A Single Processor Realtime Edge-Line Extraction System for Feature Tracking

Abstract

For robust feature tracking on a mobile robot moving with about 0.5 m/s in an office environment, it is necessary to keep up with a frame rate of at least 10Hz. For this case we introduce an unpublished method for real-time line segment extraction from grey level images by contour tracing. A special emphasis is laid on efficient algorithms for edge finding, contour tracing and symbolizing, to achieve the demanded frame rate for 512x512 8bit grey level images on a single processor system. All algorithms are geared to real time operation, avoiding recursions, enabling to priorize the processing order and aborting in case the deadline has been reached. 1