. The energy minimization method is implemented for mixed model target tracking. Experimental results show that the proposed algorithm has the better ability of anti-occlusion interference during object tracking and it can realize real-time tracking of targets at the same time.