基于谱分解的F-S最佳鉴别平面及舰船识别研究

Optimal F-S Discriminant Plane Based on Spectral Decomposition and Study on Ship Recognition

Fisher最佳鉴别分析方法已在许多模式识别问题中取得成功应用。Fisher最佳鉴别分析建立在对Fisher最佳鉴别准则的最优化基础上。本文利用对类内矩阵 Sw进行谱分解 ,提出一种在类内矩阵 Sw 的零空间中求解F -S最佳鉴别平面的新方法。我们将此方法应用于红外舰船图象的特征抽取和识别的研究。实验结果表明了该方法的有效性。

Fisher's linear discriminant analysis methods(LDA)were demonstrated their success in the area of pattern recognition.LDA is based on the optimum of Fisher discriminant criterion.In this paper,a new LDA method is presented,which solves the optimal discriminant plane in the null space of within-class scatter matrix using the spectral decomposition of .The experiments were made for the feature extraction and recognition of infrared images of ship.The experimental results show the effectiveness of the new LDA algorithm.