Band Selection of Hyperspectral Imagery Using a Weighted Fast Density Peak-Based Clustering Approach

摘要

Based on the search strategy of representative bands in Hyperspectral Imagery, various existing unsupervised band selection approaches are mainly classified into two parts: ranking-based and clustering-based ones. Recently, a fast density peak-based clustering (abbreviated as FDPC) algorithm has been proposed. The product of two factors (the computation of local density and intra-cluster distance) is sorted in decreasing order and cluster centers are recognized as points with anomalously large values, hence the FDPC algorithm can be considered as a ranking-based clustering method. In this paper, the FDPC algorithm has been modified to make it suitable for hyperspectral band selection by weighting the normalized local density and intra-cluster distance. It is called a weighted fast density peak-based clustering (W-FDPC) method. Experimental results demonstrate that the bands selected by W-FDPC approach can achieve higher overall classification accuracies than FDPC and other state-of-the-art band selection techniques.