摘要
维数约简和特征提取是模式识别中的一个重要预处理步骤。由于同一架飞机目标在各种不同的空间变换(包括平移、尺度、旋转等变换)和不同的观察角度、位置以及光照等条件下的图像之间差异较大,使得很多经典的维数约简和特征提取算法不能有效地用于基于飞机图像的飞机目标识别。判别局部保持映射(DLPP)是一种有效的监督维数约简方法,DLPP较LPP具有更好的分类能力,DLPP通过最大化样本的类间距离,同时最小化样本的类内距离来构建特征子空间。基于DLPP提出了一种飞机目标识别方法。实验结果表明,该方法是有效可行的。
Dimensionality reduction and feature extracting are important preprocessing steps in pat- tern recognition. Since the images of an aircraft target are much different from each other in the various conditions of the geographic transformations (such as rotation, translation, affine transform, etc) and the influence of different observed angle, locality and illumination in the real scene, many classical di- mensional reduction and feature extracting methods are not effective to recognize the aircraft target by aircraft images. The discriminant locality preserving projection (DLPP) algorithm is an effective super- vised dimensional reduction algorithm. The DLPP algorithm outperforms the LPP method, because it constructs the subspace by maximizing the between-class distance and minimizing the within-class dis- tance. In this paper, based on DLPP, an aircraft target recognition method was proposed. The recogni- tion results show that the proposed method is very effective and feasible.
出处
《计算机工程与科学》
CSCD
北大核心
2013年第6期118-122,共5页
Computer Engineering & Science
基金
河南省科技攻关计划资助项目(122102210429)
西亚斯国际学院引进人才资助项目(2012YJRC01
2012YJRC02)
关键词
飞机目标识别
判别局部保持映射
维数约简
aircraft object recognition
discriminant locality preserving projection
dimension reduction