作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2008, Vol. 34 ›› Issue (23): 202-204. doi: 10.3969/j.issn.1000-3428.2008.23.072

• 人工智能及识别技术 • 上一篇    下一篇

基于特征点的印品图像匹配算法

张二虎,冯 江   

  1. (西安理工大学信息科学系,西安 710048)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2008-12-05 发布日期:2008-12-05

Printed Matter Image Matching Algorithm Based on Feature Points

ZHANG Er-hu, FENG Jiang   

  1. (Department of Information Science, Xi’an University of Technology, Xi’an 710048)
  • Received:1900-01-01 Revised:1900-01-01 Online:2008-12-05 Published:2008-12-05

摘要: 高速精准的图像匹配为成功地进行印品缺陷检测奠定了基础。该文提出一种改进的Harris角点检测方法,用一种稳定性评价准则测试,证明该文算子在工业环境下的优越性。建立仿射变换模型来近似相应特征点邻域的几何变换,用确定性退火方法计算模型参数,避免了耗时的穷举搜索。利用RANSAC方法鲁棒地估计基本矩阵和单应矩阵,建立对极几何约束和单应约束来剔除初始匹配中的误匹配对。算法处理速度较快,已经成功运用到印品缺陷在线检测系统中。

关键词: 特征点, 图像匹配, 对极几何, 单应

Abstract: Image matching with fast speed and high accuracy lay the foundation of successful defect detection system for printed matter. An improved Harris corner detection method is proposed and testified by a stability evaluation criterion, where the proposed detector is found to perform well in industrial environment. An affine transform model is established to approximate the geometrical changes between two windows of corresponding feature. The parameters of the model are computed by deterministic annealing to avoid time-consuming exhaustive searches. The fundamental matrix and homography matrix are estimated robustly with RANSAC algorithm, then epipolar and homography constrains are used to delete the wrong correspondence of initial matching. The algorithm with fast speed has been successfully utilized in the on-line defect detecting system for printed matter.

Key words: feature point, image matching, epipolar, homography

中图分类号: