基于PROSAC算法与ORB-SLAM2的RGB-D相机室内视觉定位研究  

Research on Indoor Visual Localization of RGB-D Camera Based on PROSAC Algorithm and OR-SLAM2

作　　者：曹蜜 胡凌燕[1] 熊彭文 彭杰[1] 曾雷 CAO Mi;HU Lingyan;XIONG Pengwen;PENG Jie;ZENG Lei(School of Information Engineering,Nanchang University,Nanchang 330031,China)

机构地区：[1]南昌大学信息工程学院

出　　处：《传感技术学报》2019年第11期1706-1712,共7页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(61563035,61662044,61663027);江西省杰出青年人才项目(20171BCB23008);江西省研究生创新专项资金项目(YC2018-S072)

摘　　要：针对Random Sample Consensus(RANSAC)在匹配误差剔除上具有处理的盲目性而导致算法效率相对较低的问题,提出一种新的用Progressive Sample Consensus(PROSAC)取代ORB-SLAM2中的RANSAC算法的室内视觉定位方法。与传统的ORB-SLAM2方法不同,本文采用PROSAC算法根据特征点的匹配质量进行高低排序,选取质量较高的匹配点对用于求取单应性矩阵H,以此完成对异常点的剔除,在图像匹配过程中大大减少迭代次数。再结合ORB-SLAM2,进行关键帧跟踪,实时建图,回环检测这三个线程,得到准确的定位。图像误差剔除匹配实验结果表明,PROSAC算法可以明显提高运算效率,相对于RANSAC算法效率提高一倍。将该算法结合ORB-SLAM2进行定位实验结果表明,该算法能够获在不降低定位精度的情况下,明显提高算法效率,以保证实际定位过程的实时性和流畅性。Because of blindness in processing matching error rejection of random sample consensus(RANSAC)leading to low efficiency,a new indoor visual localization method using PROSAC to replace the RANSAC algorithm in ORB-SLAM2 is proposed to improve the problem.To eliminate outliers and reduce the number of iterations in the process of image matching,PROSAC algorithm ranks matching points according to their matching quality,and then chooses matching points with higher quality to obtain homography matrix H.ORB-SLAM2 combined with PROSAC perform key frame tracking,mapping,and loop detection.The experimental results of image error rejection matching show that PROSAC algorithm can significantly improve the processing efficiency,which is twice as efficient as RANSAC algorithm.The experimental results of ORB-SLAM2 localization shows that the algorithm can improve the efficiency of the algorithm without reducing the accuracy of localization,so as to ensure the real-time and fluency of the actual localization process.

关 键 词：同步定位与建图 视觉定位 RANSAC PROSAC 特征匹配 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]