光学学报, 2017, 37 (9): 0915005, 网络出版: 2018-09-07
基于协方差矩阵的多特征融合跟踪算法 下载: 874次
Multiple Feature Fusion based on Covariance Matrix for Visual Tracking
图像处理 视觉跟踪 特征融合 协方差矩阵 边缘 局部二值模式 image processing visual tracking feature fusion covariance matrix edge local binary pattern
摘要
为提高视觉目标跟踪算法的稳健性,提出一种基于协方差矩阵的多特征融合跟踪算法。在量子遗传算法框架下,采用区域协方差描述子对颜色、边缘和纹理特征进行融合,并采用一种快速协方差交叉算法进行模型更新。该算法综合利用了区域协方差描述子维数较低,量子遗传算法收敛速度快且全局寻优能力强和快速协方差交叉算法快速计算的特点,能极大地提高了融合、匹配与更新过程的运算效率,实现了快速有效的多特征融合跟踪。实验结果表明,该算法能够有效应对遮挡、旋转、形变和运动模糊等多种复杂变化的干扰,实现对目标的快速稳健跟踪。
Abstract
In order to improve the robustness of visual target tracking algorithm, a multiple feature fusion tracking algorithm is proposed based on covariance matrix. Under the framework of quantum genetic algorithm, the region covariance descriptor is used to fuse color, edge, and texture features. A fast covariance intersection algorithm is adopted to update the model. The proposed algorithm makes the most use of low dimension of the region covariance descriptor, fast convergence and strong global search ability of the quantum genetic algorithm, and fast calculation ability of the fast covariance intersection algorithm, which greatly improves the efficiency of fusing, matching and updating process, and effectively realizes fast and efficient multi-feature fusion tracking. Experimental results show that the proposed algorithm can effectively cope with the interference, such as occlusion, rotation, deformation and motion blur, and achieve fast and robust target tracking.
金泽芬芬, 侯志强, 余旺盛, 王鑫. 基于协方差矩阵的多特征融合跟踪算法[J]. 光学学报, 2017, 37(9): 0915005. Zefenfen Jin, Zhiqiang Hou, Wangsheng Yu, Xin Wang. Multiple Feature Fusion based on Covariance Matrix for Visual Tracking[J]. Acta Optica Sinica, 2017, 37(9): 0915005.