基于有向面积的角点分类算法

曾接贤; 黄华川; 张桂梅

doi:10.11834/jig.20080621

本期目录 | 浏览量 : 0 下载量: 228 CSCD: 0

PDF
导出
分享
收藏
专辑

基于有向面积的角点分类算法
A Corner Classification Algorithm Based on Directed Areas
2008年13卷第6期页码：1159-1165
纸质出版：2008
DOI： 10.11834/jig.20080621
稿件说明：

移动端阅览

曾接贤, 黄华川, 张桂梅. 基于有向面积的角点分类算法[J]. 中国图象图形学报, 2008,13(6):1159-1165. DOI： 10.11834/jig.20080621.

A Corner Classification Algorithm Based on Directed Areas[J]. Journal of Image and Graphics, 2008, 13(6): 1159-1165. DOI： 10.11834/jig.20080621.

摘要

在3维场景重构、运动估计、机器视觉等领域，不仅希望能检测出图像中的角点，而且还希望获得角点附近更多的信息，即能对角点进行进一步分类。为了能对图像中检测出的角点进行分类，提出了一种基于有向面积的角点分类方法，该方法首先采用基于协方差矩阵特征向量的小波变换角点检测原理检测出图像边缘上的角点；然后根据角点两侧的边缘信息定义了6种类型的角点；最后通过计算角点附近边缘上顺序排列的3个有向面积，实现对角点的分类。实验表明，基于有向面积的角点分类，具有较高的准确性。

Abstract

In the domain of 3D scene reconstruction

motion estimate and machine vision

we need to detect the corners in a image

as well as obtain more information around corners

namely we could classify the corner further. We proposed a corner classification method based on directed area. First

we applied corner detection principle of wavelet transformation

based on the eigenvectors of covariance matrices

to detect the corner on the edge of images. Secondly

we defined six types of corners based on the edge information of both sides. Finally

corners can be classified by computing three directed areas arranged in order on the edge of the corners. The experiment indicated that the method of classifying corners has higher accuracy.