Real Time Road Lane Detection using Deep Convolutional Networks

Information Science Engineering

Detection of lanes is an essential module for autonomous vehicles and advanced driver assistance systems (ADAS). Many state of the art methods for lane detection have been suggested in recent years. Although, these techniques focus on identifying the lane from a single frame, and they usually provide arguably dissatisfying performance in dealing with certain extreme situations such as degradation of the lane line, large shadows, significant occlusion of vehicles, noisy inputs of images, etc. Practically, lanes are supposed to be on-road continuous line structures. Hence, a lane that cannot be precisely detected in the live frame can be extrapolated from the information of previous frames. Therefore, we have used multiple frames from a continuous driving scenario to approach lane detection, and for this reason, a hybrid architecture- combination of a convolution neural network (CNN) and a recurrent neural networks (RNN). In an attempt to train the model for optimum robustness, we intend to perform comprehensive experiments on two massive datasets.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.