Anomaly Detection in Traffic Patterns Using the INDOT Camera System

The Transportation and Autonomous Systems Institute (TASI) of Purdue University Indianapolis (PUI) and the INDOT Traffic Management Center worked together to develop a system that monitors traffic conditions using INDOT CCTV video feeds. Computer vision-based traffic anomaly detection has been studied for the past 20 years, and a thorough state-of-the-art analysis was produced in a recent survey paper. Although AI has contributed to improving anomaly detection, several major challenges remain, such as tracking errors, illumination, weather, occlusion handling, camera pose, and perspective. In addition, the lack of real-life datasets makes the effectiveness of anomaly detection techniques unclear. This project builds on previous research by using automatic anomaly detection and AI algorithms to identify anomalous behavior of the short- and long-term variations of traffic patterns. The research team designed the new system, including the hardware and software components; the existing INDOT CCTV system; the database structure for traffic data extracted from the videos; and a user-friendly web-based server for showing the anomalies automatically.