Online classification for SAR target recognition based on SVM and approximate convex hull vertices selection
Shuguang Ding, Xiangli Nie, Hong Qiao, Bo Zhang · 2014
In this paper, we propose a novel online learning classification algorithm, which is based on support vector machine (SVM) and approximate convex hull vertices selection. Considering the geometrical features of SVM, we can safely delete the samples inside the convex hulls. However, in general, if the dimension of the training data is high, most of the samples are the convex hull vertices and therefore only a few samples can be deleted. To solve this problem, we adopt two steps: (1) reducing the dimension by principal component analysis (PCA); (2) finding an approximate convex hull, which is the main contribution of this paper. The effectiveness of the proposed algorithm is demonstrated through several experiments on synthetic data and MSTAR data sets.