Relative Velocity Model to Locate Traffic Accident with Aerial Cameras and YOLOv4

The ambulance response time for trauma incidents depends on location and traffic conditions. High congestion noticeably increases the ambulance response time, causing delays that decrease the victim's odds for survival. The paper aims to use a UAV to locate the head of a congestion, such that upon locating accident, the UAV can send the exact location of the accident and stream live footage of the scene so paramedics can be prepared, thus reducing response time. This is implemented through a video imaging post-processing via an object detection convolutional neural network, YOLOv4, and a proprietary MATLAB algorithm.