Road identification system based on CNN

Abstract

In recent years, convolutional neural network (CNN) has been widely used in image recognition, but there is still a big gap in automatic driving and road recognition. Here we have designed a system that maps raw pixels directly from a single front-facing camera to a steering command. The system mainly uses the camera to recognize the auxiliary lines on the road to realize automatic driving, and the accuracy rate has reached 98%. Unfortunately, such data can only be used in the laboratory, and the recognition of complex road conditions needs further development of machine learning.