A novel edge detection and pattern recognition algorithm based on beamlet theory for a vision-based wheeled mobile robot

Abstract

Number of strategies have been deployed to make Wheeled Mobile Robots (WMRs) independent or with limited dependence on human assistance/intervention. In order to enhance their mobility, improvements have been made in their pattern recognition and path planning strategies. Vision-based approach is preferred over stereopsis-based, laser-range finders and GPS-based approaches on account of higher resolution and longer ranges. Vision-Based Wheeled Mobile Robot (VB-WMR) uses an active vision system where the observer is in motion. The goal of this project is to provide a novel algorithm aimed to detect and recognize direction indicators on sign boards placed on the course of the WMR. Here visual processing is devoted to noting the salient components like the edges of patterns on sign boards rather than processing the entire image. The proposed algorithm is based on the beamlet theory, which retains directional information besides the location and scale and performs well under noisy conditions. The developed algorithm in MATLAB is later to be incorporated with the VB-WMR.