Skip to content

Orthorectification Test on Formosat 2 for the Control with Remote Sensing Campaign of the CAP

Direct area-based crop income aid to farmers was introduced in 1992 under the CAP
(Common Agricultural Policy) reform and represent around 35% of the EU’s general
budget (2001, EU 15).

Appropriate administration and control systems were implemented by Member States to ensure that aid is correctly granted. In this frame, Reg. 3508/92 and 3887/92 on the IACS (Integrated Administration and Control System, covering arable and forage aid) specified the possible use of remote sensing techniques as part of the minimum 5% on-the-spot checks to be performed each year.
Since 1993, DG AGRI has promoted the use of “Control with Remote Sensing” (CwRS), which was developed with the technical support of the JRC, to increase transparency between Member States while having a strong deterrent effect on potential fraud. In 1998, DG AGRI transferred to the JRC the responsibility of coordinating and controlling this activity. The CAP reform introduced in 2005, Reg. 1782/03 and 796/04, maintains the use of remote sensing for the control of the key elements (areas, land use) and the new requirements of the control of the cross compliance.
CwRS is presently used by 23 Member States with around 180 sites (2005). Since 2000, CwRS almost always combines aerial ortho photos or satellite ortho imagery to ensure an area measurement accuracy fully compliant with the recommendations of DG Agri for the On-The-Spot Checks (Promotional materials, AGRIFISH Unit of IPSC/JRC).
For the 2006 campaign the Commission increased the area covered by VHR sensors from 126,000 km2 to approx 140,000 km2 (Quickbird and Ikonos as prime dedicated, and with backup of EROS A, and SPOT Supermode. The intention was also to use Orbview3, Formosat2 and possibly Eros B as back up). The use of such data for controlling the Single Payment Scheme and the Single Area Payment Scheme appears appropriate in conjunction with rapid field visits (RFV). (Common Technical Specifications for the 2006 Campaign of Remote Sensing Control of Area-Based Subsidies, ITT no. 2005/S 233-229332, Ispra, 2005).
Formosat 2 was programmed as VHR backup sensor for the 2006 CwRS Campaign over 12 control zones (7.474 km2). The success rate was 87.5% (area), 83.3% (10/12 sites). The delivery of cloud free imagery was more than acceptable (88.9% < 5% CC, 11.1 % < 8 % CC on a total of 27 uploads).
The test was proposed by JRC aiming to detail the difficulties encountered to reach the required location accuracy in production of Level 3 orthorectified products. The test was performed by ReSAC, Spot Image and Spacemetric AB. ReSAC was responsible for the test with ERDAS and PCI and provided high accurate DGPS measurements and Ikonos Orthoimage map for selection of the GCPs and ICPs.
Figure 1. Location of imagery acquired for the study site. The dark-grey rectangle with white border defines the footprint of the Formosat-2 imagery.
The study area covered the extent of Sofia City – the capital of Bulgaria, – and the Northern hillsides of Vitosha Mountain. The capital is situated in Sofia Valley which is an important for the agricultural plain. The average elevation inside the city is 550m a.s.l., while the nearest highest point is Cherni Vrah (“Black Peak”), 2290m, located to the South, in the Vitosha Mountain.
The study area presents various landscapes and terrain variations, thus being a suitable test site for orthorectification and geometry quality assessment.
FORMOSAT-2 (NSPO, 2004) carries two cameras that take imageries of the Earth in the visible and near infrared electromagnetic spectrum. The swath covered by these high resolution cameras is 24 km at Nadir and their nominal instantaneous geometric field of view, at Nadir, is 2 metres for the Panchromatic sensor and 8 metres for the Multispectral sensor. For the herein study the imagery with the highest spatial resolution was considered, i.e. the panchromatic one, as it requires greater accuracy for the orthorectification result.
The image was delivered as row imagery, Level 1A, with basic radiometric normalisation for detector‘s calibration, and with no geometric correction. The product is in DIMAP format and as such comprises of a GeoTIFF file for storing the imagery and XML – for the metadata.
Figure 2. FORMOSAT-2 image used in the test, acquired on 8 Sep 2006 and covering Sofia City.
A series of orthorectification tests were carried out in order to evaluate the operational performance of the FORMOSAT-2 sensor in the production of orthoimages. The study shows that it was comparatively straightforward to produce reliable products, well inside the expected performance for the CwRS requirements – 3.5 RMSE1D (i.e. in either Northing or Easting directions).
Given that the objective of the study was to determine whether FORMOSAT-2 imageries could be used in operational mode for farmers’ subsidies monitoring, only internationally recognised software was considered. Specifically, for this study, PCI Geomatica 10 and ERDAS Imagine 9.1 were tested for orthorectification performance. In order to ensure the consistency of the software performance test, all GCPs and ICPs were identically chosen for each software-respective test, and their coordinates were transferred via import, to avoid interpretation errors during the tests.
In order to eliminate the influence of the DEM accuracy over the orthorectification results the best available elevation dataset was chosen – Reference3D product by Spot Image.
Figure 3. Overview of the terrain across the test area. For orientation IKONOS orthophoto was draped over the Reference3D DEM.
The geometric assessment that was undertaken is systematic and conforms to the standard method developed by the JRC (European Commission, 2006). This method applies strict use of points other than the one used in the orthorectification, i.e. ICPs, for the evaluation of image correction performance, which allows the comparative robustness between different processing methods.
Future research should be carried out as to define the optimal number of GCPs to be used when orthorectifying FORMOSAT-2 images on a routine basis. Furthermore, the effect of the incidence angle on the accuracy of the orthorectification should also be studied. These further investigations have to be performed for both PCI Geomatica and ERDAS Imagine software packages, as it is likely that diverse models behave differently. Such profound analysis could aid a speedy and quality optimised orthorectification production having in mind that around 10,000 km2 of FORMOSAT-2 data within suitable corridor are envisaged as VHR backup for the 2007 years CwRS Campaign.
The results of the test were presented on the 12th MARS-PAC Annual Conference “Geographical Information in Support of the CAP” (27-29 Nov 2006, Toulouse, France) and will be published in a special publications edition of the conference reports.
(Source Remote Sensing Application Center – ReSAC )