中国机械工程 ›› 2010, Vol. 21 ›› Issue (21): 2541-2545.

• 机械基础工程 • 上一篇    下一篇

基于改进的SUSAN算子的棋盘格亚像素角点检测算法

杨幸芳1,2;黄玉美1;李艳1;高峰1
  

  1. 1.西安理工大学,西安,710048
    2.西安工程大学,西安,710048
  • 出版日期:2010-11-10 发布日期:2010-11-12
  • 基金资助:
    国家科技重大专项(2009ZX04014-032,2009ZX04001-065)
    National Science and Technology Major Project ( No. 2009ZX04014-032,2009ZX04001-065)

Sub-pixel Corner Detection Algorithm of Chessboard Image Based on Improved SUSAN Operator

Yang Xingfang1,2;Huang Yumei1;Li Yan1;Gao Feng1   

  1.  
    1.Xi’an University of Technology,Xi’an, 710048
    2.Xi’an Polytechnic University,Xi’an, 710048
  • Online:2010-11-10 Published:2010-11-12
  • Supported by:
    National Science and Technology Major Project ( No. 2009ZX04014-032,2009ZX04001-065)

摘要:

针对SUSAN算子在检测棋盘格角点这一特殊类型角点时出现的不足,提出了改进的SUSAN角点检测算子。改进的算子在分析原有算子仅考虑USAN区域的大小而不考虑USAN区域的形状这一局限性的基础上,通过在SUSAN圆模板内附加两个圆环模板来实现对所有棋盘格角点的有效检测。此外,鉴于棋盘格角点图像形状和灰度分布具有对称性以及灰度对比强烈的特点,采用灰度重心法对角点进行了亚像素定位。为了提高算法性能,具体操作时采用区域分离思想,将USAN区域分成内部像素区域和边缘像素区域两部分,通过对内部像素区域灰度的均值化,达到了抑制内部像素区域噪声的目的。实验结果表明,所提出的算法准确、有效,能达到亚像素级的精度。

关键词:

Abstract:

An improved SUSAN corner detection operator was proposed herein aiming at SUSAN operator’s deficiency revealed in detecting chessboard corners. On the basis of the limitation analysis about the SUSAN operator, the authors pointed out that one of the disadvantages of the SUSAN operator is that it just takes into account USAN’s size without considering USAN’s shape in detecting feature points and which may result in incorrect detecting results. Therefore, two ring-shaped masks were added within the SUSAN circular mask to overcome this defect. Moreover,in view of chessboard image has sharp brightness contrast, its shape is symmetric and its brightness distribution is symmetric, the authors utilized the center of the gravity of brightness to position chessboard corners in order to achieve sub-pixel precision. In practical operation, the USAN was separated into two regions, that is, the internal pixel region and the external pixel region and the algorithm performance was enhanced by restraining the noise of the internal pixel region by means of averaging the brightness of the internal pixel region. Experiments show that the proposed algorithm can implement effective detection of all chessboard corners and can achieve sub-pixel precision.

Key words:

中图分类号: