摘要
本文运用改进的Harris和SIFT算法实现了图像的快速匹配,旨在提高SIFT算法的实时性。该算法首先利用优化的Harris算法提取图像的角点;然后采用SIFT算法对特征点进行单尺度的特征描述,提取出特征点的128维特征描述符;最后,采用特征点特征向量的欧式距离作为两幅图像的中特征点的相似性判定度量,并用RANSAC剔除错误匹配,实现两幅图像特征点的匹配。实验证明,本文算法在保证精度的同时,大大提高了算法的实时性。
This paper uses the improved Harris and SIFT algorithm to achieve the fast matching of images,aimed at improving the SIFT algorithm real-time.Firstly,The algorithm use the optimization algorithmof harris to extract image corners.Then We use the sift algorithm of single scale to descripfeature pointsand extract the feature points of128d feature descriptor.Finally,this paper use the Euclidean distanceof the feature vectors as the similarity of feature points in two images and use the RANSAC to eliminatemismatch.Experiments show that this algorithm greatly improve the real-time performance of thealgorithm as well as ensures the accuracy.
作者
马领月
潘禹
杜必强
Lingyue Ma;Yu Pan;Biqiang Du(North China Electric Power University. Energy Power and Mechanical Engineering Academy,Baoding,Hebei,071000,China;ChangSha University Of Science &Technology. Energy and Power Engineering Academy,Changsha,Hunan,410000,China)
出处
《电子科学技术》
2017年第4期45-49,共5页
Electronic Science & Technology
基金
中央高校基本科研业务费项目:2014MS118
