计算机应用 ›› 2011, Vol. 31 ›› Issue (05): 1202-1204.DOI: 10.3724/SP.J.1087.2011.01202

• 图形图像技术 • 上一篇    下一篇

基于主元分析和聚类的直线检测算法

刘巍1,金文标1,2,肖仙谦2   

  1. 1.重庆邮电大学 计算机科学与技术学院,重庆400065
    2.杭州电子科技大学 理学院, 杭州310018
  • 收稿日期:2010-11-01 修回日期:2010-12-21 发布日期:2011-05-01 出版日期:2011-05-01
  • 通讯作者: 刘巍
  • 作者简介:刘巍(1983-),男,湖南邵阳人,硕士研究生,主要研究方向:数字图像处理;金文标(1964-),男,浙江浦江人,副教授,博士,主要研究方向:数字图像处理;肖仙谦(1988-),男,浙江杭州人,主要研究方向:数字图像处理。
  • 基金资助:

    杭州电子科技大学科研项目科研启动基金资助项目(KYS075609066)。

Approach for line detection based on principal component analysis and clustering

LIU Wei1, JIN Wen-biao1,2, XIAO Xian-qian2   

  1. 1. College of Computer Science and Technology, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
    2. College of Science, Hangzhou Dianzi University, Hangzhou Zhejiang 310018, China
  • Received:2010-11-01 Revised:2010-12-21 Online:2011-05-01 Published:2011-05-01
  • Contact: LIU Wei

摘要: 针对现有的直线检测算法中,基于霍夫变换类算法开销大且易产生虚假结果,基于链码跟踪类方法鲁棒性和适应性较差的问题,提出一种新的直线检测算法。对边缘图像做分块链码跟踪产生链码串,然后对链码串做主元分析(PCA)构造线段,最后采用聚类方法合并线段以产生直线。实验结果表明,该算法速度较快,检测结果较理想,且对较复杂、细节丰富的图像也具有良好的检测结果。

关键词: 直线检测, 主元分析, 链码跟踪, 线段, 聚类

Abstract: In the existing line detection methods, those based on Hough Transformation (HT) have a huge cost and always bring false results, and others based on chain tracing are weak on robustness and adaptability. This paper proposed a new approach for line detection, in which, the chain was generated by chain tracing in edge image block by block, then the Principal Component Analysis (PCA) was used on the chain to construct segments, at last the lines were got by merging the segments through clustering. The experimental results show the approach is fast and gives good results, and especially it performs well in highly complex and detail rich images.

Key words: line detection, Principal Component Analysis (PCA), chain tracing, segment, cluster