Geolocation evaluation of RapidEye 3A and 1B orthorectified: case study in São Sebastião/SP/Brazil
ISBN 978-85-88783-11-9
Authors
1Rafael, B.; 2Alexandre, T.; 3Carla Madureira, C.; 4Lis, R.; 5Fabio, F.
1FEDERAL UNIVERSITY OF RIO DE JANEIRO Email: rafael.barros@ufrj.br
2IBGE Email: alexandre.teixeira@ibge.gov.br
3FEDERAL UNIVERSITY OF RIO DE JANEIRO Email: cmad@ufrj.br
4CENPES/PETROBRAS Email: lismaria@petrobras.com.br
5TRANSPETRO/PETROBRAS Email: fabio.fagundes@petrobras.com.br
Abstract
The purpose of this study is to present the results of geometric evaluation of a RapidEye image provided in two different processing levels: 1B and 3A. Since RapidEye images in 3A processing level are distributed in pre-defined tiles, it may be composed by more than 1 scene. In this case study tile was composed by two scenes, acquired in February and May 2013. For the evaluation of RapidEye 1B it was used both scenes that compose the 3A tile. The two scenes were orthorectified using RPC and 8 ground control points (GCPs) and SRTM MDE, in the PCI Orthoengine 2013. Orthorectification process presented RMS of 5.2 and 5.72 meters for February and May, respectively, 1Bimages. The evaluations were done comparing coordinates of 30 Checkpoints (not used in the orthorectification process) per orthoimage with its homologues identified all over the orthoimages (3A product and orthoimages generated from the 1B scenes). Results show that geometric quality was quite similar considering the 3A product (orthorectified by RapidEye owner/dealer) and orthoimage generated from 1B product. RapidEye 3A product presented, in this case study, CE90 =7.76 meters and RMS=2.52 meters. RapidEye 1B orthorectified presented CE90=8.04 meters and RMS=2.48 meters (image from February 2013) and CE90=6.28 and RMS=2.57 meters for image from May 2013. Considering Brazilian Standards for Cartography (PEC), values obtained in this evaluation would indicate that orthoimages presented geolocation accuracy according to classes B (May 2013 orthorectified image) and C (RapidEye 3A product and February 2013 orthorectified image) for 1:10.000 scale. Considering statistical inferences proposed by Vieira & Genro (2013) that estimate accuracy for error population and trend analysis, results are not so good: 3A product and February and May 2013 orthorectified images would present accuracy according to class B for 1:25.000 scale. Since this approach is more robust than the one proposed in PEC, authors consider that RapidEye orthoimages evaluated in this case study (no matter if it is a 3A product or the orthoimages generated from 1B product) present geolocation accuracy according to class B for 1:25.000 scale. Authors also emphasize that only geolocation accuracy was evaluated in this study. These products must be evaluated considering thematic accuracy - which is ongoing and results will be presented in the near future.
Keywords
Accuracy; Orthoimage; Remote Sensing