Lane Detection for a Vehicle Using Computer Vision

Devashish Porwal | Dhruv Patel ^[4] | Hussain Doriwala | Alex Victor

Abstract: Lane detection is a key feature of autonomous and intelligent driving systems. It is one which helps position the vehicle and gives it an idea regarding its movement in real time. A robust lane detection system is extremely necessary given how sensitive environment conditions such as sudden changes in illumination or absence of lane markings etc. on the road can lead to failure of the software. Furthermore, advancements in lane detection systems can help diminish long standing problems such as driver fatigue, distractions during driving, unsafe lane changing etc. and make road traveling by vehicles considerably safer with the intervention of computers onboard vehicles. Lane detection is a key hurdle which is delaying the progress from semi-autonomous vehicles to completely autonomous vehicles since flawed detection by software brings up severe safety concerns. To increase processing speed the software uses Canny Edge detection, selective oriented Gaussian filters and Hough Transform along with Matplotlib to process and accurately analyze the captured real time images of the top view of the road. Apart from images we have further made use of moving pictures (videos) as input to analyze the working and accuracy of the software in order to test its robustness for use in the real world, which is our final goal.

Keywords: Autonomous Vehicle, Deep Learning, Canny Edge Detection, Hough Transform, Computer Vision

Edition: Volume 11 Issue 12, December 2022,

Pages: 57 - 63