Orthorectification Tests Continued... Formosat-2 Orthorectification Of Formosat-2 Data for use in The Common Agricultural Policy Control with Remote Sensing Programme

Abstract

FORMOSAT-2 (NSPO, Taiwan) was launched on 21st of May, 2004. FORMOSAT-2 was programmed as Very High Resolution backup sensor in the CwRS campaign for the first time in 2006 [Ref 1]. Acquisition success rate has been high since it was introduced due to its high (daily) revisit capacity, but difficulties were initially encountered to reach the required location accuracy in production of orthorectified imagery. This resulted in a 1st study (2006) where FORMOSAT-2 imagery over Sofia (BG) was assessed; 4 software suites were tested on this image with low off-nadir viewing angle [Ref 2]. Results were promising, demonstrating that it is possible to perform good orthorectification using standard software packages reaching results inside the CwRS requirements for such imagery (location accuracy preliminary set to 3.5m RMSE1D). In this 2nd study (2007) the aim has been to assess the effect of large off-nadir angles on the accuracy of the orthorectification, and to define the optimal number of GCPs to be used when orthorectifying FORMOSAT-2 images on a routine basis. Results of orthocorrection of 4 images of different off-nadir angles (along/across angles), over 2 sites in France and Bulgaria, using 4 different sw suites (PCI, ERDAS Imagine, PRODIGEO, and Keystone SIPOrtho,) and with varying number of GCPs are discussed. The results are consistent with theoretical expectations; x error increases when across angle (roll) increases, the y error increases when along angle (pitch) increases. Basically the accuracy of 5m RMSE1D is reached with all tested softwares, the 3.5m RMSE1D accuracy may be reached if limits are placed on the acquisition angles. Concerning the GCP requirement a total of minimum 10 GCPs should be used: four GCPs spread in the corners of the scenes, the others evenly distributed, and clearly visible.