Research Article Open Access

Investigation of Optimum Spectral Bands for Urban Area Classification from World View2 (WV2) Satellite Image

Abstract

Satellite imaging provides not only color images with RGB data but also other information related to invisible bands which is called multi-spectral bands. Images with multi-spectral bands give rich data for features which is useful in description, classification and extraction. The research objective is to investigate the optimum bands for urban area classification from multi-spectral images based on available data. Images from World View2 (WV2) satellite with eight spectral bands in addition to panchromatic image for the captured area are available. This amount of spectral information together with the very high spatial resolution of WV2 imagery provides feature details which is suitable for mapping of land cover. Maximum Likelihood (ML) is a classification technique that is used to classify image of area of study while testing a combination of multi-spectral bands of WV2 satellite image. The proposed research algorithm is based on forming a different set of multiband. Each set is composed of three different bands from input WV2 satellite image. ML classification technique is applied on all sets to extract classified images of study area from all used sets. The assessment of classification results is represented in confusion matrix format and determination of Kappa Coefficients. The research work flow is processed commercial software ENVI( an acronym for “Environment for Visualizing Images”). The preliminary results shows that a set of (Near Infrared (NIR), Green (G) and Costal (C) bands) give accurate classification result for the area of study. Image with NGC bands gives a classification results with overall accuracy 91 % and it is evaluated by Kappa coefficient which is 0.88.

Mahmoud A. Shawky, Essam H. Hamza, Ahmed S. Elsharkawy, Hassan E. Elhifnawy

To read the full article Download Full Article