


摘 ?要: 對于遙感圖像分類過程中的問題,提出遺傳算法LVQ神經網絡來實現遙感圖像的分類。將LVQ神經網絡結合遺傳算法,使用遺傳算法最優閾值與權值實現網絡訓練,使分類精度得到提高。之后融合相似灰度值創建分類圖像特征矢量,使特征矢量在神經網絡中輸入實現訓練。學習矢量量化神經算法對初值非常敏感,對遙感圖像分類精度具有一定影響。最后,為了對性能進行測試,在實驗過程中對比本文分類方法和SVM決策樹分類方法,通過實驗結果表示,文中提出的分類方法的遙感圖像分類精度為95.82%,與其他分類方法相比,分類精度得到進一步提高。
關鍵詞: 遙感圖像分類; 遺傳算法; LVQ神經網絡; 網絡訓練; 性能測試; 精度評估
中圖分類號: TN911.73?34 ? ? ? ? ? ? ? ? ? ? ? ? 文獻標識碼: A ? ? ? ? ? ? ? ? ? ? ?文章編號: 1004?373X(2020)01?0040?04
Application of genetic algorithm LVQ neural network
in remote sensing image classification
DENG Lingyun
Abstract: In order to solve the problems in the process of remote sensing image classification, a genetic algorithm LVQ ?(learning vector quantization) neural network is proposed to realize remote sensing image classification. The LVQ neural network is combined with genetic algorithm, and the optimal threshold and weight of genetic algorithm are used to train the network, so that the classification accuracy is improved. Then, similar gray values are fused to create characteristic vector of classified images, which are input into the neural network for training. LVQ neural algorithm is very sensitive to initial values and has a certain impact on the classification accuracy of remote sensing images. Finally, in order to test the performance, the classification method proposed in this paper was compared with the SVM decision tree classification method in the experimental process. The experimental results show that the classification accuracy of remote sensing images with the proposed method is 95.82%, and has been further improved in comparison with other classification methods.
Keywords: remote sensing image classification; genetic algorithm; LVQ neural network; network training; performance testing; accuracy assessment
0 ?引 ?言
在現代圖像處理技術不斷發展的過程中,遠距離遙控圖像目標識別技術備受人們的重視,此技術被廣泛應用到工業、探測和軍事領域中。神經網絡分類方法被廣泛應用到圖像目標識別和遙感圖像分類過程中,但是此方法存在局部最佳解與識別效率比較低的問題。從70年代開始,遙感圖像分類都是遙感技術和相應領域學者所重視的問題。在現代遙感圖像分類過程中主要使用模糊數學分類法、統計模式識別法、語句法、模式識別法[1]。目前,人工神經網絡(ANN)在遙感圖像分類過程中廣泛使用,主要包括自組織特征映射、BP和ART等。在1990年,Kohonen提出了學習矢量量化(LVQ)算法,能夠實現聚類中心的監督和學習,也能夠使此中心代表數據都歸類到中心所屬類別中。……