Edge Intelligence for Traffic Flow Detection: A Deep Learning Approach

Intelligent Transportation Systems (ITS) are crucial for regulating public transportation, enhancing security, and addressing urban challenges. ITS offers smart guidance to mitigate traffic congestion, minimize environmental impact, and improve overall transport efficiency. In cloud-based ITS, traffic flow detection involves sending video data from edge devices to cloud data centers. However, this approach faces hurdles due to increased monitoring demands. To overcome this, a solution is proposed: employing deep learning on edge nodes. The approach includes a YOLOv4-based vehicle detection model, retrained for multi-object tracking using DeepSORT. Edge computing migrates and installs the detection and tracking systems, ensuring accurate performance at the edge. This approach optimizes traffic flow detection, addressing challenges faced by conventional cloud-based systems.