Real-time highway traffic information extraction based on airborne video

Real-time image processing is a difficult work for traffic video monitoring. This paper proposed a method to detect and track vehicles on highway based on airship video and therefore calculate traffic parameters in real-time. A blocking road extraction was performed to determine the ROI, and automatically calculate the tilt of the road which contributes to vehicles detection. A lane marks registration was applied to help vehicles detection and calculate geometric size and displacement of vehicles. Vehicles were tracked by an amended mean-shift method whose parameters were RGB color difference instead of color components. Detection was implemented only in key frames and tracking was applied in other frames to reduce the cost of calculation, which made it possible for real time traffic information extraction.