Traffic Sign Detection and Recognition for Driverless Vehicles using Convolutional Neural Network in Deep Learning

Abstract

Traffic Signs are essential for assisting drivers in the world to ensure smooth flow of traffic avoiding accidents. Road Symbols are the symbolic representations for communicating the drivers. Traffic signs convey a variety of messages about the road and what you should anticipate as a driver. They control traffic by guiding them to their destinations and informing them in advance of entry, exit, and turn spots. Drivers sometimes overlook traffic signs on the road in order to concentrate on driving, or due to bad weather conditions (e.g., fog, rain, etc.), which can be hazardous for both drivers and pedestrians. These days we are seeing many advancements in automobile technologies which is leading to replacement of human labour. Similarly, in this project we are trying to design a system which helps in detecting traffic signs on roads by ML (Machine Learning) algorithms replacing the drivers in vehicles which can be called as autonomous or driverless vehicles. These driverless vehicles can detect and recognise traffic signs and follow them respectively. Our software system would assist in identifying and detecting traffic signs without causing drivers to lose concentration while travelling. We create a CNN (Convolution Neural Network) model to classify images into their corresponding divisions. For image categorization, CNN is the most effective algorithm and is implemented using TensorFlow. The German Traffic Sign Recognition Benchmark (GTSRB), which consists of about 50,000 images captured by camera, was used to train our suggested CNN model. This system focuses to develop a CNN model that can successfully detect and recognize traffic signs in real-time using OpenCV. The results using our suggested system demonstrate 95 to 100% accuracy.