DEPTH PERCEPTION OF A STATIONARY OBJECT BY A MOVING CAMERA

PDF

Authors
  1. Robin, C.M.
  2. Aitken, V.C.
Corporate Authors
Defence Research Establishment Suffield, Ralston ALTA (CAN)
Abstract
Many methods have been proposed to determine the depth of an object from visual imagery. The report focuses on the case where a single camera is mounted on a vehicle which is in motion relative to a stationary object of interest. The configuration is examined in an effort to use an infrared camera mounted on an unmanned vehicle to accurately locate landmines. The exact motion of the camera is not known precisely, but is also estimated from noisy position and orientation measurements. One possible solution that is examined is to use a falt earth approximation to determine the location of the object from a single image. The method is compared to a Kalman filtering method in which the incoming images are recursively used to predict the object's location as well as the motion and location of the camera. The differences between using the quick flat earth approximation and the more involved Kalman filter are examined, especially in the case where the vehicle is going u a hill. The error in the Kalman filter is compared to the theoretical Cramer Rao lower bounds in order to determine the effectiveness of the filter. The data used to test the Kalman filter and that the flat earth approximation is simulated for various vheicle manoeuvres. TRUNCATED
Keywords
Monocular Motion Stereo Vision;Improved Landmine Detection System (ILDS);Unmanned vehicles
Report Number
DRES-M-1469 — Memorandum Report
Date of publication
01 Aug 1995
Number of Pages
51
DSTKIM No
96-00428
CANDIS No
154694
Format(s):
Hardcopy;Hardcopy;Document Image stored on Optical Disk

Permanent link

Document 1 of 1

Date modified: