摘要
特征点描述符在特征提取、图像识别与定位中有重要作用.针对SIFT等梯度方向描述符计算量大,ORB等二进制描述符匹配镜像图像入围率低,提出一种圆周二进制描述符(CBD)的图像点特征提取方法.首先以1:1.2的比例建立图像金字塔,对每层图像进行高斯平滑,使用FAST检测特征点;然后提出二值图像重心法计算特征点的方向,以提升计算特征点方向的速率;最后提出CBD图像点特征提取算法,有效地解决了镜像图像匹配的问题.实验结果证明,CBD具有良好的镜像不变性,且适应性强、入围率高.
Feature point descriptors play an important role in feature extracting,image recognition and detection.For the large computation of gradient direction histogram descriptors such as SIFT etc.and the binarydescriptors’defect that could not match the mirrored image very well such as ORB etc.,an image pointfeature extraction algorithm of circumferential binary descriptor(CBD)is proposed.First,set up the imagepyramid by the1:1.2ratio,Gaussian blur each image layer,detect the feature points with FAST detector.Then,the binary image’s centroid algorithm is given to compute the orientation of the feature points.Withthis method the computation speed is improved.Last,CBD is proposed to solve the problem that the otherdescriptors couldn’t match mirrored image effectively.The experimental results show that CBD has a betterinvariance on mirror image,a better adaptability and a higher inlier point percentage.
作者
张展
杨东升
Zhang Zhan;Yang Dongsheng(Intelligent Control and Equipment Laboratory, Shenyang Institute of Computing Technology, Chinese Academy of Sciences, Shenyang 110168;University of Chinese Academy of Sciences, Beijing 100049)
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2017年第8期1465-1476,共12页
Journal of Computer-Aided Design & Computer Graphics
基金
国家科技重大专项(2013ZX04007031)
关键词
图像识别
圆周二进制描述符
图像重心
镜像不变性
特征匹配
image recognition
circumferential binary descriptor
image centroid
mirror invariance
feature match
作者简介
张展(1988—),男,博士研究生,主要研究方向为图形图像处理、模式识别(1184373364@qq.com);杨东升(1965—),男,硕士,研究员,主要研究方向为数控技术、机器人视觉
