Weighted Residual Dynamic Ensemble Learning for Hyperspectral Image Classification

Recently, collaborative representation classifiers have been extensively studied as an essential method for hyperspectral image. However, how to comprehensively utilize the classification advantages of multiple has not well investigated. In this paper, two new dynamic ensemble learning methods using local weighted residual (LWR-DEL) and double-weighted (DWR-DEL) multi-collaborative are proposed. First, based on clustering is utilized introduce prior knowledge classifier. Then, with knowledge, weights each classifier a different region competence obtained. To consider global information data, K-nearest neighbor (K-NN) algorithm adopted achieve validation samples information. The can be obtained then used constrain locally residuals. Similar LWR-DEL, also residual, constrained fusion obtains final result. effectiveness proposed validated three data sets. experimental results show that both LWR-DEL DWR-DEL outperform their single-classifier counterparts. particular, provide superior performance compared state-of-the-art methods.