FURI | Summer 2021

Automated Traffic Incident Detection and Driver Behavior Analysis

Nowadays, there are over 260 million registered vehicles in the United States (most of them are passenger vehicles), and as a result, road safety is an important issue for everyone. However, road safety is often studied based on safety-critical events such as accidents. Traditionally, traffic incidents are reported by police and witnesses manually, which is inefficient, imprecise, and usually biased. To address this issue, I propose a system for automated traffic incident detection and fine-grained continuous driver behavior analysis with traffic monitoring videos. It will be able to track and localize vehicles with 3D bounding boxes on the map using a framework developed by ASU APG, and then, it can apply a set of rules to automatically detect interesting events in the traffic and calculate a set of safety metrics continuously at every video frame to analyze the driving behavior.