Parallel TLBO Framework for Structuring Elements Optimization in VHR Image Classification

This paper investigates the optimization of structuring elements in classification of very high resolution (VHR) image through mathematical morphology (MM). Indeed, the standard SEs are not suitable for the classification of VHR images from urban regions in which the spatial features are very complex and vary in shapes and sizes. Therefore, perceiving systematically the best layout of the structuring element (SE) to VHR image classification is an accelerated research subject. The aim of this paper is to propose an adaptive parallel teaching-learning based optimization (PTLBO) framework that is able to work out a unique SE whose configuration is optimal and non-fixed, which may change adaptively in conformity with the formats of the features and the classification duty. The proposed optimization framework is made independent of any algorithm-specific parameters and not constrained by using a finite quantity of SEs in the classification. Moreover, a cascade SVM (CSVM) classifier is adopted in this work during the phase of classifying the morphological features in order to achieve the highest classification results within the desired time. Finally, the experimental tests are performed in order to validate the efficacy of the proposed PTLBO framework, and the comparative results reveal the advantages of the presented approach in contrast to the prior works.