摘要
针对深度相机运动中的位姿估计问题,提出了一种无需迭代的估计方法。首先,在二维图像上应用图像特征点提取和描述方法,完成不同视点的初始匹配。其次,选择初始匹配度量距离最小的2个特征点作为种子点。以三维空间中欧式距离与坐标系的建立无关为准则,对初始匹配进行筛选。剔除误匹配点对,进而计算运动位姿参数。最后,采用nyuv2图像数据库进行实验,验证了本文算法的可行性和正确性。实验结果表明:与传统算法相比,该方法计算效率平均提高了8倍以上,特别适用于大型场景中的同步定位和地图构建SLAM(Simultaneous Localization and Mapping)问题。
Aiming at the problem of estimating the motion parameters of depth camera,a new^est imaout iteration is proposed. Firstly, the initial match of different viewpoints is completed according scr iption method of the image feature points. Secondly , the two feature points with smal lest distance from the initial matching are selected as seed points.Based on the in depen dence of Euclid ean distance and coordinate system in 3D space, the initial match is filtered to eliminate the mismatch points.Then,the motion parameters are calculated. Fi-nally ,in order to verify the feasibility and correctness of the proposed algorithm, experiments are carried out in the open dtbase nyuv2. There sults show that the proposed method can improve the computational efficiency by 8 times or more compared with the tradi tional algorithm,and it isespecially suitable for large scale SLAM.
作者
张一铭
吕耀文
徐熙平
吕梦凌
钟文婷
ZHANG Yi-ming;LU Yao-wen;XU Xi-ping;LU Meng-ling;ZHONG Wen-ting(School of Opto-E^lectronic Engineering,Changchun University of Science and Technolog,Changchun 130039,China)
出处
《激光与红外》
CAS
CSCD
北大核心
2018年第5期638-643,共6页
Laser & Infrared
基金
吉林省科技发展计划资助青年基金项目(No.20160520018JH)
吉林省重点科技攻关项目(No.20170204048GX)资助
