摘要
针对ORB特征匹配算法特征提取稳定性差,匹配精度低的问题,提出一种改进的ORB图像匹配方法。通过小波变换对图像预处理,利用自适应频谱抑制方法提取候选特征点集,再建立多尺度空间模型,筛选具有尺度不变性的精确特征点,并通过BRIEF算子生成特征描述子,最后使用Hamming距离进行特征匹配,完成配准。通过3组图像匹配实验对改进算法的有效性进行验证,实验结果表明,改进算法在兼顾ORB实时性的同时,提高了匹配精度,改善了图像尺度变化较大时特征提取的稳定性。
An improved ORB(oriented FAST and rotated BRIEF)image matching method is proposed to deal with the fact of poor feature extraction stability and low matching precision of the ORB algorithm.The image is preprocessed by wavelet transform,and the candidate feature point sets are extracted by adaptive spectrum suppression method.Then the multi⁃scale spatial model is created to select the precise feature points with scale invariance and the feature descriptors are generated by BRIEF operator.The Hamming distance is used for feature matching to complete registration.The effectiveness of the improved algorithm is verified by three groups of image matching experiments,which show that the improved algorithm improves the matching accuracy and perfects the feature extraction stability when the image scale changes greatly while taking the real⁃time property of ORB into account.
作者
张磊
郑子健
张殿明
张泽仲
ZHANG Lei;ZHENG Zijian;ZHANG Dianming;ZHANG Zezhong(Hebei University of Technology,Tianjin 300132,China;China Automotive Technology&Research Center Co.,Ltd.,Tianjin 300300,China)
出处
《现代电子技术》
北大核心
2020年第3期27-30,35,共5页
Modern Electronics Technique
基金
国家重点研发计划(2017YFB0102500)
关键词
ORB算法
特征点提取
图像匹配
自适应频谱抑制
多尺度空间模型
特征匹配
ORB algorithm
feature point extraction
image matching
adaptive spectrum suppression
multi⁃scale space model
feature matching
作者简介
张磊(1993-),男,硕士,主要从事汽车辅助驾驶系统研究工作;郑子健(1990-),男,工程师,主要从事汽车电子电气架构研究工作;张殿明(1979-),男,高级工程师,主要从事汽车电子电气架构研究工作;张泽仲(1993-),男,硕士,主要从事汽车电子电气架构研究工作。