Investigation Of Traffic Sign Image Classification For Self Driving Car

Farra Herliena, Md Zin (2022) Investigation Of Traffic Sign Image Classification For Self Driving Car. College of Engineering, Universiti Malaysia Pahang Al-Sultan Abdullah.


Download (4MB) | Preview


The way we live has changed as a result of technological advancements. Artificial Intelligence has had a good impact on all fields and is making our lives easier. The development of a technology-assisted traffic system is a significant step forward in the automotive industry. With the growth of autonomous vehicles, the automotive industry is improving rapidly. Autonomous vehicles are a certain conclusion in the future, and they are intended to be both safe and convenient. One of the most critical issues for autonomous vehicles is traffic sign classification. Half occlusion, color fade by surrounding barriers, variations in shadows, reflections on signboards during the day, and movement blurring different lighting and weather situations are some of the most typical issues that might occur when identifying and detecting traffic signs. In the classification and identification of road signs, the performance of a Convolutional Neural Network (CNN) has outperformed the same of humans. The purpose of this study is to boost the accuracy of this classification in order to minimize the accident and enhance the credibility of self-driving vehicles. Otherwise, the ecology of traffic may be jeopardised. Using image processing and machine vision processing technologies, as well as the use of in-depth learning in target classification, the traffic sign recognition method based on CNN is studied. A traffic sign detection and classification method with high efficiency and high efficiency are proposed. The German Traffic Sign Recognition Benchmark (GTSRB) is employed to test the approach method, and the results reveal that it outperforms state-of-the-art approaches. A CNN model for deep nature is suggested in this paper. The CNN Model for Traffic Signs is consists of four dense layers or layers that are entirely linked. Because it learns characteristics from all of the attributes of the preceding layers, a fully connected layer learns deeply.

Item Type: Undergraduates Project Papers
Additional Information: SV: Assoc. Prof. Ir. Dr. Fahmi Bin Samsuri
Uncontrolled Keywords: Artificial Intelligence, autonomous vehicles
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
T Technology > TK Electrical engineering. Electronics Nuclear engineering
Faculty/Division: College of Engineering
Depositing User: Mr. Nik Ahmad Nasyrun Nik Abd Malik
Date Deposited: 08 Jan 2024 06:54
Last Modified: 08 Jan 2024 06:54
Download Statistic: View Download Statistics

Actions (login required)

View Item View Item