摘要
提出了一种基于压缩感知的SIFT准稠密立体匹配算法.该算法利用压缩感知理论的稀疏投影将SIFT的高维特征描述子投影到低维空间上,选取RANSAC算法对匹配的结果进行去伪匹配处理,并将提取的匹配点作为种子点沿着极线方向生长,获取稠密的视差图.该算法利用压缩感知的稀疏投影,大大减小了特征匹配的运算量,同时利用种子生长使视差图变稠密.实验结果表明:与未加入压缩感知的种子扩散立体匹配算法相比,这种算法计算速度更快,误匹配的百分比也较低,是一种快速有效的立体匹配算法.
This paper compressive sensing. proposes a fast SIFT quasi-dense stereo matching algorithm based on By the sparse projection of compressive sensing theory, the high- dimensional feature descriptors of SIFT are projected into the low-dimensional space. The RANSAC algorithm is used to remove the false match. The extracted matching points are taken as seed points and grow along the polar line direction to obtain the dense disparity maps. This algorithm uses the sparse projection of compressive sensing to greatly reduce the amount of feature matching calculations. It also uses seed growth to make the disparity map denser. Experimental results show that the algorithm proposed in this paper is faster and the percentage of mis-matched is also lower than the seeded stereo matching algorithm without compressive sensing. It is a fast and effective stereo matching algorithm.
出处
《浙江工业大学学报》
CAS
北大核心
2017年第3期310-314,共5页
Journal of Zhejiang University of Technology
基金
国家自然科学基金资助项目(61473262)