A Novel Method for Multispectral Image Classification by Using Social Spider Optimization Algorithm Integrated to Fuzzy C-Mean Clustering

Abstract

In remote sensing, Fuzzy C-Means clustering (FCM) is a robust method in determining membership grades of a pixel belonging to 1 or more classes. This paper proposes a novel approach by using the social spider optimization (SSO) algorithm in solving the search for optimal cluster centers in FCM. Hanoi, the capital of Vietnam, was chosen as a case study because of its spatial complexity. Multispectral satellite datasets of Landsat 8, Sentinel 2A and SPOT 7 were used. The experiment started with the segmentation process, followed by an examination of the model, then the results were compared with several conventional clustering methods. For accuracy assessment, the FCM minimizing objective functions, user and producer accuracies and overall accuracy were used. The results showed that SSO significantly improved the performance of FCM and outperformed the benchmarked classifiers or other common optimization algorithms. It could be concluded that the model was successfully deployed in the study area and could be suggested as an alternative solution for urban pattern detection. In a broader sense, classification methods will be enriched with the active and fast-growing contribution of nature-inspired algorithms.