This report details our system for vehicle detection and tracking in multi-spectral UAV video data. The work builds upon the tracking framework previously developed, using the spatiogram bank tracker as a core component of the overall system. Camera motion is compensated by automatically performing image alignment and computing the homographic warping that describes the movement of the ground-plane in relation to the camera. By aligning subsequent frames, moving vehicles can be detected using image differencing, adaptive thresholding and selection of appropriately sized regions. Banks of spatiogram trackers are then used to perform object tracking and to fuse data from both the visible spectrum and thermal infrared video. Results are shown indicating the success of this approach on UAV data from the DTO VACE dataset collection. Finally, we note the challenges posed by this difficult application and propose future directions for this work in tackling these issues.
Vehicle Tracking in UAV Video Using Multi-Spectral Spatiogram Models
2007
10 pages
Report
Keine Angabe
Englisch
Photographic Techniques & Equipment , Infrared & Ultraviolet Detection , Image processing , Multispectral , Video frames , Data fusion , Drones , Infrared spectra , Ireland , Surveillance , Tracking , Video images , Visible spectra , Frame alignment , Object tracking , Spatiograms , Vehicle detection , Vehicle tracking , Video segmentation , Aerial surveillance , Foreign reports
Multi-Vehicle Tracking Using Microscopic Traffic Models
IEEE | 2019
|Automatic multi-vehicle tracking using video cameras: An improved CAMShift approach
Online Contents | 2013
|Multi-spectral and multi-perspective video arrays for driver body tracking and activity analysis
British Library Online Contents | 2007
|Commercial Vehicle Route Tracking using Video Detection
NTIS | 2010
|Automatic multi-vehicle tracking using video cameras: An improved CAMShift approach
Springer Verlag | 2013
|