摘要
针对目前主流SLAM(同时定位与建图)算法在动态环境中存在精度大幅下降的问题,提出了一种基于光流分割去除动态物体干扰的DY-SLAM(SLAM In Dynamic Environment)算法。该算法采用实例分割算法结合相邻帧图像之间的稠密光流对动态物体进行分割,在SLAM系统图像帧间匹配前剔除动态物体特征点,提高动态环境下的定位精度。使用公开数据集对算法进行评估,算法的RMSE提升最大可达21.59%,能够有效提高系统在复杂动态环境下的定位精度及鲁棒性。
At present,the traditional SLAM algorithm has the problem of greatly reducing of accuracy when dealing with complex dynamic environments,therefore,a DY-SLAM algorithm based on optical flow segmentation is proposed.The algorithm combines instance segmentation technique and the dense optical flow between adjacent frames to segment the dynamic objects,eliminate the feature points of the dynamic object before frame matching in the SLAM system,improve the location accuracy in dynamic environment.Algorithms are evaluated in open datasets,the RMSE of the proposed algorithm can be improved up to 21.59%,which can effectively improve the positioning accuracy and robustness of the system in complex dynamic environment.
作者
叶寒雨
李传昌
刘淼
汪子潇
张伟伟
YE Hanyu;LI Chuanchang;LIU Miao;WANG Zixiao;ZHANG Weiwei(School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;School of Vehicle and Mobility,Tsinghua University,Beijing 100084,China)
出处
《农业装备与车辆工程》
2023年第3期90-94,100,共6页
Agricultural Equipment & Vehicle Engineering
关键词
DY-SLAM算法
光流估计
实例分割
动态环境
DY-SLAM algorithm
optical flow estimation
instance segmentation
dynamic environments
