Real-Time Algorithm for Simultaneous Vehicle Detection and Tracking in Aerial View Videos

Automatic detection and tracking of multiple vehicles in airborne videos is still a challenging problem due to camera movement, vehicle occlusion and the need for computational resources. This paper presents a robust and efficient real-time method for automatic detection and tracking of vehicles in airborne videos. The detection process is based on a combination of Top-hat and Bot-hat transformation aided by the morphological operation. Background objects are removed through analyzing feature points motion of the obtained object regions using K-means clustering and KLT tracker. The obtained vehicles features are grouped and clustered into separate vehicles based on their motion properties. Finally, a connecting scheme is presented to determine the connectivity of vehicle cluster with the corresponding cluster in the vehicles trajectories. Experiments conducted on videos representing airborne cameras verify the excellent performance compared to other existing approaches.