Publication

Abstract : The classification of high-definition hyperspectral images from metropolitan locations needs to resolve specific critical issues. The conventional morphological openings and closures weaken object barriers and deform item forms, making the process more challenging. The morphological openings and closings by reconstruction can circumvent this issue to an extent with a few unintended effects. The images that are anticipated to vanish at a particular scale stay available later when morphological openings and closings are done. The extended morphological profiles (EMPs) with unique structuring factors and a growing number of morphological operators generate extremely high-dimensional data. These multidimensional facts may also contain duplicated data, creating a brand-new classification task for standard classification algorithms, particularly for classifiers that are not resistant to the Hughes phenomenon. In this article, we examine extended morphological profiles with partial reconstruction and directed MPs to categorize high-definition hyperspectral snap images from urban locations. Second, we expand it using a Hybrid 3D CNN to boost classification performance. In this article, the total accuracy of 99.42% is obtained with a small number of testing samples. Similarly, the average accuracy is reasonable when compared to other 2D and 3D convolutional neural networks.