WATER BODY EXTRACTION FROM RAPIDEYE IMAGES: AN AUTOMATED METHODOLOGY BASED ON HUE COMPONENT OF COLOR TRANSFORMATION FROM RGB TO HSV MODEL

Abstract

Water management and flood studies are some fields in which a map with all water bodies in a region is useful, especially in scenarios of environmental changes due to anthropogenic factors. Various detection methods of water body surfaces in remotely sensed images are available, from simple methods having a lower accuracy to more sophisticated ones. The objective of this paper is to present a simple, yet accurate method to detect water bodies in RapidEye images. The motivation is the availability of country wide coverage of these images, which makes feasible the generation of a map of all water bodies detectable at that spatial resolution. Our solution is the use the color transformation from Red-Green-Blue to Hue-Saturation-Value and the minimum radiance from all RapidEye bands to classify water bodies in seven classes of water. The water classes are ranked based on the confidence of the classified pixels being water, which accommodates for the differences in illumination and scattering that are present in such a large coverage, composed by more than 15000 scenes. In addition, users of the generated water bodies map can reclassify based on their needs. The methodology was developed on two RapidEye scenes, covering the Jacareí and Foz do Iguaçu municipalities, in Brazil. Results indicate that the classification is better than the traditional ones, with the advantage of providing seven classes with confidence levels.