A review on vision-based vehicle identification using convolutional neural network

Abstract

Vehicle Identification is a paradigm of Intelligent Traffic System (ITS) that is continuously being researched to improve current challenges on the road. As results, Intelligent Traffic Systems provides smarter and safer operational decisions with higher behavioural understanding. One of the important segments that improve identification is the paradigm of computer vision-based identification, which provides informative visual data of vehicles. In this paper, we review the current active body of knowledge on vehicle identification based on computer vision using Deep Neural Network's (DNN) sub-paradigm Convolutional Neural Network (CNN), by exploring different techniques and challenges. In proven in previous experiments, CNN presents a large accuracy and great results in object detection and classification. Therefore, the focus of the paper will be on the types of CNN in implemented in existing literature. Furthermore, a literature critique and analysis performance review of CNN methods for vehicle identification will be conducted. From the critique results, we further discuss future research that will further contribute to the body of knowledge.