Road Lane Line Detection System by Using Cnn and Rnn Algorithms in Deep Learning

Abstract

One of the most fundamental tasks in computer vision for autonomous driving is lane lines detection on road. Lane lines are painted for humans to see and follow while driving. In a very similar way, an autonomous vehicle that uses human designed infrastructure, needs to see the lane markings to steer accordingly and follow the road trajectory. In this project I implemented a computer vision algorithm that processes real data recorded with the front facing camera of a vehicle driving on a California highway. The result is a processed video that highlights the lane lines on the paved road. With the positions of the lane lines identified, the vehicle's offset from the lane's center can be calculated and feed a PD controller to compute the necessary steering angle. While only the lane lines detection is the scope. Then we create a function called draw line (the dram line method’s purpose is to draw lines or lines). The Hough line transform method is used to detect lanes or lanes from the replacement image in polygon images. The detected line or lanes are then drawn. Finally, when the lines are drawn, we blend the previous image.