摘要
针对特征向量匹配计算量较大的问题,提出了一种改进的基于区域相关约束的快速鲁棒局部特征(SURF,Speeded-Up Robust Feature)的视频帧间的特征匹配算法。相比于最近邻与次近邻之比,增加随机抽样一致性估计来去除误匹配,再结合连续帧间的像素相关性,进一步降低误匹配和加速匹配过程。在PETS数据库的仿真结果表明,该算法能够在凌乱和存在遮挡的背景下完成目标识别,去除误匹配更加有效,适用于对实时性要求较高的场合。
In consideration of complex calculation in feature matching, an improved feature matching algorithm for between video frames based on local SURF（Speeded-Up Robust Feature） features in correlation region is presented. Random sample consensus algorithm outperforms NN/SN（the ratio of first to second closest distance） in getting rid of false matches. Moreover, the pixel distance correlation between successive frames is taken into account, thus to reduce the false match and speed up the matching procedure. Simulations on PETS database show that this algorithm is more effective and could achieve real-time performance in robustly identifying objects among clutter and occlusion.
出处
《通信技术》
2012年第2期135-137,共3页
Communications Technology
基金
国家自然科学基金资助项目(批准号:61179045)
关键词
图像处理
局部特征
特征匹配
积分图
image processing
local features
feature matching
integral image
