Change detection technique based on the radiometric rotation controlled by no-change axis, applied on a semi-arid landscape

Abstract

We demonstrate a new digital technique for change detection that integrates field survey data in a methodology. The field survey was specifically designed to control the operation of this new technique for change detection. The Radiometric Rotation Controlled by No-change Axis (RCNA) technique was applied with spectral bands in the visible spectral range without atmospheric correction. The 190.37 km2 study area is located in the centre of the semi-arid region of northeast Brazil. This region is occupied by a natural shrub vegetation and low arboreal type steppe. This is known locally as 'caatinga', and in this area the predominant use is grazing by cattle and goat. Two TM/Landsat images in digital format of the dry seasons of 1984 and 1996 were used together with two ETM+/Landsat images for the dry seasons of 1999 and 2001. Three change maps were generated, all with accuracy index values of around 0.70, which is acceptable for areas with complex patterns like semi-arid lands. In conclusion, there is no need to apply atmospheric correction methods to a pair of multi-temporal images to obtain an acceptable accuracy of change detection in semi-arid regions. Although this technique can combine images from different sensors, this requires that both have similar spectral ranges.