Road Recognition and Lane Detection using Deep Learning

Abstract

 An autonomous vehicle needs to be familiar with its surroundings. The safety of the
transportation system is greatly enhanced by advanced driving assistance systems (ADASs). Road detection
is one of the steps that a driving car must do. Is it possible for a computer to recognize a road in a single
photograph for this purpose? This question is addressed using the lane detecting techniques. Roads and
lanes are tough for machine learning to differentiate because of training a machine to recognize a road.
Over the past few decades, a number of lane identification technologies have been created and integrated
into various autonomous cars. It is still very difficult to create lane recognition technology that can
effectively identify a road lane in a range of road conditions. This research provides a composite approach
for road detection from image processing using convolutional neural networks by testing 150 photographs
that include a road, jungle, muddy road, and barriers. It will decide if an image contains a road or not. In
this essay, we first establish whether a road exists. The second step is to find a lane on the finished road.
The benefit of the proposed technology is that if there is a road, the automobile can continue to move
forward; otherwise, it will stop.