Clustering for High Resolution Monochrome Satellite Image Segmentation
D. Chakraborty
Regional Remote Sensing Service Centre, Indian Space Research Organization, IIT Campus Kharagpur – 721 302, India
Gautam Kumar Sen
School of Oceanographic Studies, Jadavpur University, Kolkata – 700 032, India
Sugata Hazra
School of Oceanographic Studies, Jadavpur University, Kolkata – 700 032, India
A. Jeyaram
Regional Remote Sensing Service Centre, Indian Space Research Organization, IIT Campus Kharagpur – 721 302, India
要旨
Variability of spectral and structural information in high resolution remote sensing images requires considerable modification in pixel-based segmentation techniques for appropriate analysis and interpretation. A three step iterative procedure for clustering has been adopted to (i) identify range of pixel values, densely populated in the kernel (ii) estimate optimal frequency of range of pixel values (iii) detect shape and size of the cluster. Two thresholds corresponding to pixel intensity range and percentage of occurrence within the range in the kernel has been used to detect the shape and size of the cluster.
