Evaluation of semiautomatic geometric correction of the Landsat 7 image using as base Landsat 8 geometry for temporal analyses
ISBN 978-85-88783-11-9
Authors
1Duarte, G.S.; 2Costa, G.E.; 3Oliveira, A.P.; 4Barros, R.S.; 5Cruz, C.B.M.
1UFRJ Email: gabriel11705@gmail.com
2UFRJ Email: genilsonestacio@gmail.com
3UFRJ Email: ana.paulaoliveiran@gmail.com
4UFRJ Email: rafael.barros@ufrj.br
5UFRJ Email: cmad@ufrj.br
Abstract
Temporal analysis constitute an important kind of comparison in many ways of Geography, being a task most of times complex because of a temporal variation of the object of study and of the data that are limited by the technology of its time. A lot of times, the limited mapping accuracy of the old data, often related to the precision obtained at the time, may difficult their uses because of the necessity of a series of processes to make them consistent with new data. Among the technological development increasing is expected that this geometric correction becomes simpler to be done, more automatic and precise as if it were done manually, by assigning its product a coordinate system and a reference scale. Based on that, This paper aims to analyses the orthorectification performed by USGS / NASA in Landsat 8 images in accordance with the PEC (Cartographic Accuracy Standards), as well as test and evaluate the method of semi-automatic georeferencing offered by software ENVI module Registration in the satellite Landsat 7 images, using as a basis for the collection of points one Landsat 8 image. The PEC was established by a decree of the 80s by the Brazilian government to classify cartographic products in relation to the displacement in the field, in which up to 90% of check points in a map, when tested on the field, should not exceed the appropriate set for roster error. Likewise, there is a limit to the standard deviation of the error value. The PEC classifies cartographic products into three classes (A, B and C) according to the quality of the product. It will be georeferenced semiautomatically a raw Landsat 7 image, distributed by INPE, usying as reference the Landsat 8 geometry, and evaluating the product generated by the PEC using as a comparison the roads base of Instituto Pereira Passos (IPP) 1:2.000 from Rio de Janeiro’s city. As results, The geometric correction made by the USGS/NASA in Landsat 8 images showed surprising for the proposal and resolution of the sensor. It achieves the mapping classes A in 1:50.000 and 1:100.000 scales, and with the possibility of having the spatial resolution magnified to 15 meters using the panchromatic band, it is possible to reach even better results. Can also be used to fix other similar products or with lower resolution. This is very important for the compatibility with historical Landsat images, demonstrating that there is a huge potential for spatiotemporal studies that use the Landsat 8 as the basis for georeferencing of older images, while maintaining good accuracy in scale 1:100.000. The module Registration of ENVI showed easy to use and with numerous possibilities, including georeferencing of products with different resolutions. The georeferencing done through this software showed very satisfactory, having corrected images quickly and generating products with good quality, both in areas of flat relief as mountainous. In the relationship between expended time and product quality the software question proved to be a great tool for projects that use many images for adopting large scales, be spatial or temporal. Thus, it appears very suitable for work in Brazil where, as a country of vast territory, it is common to work with multiple images due to the adoption of extensive spatial slices. Although there is a cost to purchase the program and the module, large projects could acquire it aimed to gain speed in georeferencing, since the semi-automatic georeferencing for Registration could provide considerable savings of time.
Keywords
Geometric Correction; Georeferencing; Rio de Janeiro