SatFocuser: A Software Solution for Dealing with the Mixed-pixel Classification Problem Arising when Analyzing Raster Based Environmental Data

Abstract

Environmental modelling of terrestrial ecological processes relies heavily on information which is obtained from space born data and/or aerial photographs. Complexity of terrestrial ecological processes creates a lot of problems when analyzing raster based digital data derived from the Earth's surface as definite land cover class boundaries cannot be easily drawn between different classes/segments in the images. For example, in nature, a wetland ecosystem having a grassland ecosystem in its periphery will have no clear-cut boundary to separate it from the latter. The intermixed zone between two adjacent ecosystems is called an ’ecotone’. In ecology, ’ecotones’ are important areas of study due to the complexity of the processes they harbor within. It is often a problem for the modeler to clearly demarcate the ecotone land-cover class areas (or “fuzzy areas”). Usually, ecotones will have mixed pixel values i.e. values of both landcover classes that have combined to create them. Hence, from an image processing point of view, finding clear boundaries between the two or more different features in the real world on the respective environmental raster data harboring mixed-pixels is thus much more difficult where the transition areas can be seen. However, the identification of the demarcation of these kinds of “fuzzy area” boundaries is important for understanding of ecological processes — especially on how ecotones develop, merge and diverge. Since the standard ’off the shelf’ GIS software programs are unable to perform this difficult task - a high-level software programming solution is deemed neces5ary. In this paper, we pre5ent a software solution to identify the mixed areas in a given image via analysis of pixel elements. In this method, pixel value for mixed area is calculated by taking of arithmetic mean of the pixels values which are scattered in pure area. Through the scanning of entire pixels, corresponding pixels are selected and marked as mixed areas. During this process, the ’mixed areas’ in the given image files of format jpeg, tiff, and bmp can be clearly identified electronically. A user-friendly software tool “SatFocuser” incorporating this methodology has been designed with support of the IDE Visual studio.NET 2008, using interior coding of C sharp. Further investigative explorations of ’SatFocuser’ would no doubt provide a better understanding of the problematic mixed pixel issue arising in environmental image processing