Land cover land use mapping using image fusion techniques and machine learning algorithms

Abstract

Image fusion based Land cover classification is one of the important processes in remote sensing application using machine learning and deep learning algorithms. It is used for managing environments and urban planning. The government faces many issues related to urban planning because of increasing population. To solve the above issues a land cover classification is proposed in this thesis work. These remote sensing images are captured by using satellite which provides panchromatic (PAN) and Multispectral (MS) images, and the images are fused for land cover classification that includes both spectral and spatial information of the images. Generally, the remote sensing images have many distortions which is to be removed it from the images for obtaining better results during feature extraction (spatial and spectral) and classification. Experiments were conducted to evaluate the performance of the proposed methodology. In first work, Quantization Index modulation with Discrete Contourlet Transform (QIM-DCT) based fusion method is proposed for land cover classification using satellite images. LISS IV sensor image which are captured by Indian Remote Sensing Satellite P6 (IRS P6 satellite) and PAN image captured by CARTOSAT-1 satellite is used in this thesis work. To improve the quality of the images and to analyse the performance of the image fusion, color conversion and Bayesian filter with Adaptive type-2Fuzzy system were proposed in this work. In this process the images are converted from RGB to gray scale which eliminates the Gaussian and salt andpepper noise from the images and also increases the performance of image fusion. After completion, three processes such as feature extraction, feature selection and classification were performed. Affine transformation is used to extract the multiple features (spectral, shape, local and global features) from the images. For selecting optimal features this research calculates the coefficients which include mutual and maximal information of the features. Finally, the images are