摘要
由于各种外界因素的影响,角点检测算法检测的角点可能出现位置偏移。在分析该现象的基础上,提出对角点进行重新的定位。首先,通过计算角点的二阶模板值来进行角点模糊定位;其次,引入像素点的概率密度梯度方向,通过比较角点和角点邻域内的像素点的概率密度梯度方向的关系来判断错误角点,并找出所有的候选替代点;最后,根据角点和候选像素点之间的关系,寻找出新的角点。通过实验对比发现该算法能正确地对发生位置偏移的角点进行重新定位。
As a result of the impact of external factors, comers detected by corners detection algorithm may be offset. Based on the analysis of this phenomenon, this paper proposed a new algorithm for corners relocation. Firstly, through calculating the second-order template value for every corner, the fuzzy corner relocation was carried out; secondly, the density gradient direction of pixel was introduced. Comparing the density gradient directions of the comers and the neighborhood pixels of the corners, the wrong comers were determined precisely and all candidates were found; finally, according to the relationship between the corners and the candidate pixels, new corners was found. Experimental results show that the algorithm can accurately relocate the wrong corners.
出处
《计算机应用》
CSCD
北大核心
2010年第2期359-361,366,共4页
journal of Computer Applications
基金
国家863计划项目(2008AA10Z211)
关键词
概率密度梯度方向
模糊定位
角点重定位
density gradient direction
fuzzy location positioning
comers relocation
