Abstract: Current corner matching methods mainly focus on plane surfaces, where the chessboard patterns are only oblique. They will lose efficacy faced with warped surfaces. Aiming at this problem, this paper presents a novel method based on the Voronoi diagram. Experimental results show that compared with the other method, the new algorithm uses the distance between points to match the chessboard corners more quickly and get the final results better. This method makes the information of the points easier to save, and the computing cost is lower down.

Key words: augmented reality, geometry correction, corner matching, Voronoi diagram

摘要： 现有的角点检测方法多针对倾斜的平面投影情况，而对于投影在扭曲平面上的棋盘格则不能实现匹配。针对该问题，提出一种基于Voronoi图的匹配方法。实验结果证明，与一般方法相比，利用距离分类的方法能更快更好地实现角点匹配与几何矫正效果，且有易于存储、计算简便等优点。

关键词: 增强现实, 几何矫正, 角点匹配, Voronoi图

CLC Number: 

• TP391