CSS
 
Date : 21-01-11 12:05
   3.IOR_S5_051_4.pdf (865.8K)
Design of a Vision Based Lane Detection Method with an Evaluation Index for Autonomous Vehicles
Kwiwoo Park, Jun Gi Park, Chansik Park*


This paper proposed a lane detection method with the evaluation index. Three constraints for correct lane are defined in order to design reliable evaluation index. The errors with respect to three constraints are mapped to each evaluation index with Gaussian pdf. The total evaluation index is defined as a weighting sum of three evaluation index: vanishing point, lane color, and lane width. Using images from real autonomous vehicle, the efficiency of the designed lane detection system with evaluation index are verified. Experimental results show that the evaluation index becomes less than threshold if determined lane is violated one or more constraints and this result will be used in the integrated navigation system effectively. Therefore, designed lane detection system and evaluation index will contribute to improve navigation performance in autonomous vehicle.

Keywords: lane detection, evaluation index, autonomous vehicle