一种基于背景运动特性分析的背景配准算法

为了解决复杂背景下基于背景运动补偿的目标检测算法无法精确提取背景运动特征块的问题,该文提出了一种基于背景运动特性分析的背景配准算法。采用Harrir-Affine焦点算法提取仿射不变特征点,通过Lucas-Kande光流法估算特征区域光流矢量,通过特征区域光流主方向相似度度量对特征区域聚类实现背景特征区域的精确提取。实验结果表明:该算法能够精确提取出背景运动特征块,与传统的基于背景配准的目标检测算法进行比较,采用本文算法的目标检测其有效性提高了1.1倍。

A Novel Background Compensation Algorithm Based on Analysis of Background Moving Character

To resolve the accurate extraction of background moving characters based on background compensation algorithm in complex background, an background compensation algorithm based on analysis of background moving characters was proposed. At First Harrir-Affine algorithm was used in extracting affine invariant interest points. Then optical flow vectors of feature-regions were acquired, which adopted Lucas-Kanade optical flow algorithm Through comparing the similarity values of main orientation in feature-regions...