Moving shadow detection and removal for video-based traffic monitoring

In this paper, a shadow detection and removal framework for a video-based traffic monitoring system is proposed. The proposed framework is able to correctly detect if a video sequence has cast shadows that needs to be removed. Shadow detection results in 90.39% shadow detection rate and 88.71% shadow discrimination rate for different daytime traffic scenes with varying shadow strength. A traffic monitoring system using the proposed shadow removal framework shows improvement on traffic volume and road occupancy parameter extraction accuracy.