基于黎曼流型度量的人工鱼群算法视觉跟踪

Visual Tracking of Artificial Fish Swarm Algorithm Based on Riemannian Manifold Metric

针对经典的基于协方差算子的跟踪方法不能适应目标的遮挡及其全局搜索造成的过多计算消耗问题,提出了一种在黎曼流型度量上的人工鱼群算法的视觉跟踪方法。该方法将融合了目标的位置、颜色、梯度等特征区域的协方差算子作为目标的表观模型,以提高它对姿态变化以及亮度变化的适应性。利用人工鱼群算法搜寻目标与候选目标之间最优的匹配,其并行运算机制提高了跟踪算法的效率,其全局搜索的能力则提高了算法对遮挡问题的鲁棒性。实验结果表明,该算法在复杂背景情况下具有目标跟踪的鲁棒性。

A novel visual tracking method based on artificial fish swarm algorithm on Riemannian manifold metric was proposed.The new algorithm can well deal with the interactive occlusion,and consume less computation load comparing with global exhaustive search,both of which are the limits of classical covariance descriptor tracker.The paper used covariance descriptor combining with object information of position,color,and gradient to enhance the adaptability to change of gesture and illumination changing.The artificial fish swarm algorithm was utilized to find the best matching between object and candidate.Its parallel operation and global search ability improves the effectiveness of processing and can be more robust to occlusion.The experimental results show that the proposed method is more robust for visual tracking under complex scene.