Assessment of Parcel Area Measurement based on VHR SAR Images

In the frame of the Common Agriculture Policy, Member States have to measure parcels claimed for subsidies with a recommended precision. This is usually done using Very High Resolution (VHR) optical images with ground sampling distance of around 1m or better. However acquisition of such imagery may fail due to cloud cover. It is therefore worth examining the potential of almost weather independent VHR radar data for replacing VHR optical imagery: during this study, the identification of agricultural parcels and the assessment of the measurement accuracy on VHR SAR images were tested. Airborne VHR X band SAR data were provided over 4 agricultural test sites in France. Three of these sites were covered with 1m monopolarized (“B&W”) data from 2002-2004 whereas the remaining one was covered with 2m multipolarized (“colour”) data from 2002. Orthophotos (1m B&W and 50 cm colour ADS 40) acquired over 2001-2004 were used as reference. All parcels falling on the frame of the VHR SAR images were digitized on the orthophotos and examined on the VHR SAR data. Two sets of around 40 parcels each were selected on the two types of VHR SAR images (2m “colour” and 1m “B&W”). Each parcel was measured randomly 3 times by 3 operators on both the SAR imagery and the orthophoto. The errors on the parcel area were translated into buffer widths around the parcel perimeter. After the elimination of outlier measurements, the buffer variations were analyzed and a tolerance interval around the buffer estimated. The results indicate that (1) about 30% of the parcels were not visible on the X-band SAR data; (2) the estimated tolerance intervals of the buffer values were of 4.14 m and 4.81 m on 2 m colour composition and 1 m black-and-white SAR data respectively, which is larger than requested by the EU Regulation.

PLUTO-KOSSAKOWSKA Joanna;  GRANDGIRARD David;  KERDILES Herve; 

2007-12-20

Remote Sensing and Photogrammetry Society

JRC37819