Abstract

Semi-automated delineation of target from high resolution satellite imagery
Track: Imagery and Raster
Authors: Suparn Pathak

For large scale mapping & manual digitization process is most cumbersome and time consuming its constraint necessitates automation. Delineation of target of interest has been carried out using Cartosat -1 satellite data along with DEM derived from IRS P5 stereo pair. High resolution satellite data does not follow the normality of class's assumption and therefore edge detection followed by segmentation is viable solution. Object oriented classification has its own advantages but jagged outlines are generated. Morphological operator in ArcGIS is used for shape preserving edge detection. Edge generate forms the boundary of segment. Binary image does not require line thinning. Auto vectorisation attempted has some in-built processing 'cleaning' of images. Post processing usually requires editing still not comparable with human interpreter. Improvement consists of generalisation in GIS environment with no topological problem. The result thus obtained serves both qualitative and quantitative detection of feature of interest.