Intelligent Technologies for Automated Electronic Systems

Machine Learning For Traffic Sign Recognition

Author(s): U. Lathamaheswari* and J. Jebathagam

Pp: 182-191 (10)

DOI: 10.2174/9789815179514124010017

* (Excluding Mailing and Handling)

Abstract

The recently developed technology in autos makes traffic signal prediction devices obligatory. It teaches users how to drive safely and manoeuvre their vehicles effectively. Due to drivers' various forms of attention, the number of accidents is rising alarmingly nowadays. The danger of distracted driving, which causes accidents, is decreased thanks to this technology, which also assists in identifying and providing information based on data. The notion of machine learning is presented, and the concepts of supervised learning, unsupervised learning, and reinforcement learning are covered under the heading of categorization and serve as the main principle. Linear regression, neural networks, naive Bayes, random forests, support vector machines, clustering, etc. are some types of models that machine learning may give. This study describes how to train a model using machine learning, with the basic principle being to divide the data into training, testing, and validation. The last section of this chapter discusses how to access machine learning methods to improve the quality of a machine learning project. The suggested approach provides an explanation of the combined model of the modern convolutional neural network (CNN) and the classic support vector machine (SVM) for traffic sign identification. Essentially, a CNN model was trained to produce this model. Several CNN model designs, including LeNet, AlexNet, and ResNet-50, may be used here. The subsequent layers of CNN's output may be utilised as features. These characteristics were added to SVM for categorization purposes. 

© 2024 Bentham Science Publishers | Privacy Policy