摘要
针对经典角点检测算法存在角点检测准确性或抗噪性不佳的问题,根据常见的X型、T型、Y型3类角点的分布特点,提出一种基于图像邻域灰度变化的角点检测改进方法。首先利用图像灰度变化自相关性,初筛角点。然后采用USAN(Univalue Segment Assimilating Nucleus)模板遍历筛选角点集,基于模板内的分布离散度,进行角点二次定位。最后采用非极大值抑制法,精准定位角点。采用模拟几何图像和真实图像进行角点检测,并与Harris算法、SUSAN算法和基于灰度差分与模板的Harris角点检测算法对比。结果表明,本文改进算法的角点检测准确性(ACU)和一致性(CCN)均有明显提升,具有较好的综合检测性能。
Aiming at the problems of corner detection accuracy or anti-noise performance for classical corner detection algorithms, according to the distribution characteristics of X,T and Y corners,an improved corner detection method based on grayscale changes of image neighboring was proposed.Firstly screening the corner by autocorrelation of image grayscale changes.Then, traversing and screening the corner set by USAN(Univalue Segment Assimilating Nucleus) template. Based on the distribution dispersion in the template, the corner points are located twice.Finally the nonmaximum suppression method was used to locate corner accurately.Corner detection was made by the simulated geometric images and real images,and compared with Harris algorithm,SUSAN algorithm and grayscale difference and template based Harris algorithm.The results show that the accuracy and consistency of the improved corner dection method are enhanced obviously,and the method has better comprehensive detection property.
作者
杨佳豪
董静静
袁彤
何雨恒
杨丹
石美红
YANG Jiahao;DONG Jingjing;YUAN Tong;HE Yuheng;YANG Dan;SHI Meihong(School of Computer Science,Xi′an Polytechnic University,Xi′an 710600,China)
出处
《纺织高校基础科学学报》
CAS
2019年第3期337-344,共8页
Basic Sciences Journal of Textile Universities
基金
国家级大学生创新创业训练计划项目(201810709018)
