Towards AI-Based Traffic Counting System with Edge Computing

The recent years have witnessed a considerable rise in the number of vehicles, which has placed transportation infrastructure and traffic control under tremendous pressure. Yielding timely accurate flow information is essential development strategies. Despite continual advances wealth literature available intelligent system (ITS), there lack practical counting system, readily deployable on edge devices. In this study, we introduce low-cost effective edge-based integrating object detection models to perform vehicle detecting, tracking, counting. First, dataset (VDD) representing conditions Vietnam was created. Several deep learning for VDD were then examined two different device types. Using detection, presented lightweight method seamlessly combining with traditional tracking increase accuracy. Finally, obtained based counted categories their directions. experiment results clearly indicate that proposed achieves top inference speed at around 26.8 frames per second (FPS) 92.1% accuracy VDD. This proves our proposal capable producing high-accuracy can be applicable ITS order reduce labor-intensive tasks management.