改进的自适应SUSAN角点特征提取方法  

An improved self-adapting corner feature extraction method

作　　者：陈洪[1] 李进强[2] 

机构地区：[1]福州市勘测院,福州350108 [2]闽江学院,福州350108

出　　处：《测绘科学》2017年第12期118-121,126,共5页Science of Surveying and Mapping

摘　　要：针对目前大多数的角点特征提取算法存在算法结构过于复杂、运行效率偏低及可推广性偏差等方面的局限性,该文通过改进SUSAN算法中灰度差阈值的获取方法,提出一种自适应的角点特征提取方法。该方法首先采用高斯滤波对原始影像做预处理,然后利用Ly算子初步探测概略角点特征集合,最后利用改进的SUSAN角点检测算法从概略角点特征精确确定角点特征。实验结果表明,该方法提高了角点检测的精度,缩短了角点特征提取时间,具有较好的鲁棒性。Aiming at the problem that most of the corner feature extraction algorithm for structure is too complex,is of low operating efficiency and poor generalizability,and so on,SUSAN algorithm was improved primarily by adjusting the computing method of the grayscale difference,then a self-adapting corner feature extraction method was proposed.The proposed method could be described as follows,firstly,use Gaussian filter to smooth the original image;secondly,use the Lv-Yan operator to implement the primary corner extraction;finally,with the improved SUSAN algorithm,this method extracted the corner features precisely.The experimental results showed that the approach improved the precision of the corner extraction,meanwhile,the elapsed time was shorten.The results of performance evaluation also support the high robustness of the approach developed.

关 键 词：角点提取 自适应 SUSAN算法 高斯滤波 

分 类 号：TN911.73[电子电信—通信与信息系统]