Vehicle Identification and Classification for Smart Transportation using Artificial Intelligence- A Review

Abstract

The vehicles are the core elements of smart transportation, and vehicle identification & classification is an essential task in many sub-areas of smart transportation. With the change in technologies, it is time to shift in the implementing methods for smart transportation systems, such as toll management systems and advanced traffic management systems, from traditional methods to artificial intelligence (AI) based methods. This paper presents various AI-based object detectors that detect the objects using recorded or real-time images and are suitable for vehicle identification and classification. This paper suggests use of single-stage object detectors as the preferred choice over two-stage object detectors. For processing images and videos, we have also presented a comparative study between single stage-object detectors You Only Look Once (YOLO) and Single Shot Multi-Box Detector (SSD), including their incremental versions and differences between them. Furthermore, YOLOv3, from the family of single-stage object detectors, has been suggested for the task of identifying & classifying of vehicles for toll management systems and advanced traffic management systems. This paper compares single-stage object detector algorithms YOLO & SSD and suggests that YOLO being faster than SSD and comparable mAP makes YOLO a suitable algorithm for use in the toll management system.