摘要
传统特征点匹配方法在训练程序完成之后将实时图像与样本图像直接进行匹配,计算开销大,匹配时间长,实时性不强;提出一种基于分类的快速特征点匹配思想,在训练阶段引入视点的压缩和分类机制,改进匹配过程,在减少计算开销的同时保证匹配精度。与直接匹配方法进行比较,进一步说明改进过程对于匹配速度提升的作用。
The traditional point matching methods rely on direct matching between real time images and sample ones after training. It requires computational consumption and long time to match, therefore the methods perform not very well in real-time application. We propose a fast point matching method, which based on the compaction and classification of view points in training, and we develop the matching procedure for lower computational cost and robust matching. Compared with the direct methods, the improvement of matching speed is distinct.
出处
《计算机与数字工程》
2007年第9期123-124,163,共3页
Computer & Digital Engineering
关键词
特征点
快速
匹配
分类
feature points,fast,matching,compaction,classification
