ASEE NCS Conference 2019

Full Program »
PDF File
View File
pdf
1.8MB

An algorithm using geometric constraints for detecting and marking roads for autonomous golf cars

Recent years, Driver Assistance System (DAS) and Autonomous Vehicle technology to help improve driver’s safety and experience have become very popular research topics. Lane detection and marking is the first step for extracting road information as the input data for subsequent processes, thus it is a critical component in autonomous vehicle systems. Over years, people have done a great deal of research in this area and have developed different methods and technologies to solve real world problems. Usually, people use straight lines or different fit curves to model roads. However most methods are based on highway or local road environment and they are not suitable for golf courses detection.

In this paper, we propose a new lane detection and marking algorithm with a simplified filter to help identifying multiple-curved roads in golf courses based on the images acquired from a forward-looking camera. This algorithm can detect the whole road surface and get all possible edge points. After processed by the proposed filter, the algorithm can precisely mark the left and right edges of the lane for the golf car to autonomously drive along.

In this algorithm, we use geometric constraints of the road to help build the road model, thus the shape of the road to be detected can be flexible and can have multiple curves. The algorithm provides more information for identifying the direction and shape of the lane at far point. Compare to traditional lane detection methods, this method does not need white and yellow lane marks as the initial input condition. It also does not need too much computer capability for calculation. Furthermore, it is easy to implement this algorithm as a real-time algorithm for golf lane detection and marking. The road model got from this algorithm can be used for extracting further road information for DAS and Autonomous Vehicle.

The experimental results on local golf course show that this algorithm is reliable and has good robustness.

Luodai Yang
Eastern Michigan University
United States

Lijuan Zhang
Eastern Michigan University
United States

Jonathon Lin
Eastern Michigan University
United States

Qin Hu
Eastern Michigan University
United States

 



Powered by OpenConf®
Copyright©2002-2018 Zakon Group LLC