计算机应用

• 人工智能 • 上一篇    下一篇

基于分类的自动初始化VVS跟踪算法

孙敏 顾耀林   

  1. 江南大学信息工程学院 江苏省无锡市江南大学信息学院
  • 收稿日期:2008-01-09 修回日期:1900-01-01 发布日期:2008-07-01 出版日期:2008-07-01
  • 通讯作者: 孙敏

Classification-based automatic initialization VVS tracking algorithm

SUN Min Yao-Lin Gu   

  • Received:2008-01-09 Revised:1900-01-01 Online:2008-07-01 Published:2008-07-01
  • Contact: SUN Min 孙敏

摘要: 自动设定摄像机初始值成为增强现实领域中跟踪技术的热点问题。文中在虚拟视觉伺服(VVS)跟踪算法基础上利用Randomized Tree分类技术,将特征点匹配作为分类问题处理,再利用线性位置算法计算出摄像机的初始位置。通过实验验证了该方法的快速准确性和有效性。

Abstract: Automatic initialization becomes the hot issue in the tracking of augmented reality. The paper based on Virtual Visual Servoing (VVS) tracking algorithm used randomized trees as the classification technique and treated the feature point matching as a classification problem. Then, a linear pose computation method was used to estimate the initial camera position. Finally, the fast veracity was verified and the method had a better performance.