Extraction of infrastructure details from fused image

This paper demonstrates a method of extracting features included edge features from a data fusion image. The intensity-hue-saturation (IHS) transformation is used for producing the data fusion image. This is achieved by merging the SPOT panchromatic 10-m resolution image with the Landsat TM 30-m resolution multispectral channel image. The resultant imagery has a high resolution and spectral characteristics that enhance its visualization. The image was then subjected to the processes of thresholding, Gaussian filtering using a low pass filter and the Iterative Self-Organizing Data Analysis (ISODATA) unsupervised classification to derive the feature classes. Ground truthing of this classification showed an overall accuracy of 89 %. To perform the edge features extraction, an edge detection operator (Sobel edge filter) is applied to the classification image and produces a gradient image. This gradient image is threshold to produce a binary digital edge map.