Abstract: This paper applies manifold learning method in aircraft image identification, and proposes an algorithm of aircraft identification based on improved Isometric Mapping(ISOMAP). For the characteristics of high-dimensional of aircraft images, it adopts an improved algorithm of ISOMAP to reduce the dimension. In the process of constructing nearest neighbor graph, it uses Procrustes distance to replace traditional Euclidean distance. Experimental results demonstrate that the algorithm can achieve high recognition rate.

Key words: manifold learning, aircraft image recognition, Grassmann manifold, Isometric Mapping(ISOMAP), Procrustes distance

摘要： 将流形学习方法应用于飞机图像识别中，提出一种基于改进等距映射(ISOMAP)的飞机识别算法。根据飞机图像数据的高维性质，采用改进的ISOMAP对数据进行降维，在构造近邻图的过程中，利用Procrustes距离取代传统的欧氏距离。仿真实验结果证明，该算法的的识别率较高。

关键词: 流形学习, 飞机图像识别, 格拉斯曼流形, 等距映射, Procrustes距离

CLC Number: 

• TP317.4