上海大学学报(自然科学版)

• 通信与信息工程 • 上一篇    下一篇

非线性相机模型对AR系统注册精度的改进

丁伟奇,吴家麒,袁政鹏,陈金波   

  1. 上海大学 机电工程与自动化学院,上海 200072
  • 收稿日期:2005-09-06 修回日期:1900-01-01 出版日期:1900-01-01 发布日期:1900-01-01
  • 通讯作者: 吴家麒

Improvement of AR Registration Accuracy with a Non-linear Camera Model

DING Wei-qi,WU Jia-qi,YUAN Zheng-peng,CHEN Jin-bo   

  1. School of Mechatronical Engineering and Automation, Shanghai University, Shanghai 200072, China

  • Received:2005-09-06 Revised:1900-01-01 Online:1900-01-01 Published:1900-01-01
  • Contact: WU Jia-qi

摘要:

注册是AR系统成功的关键技术之一.最近提出的采用基本矩阵的注册算法,不需要辅助的定位器材,也不需要人工标志物,易于执行,适应范围广,但注册精度有待改进.通过Tsai算法引入了非线性相机模型,对原有图像进行畸变校正,然后用基于基本矩阵的注册算法采用校正后的特征点进行计算.从实验所获得的结果来看,这种方法可以获得较原注册算法更好的精度.

关键词: Tsai算法, 增强现实, 注册方法

Abstract:

Registration is a key technique in an augmented reality (AR) system. A recent registration method based on fundamental matrix does not need auxiliary instruments and manmade markers, therefore is easily implemented and widely adaptable. However its registration accuracy needs improvement. In this paper the Tsai algorithm is used to introduce a non-linear camera model and correct distortion in the image feature points. The corrected feature points are then applied to the registration method based on a fundamental matrix. Experimental results show that the new registration technique provides better accuracy than the original method.

Key words: registration method, Tsai algorithm, augmented reality

中图分类号: