Real-Time 6-DOF Monocular Visual SLAM in Large-Scale Environments

Hyon Lim (Seoul National University)

COMPUTER VISION AND ROBOTICS SERIES

DATE: 2013-06-27
TIME: 16:00:00 - 17:00:00
LOCATION: RSISE Seminar Room, ground floor, building 115, cnr. North and Daley Roads, ANU
CONTACT: JavaScript must be enabled to display this email address.

ABSTRACT:
We present a real-time approach for monocular visual simultaneous localization and mapping within large-scale environment. From a monocular video sequence, our method continuously computes the current 6-DOF camera pose, keyframe poses and 3D landmark locations. One of our main contributions lies in utilizing inexpensive binary feature descriptors from the fast keypoint tracker to recognize previously visited scenes by using bags of binary visual words. This is the first approach that employs the unified feature descriptor in both 2D feature tracking and loop closure detection in real-time. By doing so, our algorithm can successfully build the consistent maps from challenging outdoor sequences using a monocular camera as the only sensor, while existing approaches with a monocular camera utilize additional information such as the known geometry of the vehicle to obtain similar results. To our knowledge, this is the first monocular visual SLAM approach which demonstrates real-time performance on large-scale sequences like the KITTI dataset without resizing the input images. The effectiveness of the proposed method is demonstrated on various challenging video sequences including the KITTI dataset and indoor video captured on a micro aerial vehicle.

BIO:
Mr Hyon Lim is a Ph.D. student from Seoul National University currently visiting ANU.