基于时空约束和稀疏表示分类的目标跟踪算法

针对经典稀疏分类目标跟踪算法中目标模板和目标基的建模及更新方式效率低，跟踪性能不可靠等问题，提出一种新的目标跟踪算法，解释了时空约束原理，目标基、背景基、时序特征池的创建方法以及选择与抛弃两种基更新机制；该算法采用时序循环更新方式解决模板更新问题，结合稀疏表示分类和标准对冲实时计算目标坐标。相比其他几种经典目标跟踪算法，有效提高了在复杂背景下的目标跟踪性能。

Visual object tracking via time-space constraints and sparse representation classification

In order to solve template updating, basis construction and low efficiency problems in visual tracking tasks, a
novel algorithm based on the space-time constrained and sparse representation classification is proposed. The details of
space constraint, time constraint, both target and background basis construction methods, basis choosing, basis abandoning
mechanisms and temporal feature pool construction methods are given. The temporal looping updating method is used to
solve the template updating problem. The sparse representation classification method and normal hedge are combined for
calculating target locations. The proposed tracking algorithm outperforms better than other state-of-art tracking algorithms
in many difficult situations.