中文摘要：

针对复杂跟踪环境下,单模态方法不能很好地跟踪目标的问题,提出了一种基于多模态特征联合稀疏表示的目标跟踪方法。该方法对每个候选样本的多模态特征进行联合稀疏表示,将各模态重建误差之和用于计算候选样本的观察概率,并将具有最大观察概率的候选样本确定为目标。通过与其他一些流行跟踪算法进行对比实验,结果表明本方法在遮挡、光照变化等场景下均能可靠跟踪,具有更好的跟踪效果,从而验证了方法的可行性。

英文摘要：

The single feature usually cannot distinguish the target from background well in the complex environment, and thus a multi-cue joint sparse representation based tracking method was proposed. The multi-cue features of each candidate target were represented sparsely and jointly, and the sum of their reconstruction errors was used to compute the observation probability of each candidate. The candidate with maximum observation probability was determined to be the target. Comparative experiments with other state-of-the-art tracking algorithms show that the proposed method can reliably track in various scenarios such as occlusion and illumination variation. It has better tracking performance, which verifies the feasibility of the proposed method.