Geolocation Evaluation of Rapideye 3a and 1b Orthorectified: Case Study in São Sebastião/Sp/Brazil

Abstract

This study aims to present the results obtained in the geometric evaluation of two RapidEye scenes, acquired in two levels of processing: 1B and 3A. To evaluate the RapidEye 1B data, the parties were considered the two scenes that make up the 3A cut. Both 1B orthorectified scenes were considering two scenarios: using only the RPCs and also using nine-point control in the field (GCPs), and the MDE SRTM, using PCI Geomatic OrthoEngine 2013. The determination of the accuracy was made using 30 evaluation points. All points (control and evaluation) were collected in the field by making use of GNSS receivers of two frequencies through static relative positioning. For the assessment of accuracy was made comparing the coordinates of the evaluation points with their counterparts in orthoimagery (either 3A or that have been generated from 1B scenes). The 3A product has CE90 = 7.76 meters and standard deviation = 2.52 meters. When considering orthoimagery derived from the data 1B with GCPs, we have CE90 = 8.04 meters and standard deviation = 2.48 meters for February 2013 scene, and CE90 = 6.28 meters and standard deviation = 2.57 meters to the scene May 2013. For orthoimagery corrected without GCPs met CE90 = 6.8 m and standard deviation = 2.38 meters for the scene, 2013 and CE90 = 7.64 meters and standard deviation = 2.76 meters for the scene February 2013. Considering the PEC (Cartographic Accuracy Standards), the values obtained allow saying that the observed accuracies are compatible with the specified product to up the scale 1: 10,000, class B (orthoimage, 2013) and class C (orthoimagery February 2013 and 3A product). Making use of statistical inferences proposed by Vieira & Genro (2013), which estimate the accuracy for the population of the errors and the trend in the data, the results were not so good: all orthoimagery (3A, February and May 2013) would accuracy compatible with the specified product to scale 1: 25,000, class B. As the approach statistical inferences is more robust than specified in the PEC, the authors believe that the geolocation of RapidEye images evaluated in this study should be considered compatible with products in scale 1: 25,000, class B. The authors emphasize, though, that only the planimetric accuracy or the geolocation was evaluated. Products must undergo assessment of thematic accuracy to have a full assessment of your application, which is in progress.