Vision-and-lidar based real-time outdoor localization for unmanned ground vehicles without GPS - Details

author：

Wu, Fangyi (Wu, Fangyi.) ^[1] | He, Bingwei (He, Bingwei.) ^[2] (Scholars：何炳蔚) | Zhang, Liwei (Zhang, Liwei.) ^[3] (Scholars：张立伟) | Chen, Shuiyou (Chen, Shuiyou.) ^[4] | Zhang, Jianwei (Zhang, Jianwei.) ^[5]

Indexed by：

EI Scopus

Abstract：

This　paper　describes　a　vision-and-lidar　based　outdoor　localization　method　for　Unmanned　Ground　Vehicles　(UGV)　without　GPS.　We　present　a　real-time　method　for　pose　estimation　by　combining　visual　odometry　and　lidar　odometry.　Instead　of　using　a　GPS,　a　laser　scanner　and　a　RGB-D　camera　are　mounted　on　our　UGV.　Visual　odometry　and　lidar　odometry　are　fused　by　Extended　Kalman　Filter　(EKF).　Bundle　Adjustment　(BA)　is　used　to　optimize　the　fused　odometry　and　build　a　3D　map.　The　method　has　been　evaluated　by　four　test　routes　set　in　our　university.　Experimental　results　indicate　that　a　combination　of　visual　and　laser　information　on　pose　estimation　is　better　than　using　only　one　of　them.　In　addition　to　test　robustness　of　the　method,　experiments　are　performed　both　in　daytime　and　at　dusk.　©　2018　IEEE.

Keyword：

Computer vision Extended Kalman filters Ground vehicles Intelligent vehicle highway systems Optical radar Vision

Community：

[ 1 ] [Wu, Fangyi]School of Mechanical Engineering and Automation, Fuzhou University, Fuzhou, China
[ 2 ] [He, Bingwei]School of Mechanical Engineering and Automation, Fuzhou University, Fuzhou, China
[ 3 ] [Zhang, Liwei]School of Mechanical Engineering and Automation, Fuzhou University, Fuzhou, China
[ 4 ] [Chen, Shuiyou]School of Mechanical Engineering and Automation, Fuzhou University, Fuzhou, China
[ 5 ] [Zhang, Jianwei]TAMS, Department of Informatics, University of Hamburg, Germany