现代武器装备的高度复杂性,装备故障信息具有高度的非线性,这给装备技术保障带来新的挑战.针对这一问题,该文采用保局投影流行学习方法作为装备故障信号的特征提取器,将故障信号一类信息通过显式映射矩阵投影到低维空间,获得关于故障的特征信息;在此基础之上,利用特征信息训练SFAM网络形成一个故障诊断分类器,定位故障源及性质.仿真实验表明,该文提出的故障诊断技术方案对较高维数的故障信息具备较高的故障识别率及鲁棒性.%A ship recognition method based on manifold and Simplified Fuzzy ARTMAP network is proposed. Firstly, by using the geodesic approach to measure the distance between data, a manifold learning method based on the locality preserving projections (LPP) is introduced. Secondly, the supervised form of ISOLPP, termed as S-ISOLPP, is proposed to extract the features of ship image objects. Eventually, a network classifier with high approaching accuracy is trained with the S-ISOLPP features, and the classifier is used to classify the new samples. The simulating results show that ISOLPP can not only inherit the advantage of explicit projection matrixes but also solve the problem of the under fitting state in LPP method, and the adaptability of feature extraction is improved remarkably. Moreover, the proposed ship recognition method can accurately classify the ship image objects with high robustness; as shown in the example, the selection probability of the five sorts of ships is above 97.25%.
展开▼