Open AccessWater Feature Extraction, Enhancement and Change Detection of Multi-Temporal Satellite Images using MNDWI2-PCA
Author(s) -
M. Hemalatha
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1049/1/012005
Subject(s) - pixel , remote sensing , artificial intelligence , computer science , satellite , water extraction , feature (linguistics) , confusion matrix , principal component analysis , pattern recognition (psychology) , cohen's kappa , feature extraction , image resolution , computer vision , environmental science , extraction (chemistry) , geology , linguistics , chemistry , philosophy , chromatography , machine learning , engineering , aerospace engineering
Water feature extraction is a challenging task in remote sensing. In this research work, a new water index is implemented for easy identification of water pixels. The Area of interest is extracted with desired shape file. Here water bodies from kalahasti region are extracted, which is in Chittoor district. The Water indices are used to identify water pixels from Landsat-8 image, which has high spectral resolution. This image is multi-spectral image comprising of eleven bands. Interactive supervised classification is implemented for segmenting the satellite image. The image is classified into two categories i.e. water bodies and non-water bodies. Then MNDWI2-PC (Normalized Difference Water Index2-Principal Component) is applied to LANDSAT-8 image. Then this image is segmented into water bodies and non-water bodies. Finally accuracy assessment is carried out by confusion or error matrix. Quantitative parameters such as Overall Accuracy (OA), Kappa Coefficient (KC), Overall Kappa Coefficient (OKC) User’s accuracy (UA), Producer’s Accuracy (PA), and F1 score (F-Measure) are calculated for this multi-spectral satellite imagery. The algorithm reduced misclassification of water pixels with urban pixels, vegetation and other land covers. The algorithm outperforms in terms of quantitative performance metrics.