Impact of cluster sampling on the classification of landsat 8 remote sensing imagery

Abstract

Remote sensing images are rich enough in terms of both qualitative and quantitative information. Extraction of different land features on earth surface from imagery is highly required. Since satellite imagery includes very large data sets, it is quite difficult to deal with such large data sets, specially extracting the land features. Classification techniques help researchers and analyst to extract such features available on earth surface which further depends on the sampling methods applied for testing purpose. In this study, the focus is on sampling methods, how the samples should be collected so that the classification of satellite imagery can be done effectively and efficiently. Out of many sampling schemes stratified random sampling and cluster sampling Congalton RG, Green K (Assessing the accuracy of remotely sensed data: principles and practices. CRC Press, 2008 [1]), Hashemian MS, Abkar AA, Fatemi SB (Int. Congr. Photogramm. Remote. Sens. 2004 [2]) and Vieira CAO, Santos NT (Analysis of sampling methods and its influence on image classification process of remotely sensed images through a qualitative approach, pp. 6773–6780, 2009 [3]) are used to classify Landsat 8 imagery, and it is found that cluster sampling outperforms with an accuracy up to 95.8%.