A Study on the Development of the Driver's Intensive Warning System during Tunnel Driving Based on Real-Time Vehicle Detection and Distance Estimation

In this paper, we propose to detect vehicles driving in a tunnel in real time, estimate the distance between vehicles by comparing the size of the detected auto ROI area, and provide driving safety information to the driver to maintain the safe distance from the omnidirectional vehicle. The proposed method uses the YOLO v3 deep learning model to detect cars in real time in a tunnel environment. The YOLO model has the ability to detect objects in real time. In addition, to estimate the distance from the detected forward-driving vehicles, it is estimated by separating them into three steps (very close, close, normal) distances, not the actual measured distances. The proposed method uses the size information of the detected vehicle ROI region for distance estimation. Therefore, the proposed method presents that a safe driver support system is possible by detecting vehicles in real time in various tunnel environments and providing three-stage distance information. As a result of experimenting with the proposed method, vehicle detection within the tunnel presented approximately 90% of the results and distance estimation accuracy of approximately 97%.