基于改进李群结构的特征协方差目标跟踪

Target tracking with feature covariance based on an improved Lie Group structure

最近的研究发现,在对称正定流形上可构造一种改进的李群结构,并赋予具有双不变度量性质的对数-欧几里得黎曼度量,所得到的距离公式和黎曼均值均呈现简单形式。据此,利用目标的综合特征构建区域协方差阵为目标建模,提出一种基于改进李群结构的特征协方差目标跟踪方法。实验表明,这种跟踪方法实用有效,在相同的条件下,因为算法的计算量的减少,跟踪性能略优于基于仿射黎曼度量的协方差目标跟踪。

Recent research shows that an improved Lie group structure can be constructed on the symmetric positive manifold. This will lead to a bi-nvariant log-Euclidean metric, which makes the distance formula and Riemannian mean take a much simpler form. We model the tracked object with its covariance feature of the interest region and propose a feature covariance tracking method based on the improved Lie group structure. Experimental results show that this method is practical and efficient. Under the same tracking condition, its performance is slightly superior to that of the method based on widely used affine invariant Riemannian metric