摘要
景物的特征点抽取是模式识别及计算机视觉中的一个重要问题,已出现的多种检测特征点的方法中主要有角检测法和多边形逼近法。在这两种方法基础之上,人们又提出了结合两种方法的综合方法。提出了一种新的综合方法,首先应用一个简单的角检测方法,然后利用前面计算曲率时的一些值在检测到的角点之间加入一些特征点。实验结果表明新方法比传统方法执行速度更快,并且克服了传统方法的缺陷。
Detecting dominant points is an important subject in pattern recognition and computer vision. Corner detection and polygonal approximation are two major approaches for dominant point detection. Based on the two approaches, combined methods have been proposed. In this paper, a new combined method is presented to detect the dominant points. The new method first applies a simple corner detection procedure, then adds some points between detected corner points with the value of curvature computation. Experimental results show that the new combined method is more efficient than the conventional methods. And it overcomes some defects in the conventional methods.
出处
《计算机应用研究》
CSCD
北大核心
2006年第6期148-152,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(60473108)
关键词
模式识别
曲率
特征点
角检测
多边形逼近
Pattern Recognition
Curvature
Dominant Points
Corner Detection
Polygonal Approximation