Traffic Signs Recognition and Detection using Deep Convolution Neural Networks for Autonomous Driving

• Published in: 2022 IEEE 11th International Conference on Communication Systems and Network Technologies (CSNT) pp. 207 - 214
• Main Authors: Raut, Roshani, Desarda, Khushi, Gulve, Saylee, Bodas, Abha, Jadhav, Prem
• Format: Conference Proceeding
• Language: English
• Published: IEEE 23.04.2022
• Subjects: Automation; Autonomous automobiles; CNN; Computational modeling; Driver support systems; Feature extraction; Intelligent vehicles; Lane detection; Pedestrian detection; Pothole detection; Real-time processing; Road crack detection; Roads; Safety; Traffic sign recognition
• DOI: 10.1109/CSNT54456.2022.9787584

Automation is a step in the right direction towards increasing road safety. It helps reduce the number of accidents on the roads. It also increases independence by giving individuals more freedom with full automation. Several causes of traffic congestion can also be avoided, potentially reducing fuel consumption and carbon emissions. Therefore, we have integrated four different systems into the study that contribute to the smooth functioning of self-driving cars. The Traffic Sign Recognition System is one of the most essential historical past studies subjects for figuring out self-sufficient automobile riding structures. It is used to recognize various signs and traffic lights on the road. This ensures more traffic safety by complying with traffic rules. The pothole and road crack detection systems help detect road depressions. This allows the vehicle to change course accordingly and prevent damage to the vehicle. The lane detection system allows the vehicle to stay in the lane without deviating, which is an important aspect of autonomous driving. A Pedestrian Detection System helps detect pedestrians walking down the street and prevent accidents and deaths. This paper proposes the implementation of a traffic sign recognition algorithm using a convolutional neural network (CNN). With high detection rates and speedy execution, convolutional neural networks have improved the maximum performance of computer vision tasks. The developed CNN model provided an accuracy of 95%. This model allows us to examine and apprehend visitor signs, which is a totally essential mission for all self-driving cars. The proposed composite system allows self-driving cars to record and manipulate themselves and perform the required functions without human intervention.