Real-time lane detection for outdoor autonomous guided vehicles

Abstract

This dissertation purposes a vision-based lane detection and tracking algorithm for outdoor Autonomous Guided Vehicles (AGVs). The basic approach in this project is to work on the ground image of the road surface, in real world space, with perspective effect removed. The advantage of working on the ground image is that we can make valid assumptions of the road geometry such as, the lane markings on two sides of the road are parallel and the lane width is constant.