Automatic Stabilization of Aerial Traffic Images
Peter Knoppers, Hans van Lint, Serge Paul Hoogendoorn · Transportation Research Board 91st Annual MeetingTransportation Research Board · 2012
For analysis of car traffic a camera taking images was mounted underneath a helicopter and flown over the studied road area. From these images, vehicles are to be detected and tracked to allow microscopic traffic flow analysis. In analyzing such image sequences, a crucial step is the image stabilization, entailing removing effects of vibration and movement of the helicopter between successive images. This contribution describes a new system that was developed to remove the impact of vibration and movement of the helicopter from the images. This system stabilizes the images without user support (except initial setup). The implemented algorithm searches for the maximum value of the normalized cross correlation of a reference image and the image that is to be stabilized by varying six image transformation parameters. The algorithm is robust, user-friendly and efficient: on late 2007 personal computer hardware (quad core), the software takes about 1 minute of CPU time per 2448 x 2050 color image. A test case shows the application of the method. Keywords: image stabilization, image registration, normalized cross correlation, remote sensing, road traffic, helicopter, UAV