Entity

Time filter

Source Type

Bells Corners, Canada

Secker J.,Defence R and D Canada DRDC Ottawa | Gong S.,Defence R and D Canada DRDC Ottawa | Parsons G.,Defence R and D Canada DRDC Ottawa | Schlingmeier D.,Defence R and D Canada DRDC Ottawa
Canadian Journal of Remote Sensing | Year: 2013

During the winter of 2009-2010, Defence R&D Canada - Ottawa collected time series data to characterize the Spotlight mode on RADARSAT-2 for change detection in urban vehicle compounds, specifically for small vehicles during wintertime conditions. Corresponding to each image acquisition, three trihedral corner reflectors and five full-sized pickup trucks were deployed with known locations and orientations. This paper describes the collection, processing, and analysis of two time series of RADARSAT-2 Spotlight mode images acquired over the Shirleys Bay Campus (west Ottawa) between December 2009 and April 2010. The first time series consists of six Spotlight A (SLA) 19 mode ascending orbit images acquired on consecutive 24-day repeat cycle passes. The second time series consists of six SLA17 mode descending orbit images, also acquired on consecutive 24-day repeat cycle passes. The 24-day separation of the images ensured consistent geometry and allowed for precise (sub-pixel) co-registration of the individual images using interferometric processing. This paper describes the measurement of radiometric properties, including the radar cross section and the peak-to-clutter ratio, for the known targets and it describes the use of an automatic target detection algorithm and interactive target validation tools to assist with site monitoring. It was found that precise (sub-pixel) orthorectification and co-registration of the imagery is essential for use of a vector mask to constrain the automatic target detection analysis. A single set of detection parameters could be applied to all images (both ascending and descending); tuning of parameters for specific images was not required. Most target vehicles (pick-up trucks) were detected and there were few false alarms. For the 12 images and the selected detection parameters, the average probability of detection was greater than 98%, whereas the average False Alarm Rate was on the order of 10-4 per pixel (i.e., one false detection per 104 pixels). Identification of small vehicles was not possible via visual inspection; neither length nor orientation could be correctly estimated. However, the detection of targets with RADARSAT-2 Spotlight mode data can contribute to site monitoring. © 2013 Government of Canada. Source

Discover hidden collaborations