從過去的文獻可以發現「粒子群優法(Particle Swarm Optimization, PSO)」應用於「貪婪模組特徵空間法(Greedy Modular Eigenspace method)」趨近最佳解的方法中,在選取具代表性波段的步驟,沒有一個有效、低運算量的維度優先權方法。因此,本論文提出「非純度維度優先權(Impurity Function Band Prioritization, IFBP」方法:先使用PSO演算法,對高光譜影像資料的相關係數矩陣聚合多個高相關度模組,形成一組群聚特徵空間;再藉由IFBP公式,計算出類別覆蓋率,而得到各波段的優先權,最後在每一個高相關度模組中,挑出該模組最優先的波段作為最具代表性波段,將這些最具代表性波段交由分類器進行分類。本文使用 Northwest Tippecanoe County 的AVIRIS 遙測影像以及鰲鼓溼地的 MASTER 遙測影像作為實驗所使用的圖資。從實驗結果可以得知,相對於現有文獻提出的維度優先權法而言,運用本文提出的IFBP方法可以挑選同一模組內對分類最有貢獻的波段,獲得高正確率的降維品質;並解決過去使用PSO波段選取法,在選取具代表性維度時沒有一個可靠依據的問題,此為本論文最大貢獻。
Band selection for hyperspectral images is an effective technique to mitigate the curse of dimensionality. Many band selection methods have been suggested in the past. In this paper, a novel band prioritization based on impurity function (IF) is considered for the band selection of hyperspectral images. The proposed IF band prioritization (IFBP) is incorporated with particle swarm optimization (PSO) band selection which has been developed to effectively group highly correlated bands of hyperspectral images into high corrected modules. It makes use of a particle swarm optimization scheme, which is a well-known method to solve the optimization problems, to develop the effective feature extraction algorithm for hyperspectral imagery. The proposed IFBP is introduced to enhance the efficiency of band selection after PSO method is applied to the band reduction of hyperspectral images. The propose method are evaluated by AVIRIS which built by Jet Propulsion Laboratory and MODIS/ASTER airborne simulator (MASTER) during the Pacrim II campaign. The performance of IFBP is validated by supervised k-nearest neighbor (KNN) classifier. Experimental results demonstrated that the proposed IFBP approach is an effective method for dimensionality reduction and feature extraction. Compared to several band selection methods, IFBS can effectively select the most significant bands for the image classification of hyperspectral images.