Vehicle Categorical Recognition for Traffic Monitoring in Intelligent Transportation Systems

Automatic vehicle detection and recognition play a vital role in intelligent transport systems (ITS). However, study results in this field remain certain limitations in terms of accuracy and processing time. This article proposes a solution to improve the accuracy of vehicle recognition in order to support traffic monitoring on vehicle restricted roads. The proposed solution to vehicle recognition consists of two basic stages: (1) Vehicle detection, (2) vehicle recognition. This study focuses on proposing solutions for improving the accuracy of vehicle recognition (stage 2). The vehicle recognition solution is based on the combination of architectural development in deep neural networks, SVM model, and data augmenting solutions. It aims at achieving a greater accuracy than traditional approaches. The proposed solution is experimented, evaluated, and compared with different approaches to the same set of data. Experimental results have shown that the proposed solution brings a higher accuracy than other approaches. Along with an acceptable processing time, this promising solution is able to be applied in practical systems.