A Novel Optimized Land Cover Classification Framework Using Data Mining Techniques

Abstract

The main purpose of this study is to highlight the importance of data mining techniques for the classification of land cover (LC) types such as fertile cultivated land, green pasture, desert-rangeland, bare land and Sutlej-river land. A novel framework is designed to classify the subjective land cover types. Visually three selected land cover, desert rangeland, Sutlej river land and bare land have almost similar physical features and remaining two, fertile cultivate land (cropland) and green pasture (grass) have also to some extent similar physical features. It seems very difficult to discriminate these vast land cover areas when remotely sensed. For this study, data have been acquired by handheld crop scan device Multispectral Radiometer (MSR5) in the form of five spectral bands such as, blue (B), green (G), red (R), near-infrared (NIR) and shortwave infrared (SWIR), while texture data have been acquired from the digital photograph at the same site and location in the region of Bahawalpur Punjab Pakistan. Feature selection and reduction techniques are employed on statistical texture data to get an optimized set of features, while for MSR5 dataset, there is no such type of processing is required. These both type of data sets are deployed to WEKA software version (3.6.12) for classification. A comparative analysis is performed on the results of Multilayer Perceptron (MLP), Random Forest (RF), j48 and Naïve Bayes (NB). It is observed that MLP outperformed exceptionally and received an overall accuracy of 97.333% for texture dataset and 96.66% for multispectral dataset respectively.