Towards Detection of Abnormal Vehicle Behavior Using Traffic Cameras

Throughout the world, many surveillance cameras are being installed every month. For example, there are over 18,000 publicly accessible traffic cameras in 200 cities and metropolitan areas in the United States alone. Live video streams provide real-time big data about behavior happening in the present, such as traffic information. However, until now, extracting intelligence from video content has been mostly manual, i.e. through human observation. The development of smart real-time tools that can detect abnormal vehicle behaviors may alert law enforcement and transportation agencies of possible violators and can potentially avoid traffic accidents. In this study, we address this problem by developing an application for detection of abnormal driving behavior using traffic video streams. Evaluation is performed using real videos from traffic cameras to detect stalled vehicles and possible abnormal vehicle behavior.