SUSAN角点检测算法改进  

The improvement of corner detection of SUSAN algorithm

作　　者：席志红[1] 刘利彬[1] 许新利 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2006年第9期29-31,36,共4页Applied Science and Technology

摘　　要：提出了一种改进的角点检测算法.本着好的算法不依赖于人为干涉的思想,在SUSAN算子基础上,通过对图像灰度值和对比度分析,提出灰度阈值t和比较函数C的快速自适应选取.针对SUSAN算法中对某些特殊型角点检测会失败的情况,构造一种针对SUSAN圆形模板的二圆环模板,对一些与边缘点难以区分的角点进行检测.试验表明,改进的算法近一步提高了检测的准确性.An improved corner detection algorithin is put forward. With the idea that a good algorithin does not rely on human＇ s interference, a fast adaptive selection is proposed for gray threshold t and comparison function c on the basis of SUSAN operator through the comparative analysis to the gray scale and contrast of image. For the case in which the corner detection of some special comers will fail in SUSAN algorithm,a double ring template is constructed by virtue of SUSAN circular template, which can detect the corner points distinguished hard from the points on the edges. Tile experiments show that the proposed algorithm further raises the accuracy of detecetion.

关 键 词：SUSAN算法 角点俭测 特征提取 自适应阈值 

分 类 号：TN919.8[电子电信—通信与信息系统]