PREDICTION OF CHANGES IN LAND COVER/LAND USE REGION USING PARALLEL PROCESSING METHOD

Abstract

The adoption of high-resolution hyperspectral images to view the land cover regions has turned out to be a hotspot in the field of remote sensing. The land cover detection process is merged with some preliminary pre-processing methods to extract essential information; however, the manual sample selection process is more complex for a wide range of remote sensing images. Thus, this research concentrates on modelling an approach to predict the changes in land cover regions and forward the preliminary stage to successive stages for measuring the structural similarities. Here, the input is taken from the hyperspectral dataset and fed to the next stage for further processing. It is explained with four stages, and they are: pre-processing with Principal Component Analysis (PCA)-based dimensionality reduction; cluster formation is done with parallel processing of Bacterial Foraging Optimization (BFO) in a parallel processing manner for weight updation, and it is merged with k-means clustering to measure the fitness function. Finally, post-processing is done with Region-based segmentation for showing the colour representation based on the bands. All these processes are done with MATLAB 2016b simulation, and the proposed model outperforms the existing approaches. The evaluation of the anticipated model is done with various metrics like accuracy, precision, recall, F-measure, minimal weight measure, True Positive (TP), True Negative (TN), False Positive (FP) and False Negative (FN). This model attains 86.37% prediction accuracy, 83.32% precision, 100% recall, and 86.08% F-measure, respectively.