Vinav a system for VIsion supported NAVigation
Claus S. Andersen, Niels O. S. Kirkeby, Lasse Knudsen, Claus Brøndgaard Madsen · 2005
We describe a highly modular vision svstem enabling a mobile robot to navigate in a lab environment. The system uses a coarse model of the la, motion information from the mobile robot, and input from a single video camera to achieve this goal. The system consists of the following modules: Filtering of the input. Image to remove noise. Edge extraction. Line linking. Prediction of lines from t he model. Matching of extracted and predicted lines. Estimation of robot position based on rnatched lines. The system has been implemented in order to have a testbed facilitating experiments with either improvements to existing modules, or completely new modules. In its present state the vision system is able to determine an accurate estimate of the position of the robot thus correcting the inevitable errors of the motion sensors of the mobile robot.