摘要
基于图像的三维重建近年来成为摄影测量和计算机视觉等学科的研究热点,其关键技术是将图像中的特征点进行提取和匹配,本文提出了一种混合式的图像特征检测与匹配算法。混合算法基于传统的Harris角点检测算法和SIFT算法,通过降采样构建图像尺度空间并提取Harris角点,采用综合图像色彩标准化算法(CCIN:Comprehensive Color Image Normalization)预处理图像,计算三色通道下的SIFT描述子,并二值化,最后通过欧式距离作为相似形度量实现两幅图像特征向量的匹配。在MATLAB 2014平台上实现混合算法,测试3组不同图像,结果表明,混合算法提取的特征数量优于Harris算法和SIFT算法,且特征分布均匀,采用了具有色彩属性的描述子,特征匹配正确率达到了90%以上。混合算法能提高特征提取的数量,提高特征匹配正确率,为后期三维重建建立基础。
3 D reconstruction based on image has become a hot topic in photogrammetry and computer vision in recent years. Its key technology is to extract and match the feature points in the image. This paper presents a hybrid image feature detection and matching algorithm. Based on the traditional Harris corner detection algorithm and SIFT algorithm,first build image scale space by over sampling and extract Harris corners,and then preprocess images by CCIN algorithm,second,calculate the SIFT descriptor in three color channel and binary. Finally,the matching of two image feature vectors is realized by the Eulerdistance as the similarity measure. The hybrid algorithm was implemented on MATLAB 2014 to test three groups of different images. The results show that the number of feature extracted by hybrid algorithm is better than Harris algorithm or SIFT algorithm,and the feature distribution is uniform of the descriptor,the successful matching rate reached more than 90%. The hybrid algorithm can increase the number of feature extraction,improve the correctness of feature matching,and establish the foundation for the later3 D reconstruction without significantly increasing the computation time.
作者
缪盾
刘燕萍
Miao Dun;Liu Yanping(Civil Engi. Department, Tongji Zhejiang College, Jiaxing 314051, China)
出处
《工程勘察》
2018年第6期51-54,共4页
Geotechnical Investigation & Surveying
基金
浙江省教育厅科研项目资助(Y201636365)
