Driving Recorder Based On-Road Pedestrian Tracking Using Visual SLAM and Constrained Multiple-Kernel

IEEE Conference on Intelligent Transportation Systems (ITSC), Oct. 8 ~ 11, 2014

Kuan-Hui Lee, Jenq-Neng Hwang, Dept. Electrical Engineering, University of Washington

Greg Okopal, James Pitton, Applied Physics Laboratory, University of Washington

paper: pdf


The proposed system systematically detects the pedestrians from recorded video frames and tracks the pedestrians in the V-SLAM inferred 3-D space via a tracking-by-detection scheme. In order to efficiently associate the detected pedestrian frame-by-frame, we propose a novel tracking framework, combining the Constrained Multiple-Kernel (CMK) tracking and the estimated 3-D (depth) information, to globally optimize the data association between consecutive frames. By taking advantage of the appearance model and 3-D information, the proposed system not only achieves high effectiveness but also well handles occlusion in the tracking. Experimental results show the favorable performance of the proposed system which efficiently tracks on-road pedestrian in a moving camera equipped on a driving vehicle.

ETHMS Seq #1
ETHMS Seq #4

Downtown Seq#2

3-D Visualization

  W = Forwards
  S = Backwards
  A = Strafe Left
  D = Strafe Right
  Page Up = Move Up
  Page Down = Move Down

Camera View:
  Up Arrow = Tilt up
  Down Arrow = Tilt down
  Right Arrow = Turn right
  Left Arrow = Turn left
Note: if demo does not work, please try to refresh the page several times, to help loading the 3-D models. If "need your permission to run GEPlugin", please permit it in your browser (for chrome, type "chrome://plugins" in search bar, and enable it).