Lane Detection and Traffic Sign Recognition from Continuous Driving Scenes using Deep Neural Networks
R Kavya, K Md Zakir Hussain, Naidu Sree Nayana, Sanjana S Savanur, M S Arpitha, R. Srikantaswamy · 2021 2nd International Conference on Smart Electronics and Communication (ICOSEC) · 2021
Increasing analysis on automatic driving technology predicts that by upcoming years it will outnumber the manual vehicles. Ensuring safety and reducing road accidents is the primary concern in automated vehicles. Among many challenging and complex tasks to build an automated vehicle, Lane detection and Traffic sign recognition are very crucial. Though these are challenging tasks due to various road conditions, environmental or weather conditions that drivers can encounter while driving, researches have shown that advancement in deep learning and computer vision has gone beyond imagination. This paper proposes and analyses deep learning techniques compared with the existing different approaches to implement an efficient Lane detection and Traffic sign recognition model that acts as a driver support system to operate in real time. It is a significant module for unmanned vehicles and Advanced Driver Assistance Systems (ADAS). This system is aimed to operate in a real time environment for enhanced safety by faster acquisition of traffic signs and detecting the lanes to assist the driver in controlling and performing scrutinized operations. For this purpose, Convolutional Neural Network (CNN) based on SegNet architecture for lane detection and YOLOv4 (You Only Look Once), a clever CNN algorithm meant for object detection in real time can be used to detect and segment the sign boards along with recognising them.