论文类型：期刊论文
发表时间：2010-06-01
发表刊物：Journal of Information and Computational Science
收录刊物：EI、Scopus
卷号：7
期号：6
页面范围：1261-1266
ISSN号：15487741
摘要：Principal curves are extensions of Principal Component Analysis and are smooth curves, which pass through the middle of a data set. We have smooth principal curves based on subdivision scheme which go through the centre of data sets. The core of the method is to iteratively subdivide the fold lines linked from notes of sparsified set in order, which came from the average of the current local projections of the data sets. The Principal curves are generated in three ways: by joining up the nodes in order, by fairing the fold lines and by subdividing the last fold line. The Principal Curves algorithm presented in the paper can gain connected skeleton from initial data set with arbitrary shapes, which is fit for handwriting numbers recognition very well. ? 2010 Binary Information Press.
