摘要
针对传统的基于增强现实系统中自然特征识别技术数据量大、计算耗时长的问题,提出一种改进SURF算法的虚实场景注册融合方法并运用在东巴古迹展现中。利用FAST检测图像的角点,用SURF算法进行增强现实系统中图像自然特征的提取,建立图像的特征描述。通过对东巴古迹的跟踪注册过程中图像特征描述符的匹配,使用RANSAC算法和最小二乘匹配提高匹配率,实现虚拟信息的注册;同时将虚拟信息在真实场景中叠加,用于完成对东巴文字、图像等古迹的展现过程。在对东巴古迹的增强现实实验中表明:利用该算法进行虚实注册,既能满足匹配准确性要求,又具有计算量小、计算速度快的优点,有较好的鲁棒性和实时性。
Aiming at the problem that the technical data amount of natural feature recognition in the augmented reality system is large, and, computing time is long, and etc, an improved SURF algorithm for virtual and real scene registration fusion method was proposed. Using FAST corner detection image, the natural characteristics of image with SURF algorithm were extracted, to establish the feature description. By matching the image feature descriptor in tracking registration process of Dongba monuments, we used the RANSAC algorithm and least squares matching to improve the matching rate, and realize the virtual information registration. At the same time, the virtual information was superimposed on the real scene, to show the Dongba text, images and other monuments. The experimental results show that the registration has good robustness and real -time performance by using this algorithm.
出处
《计算机仿真》
CSCD
北大核心
2015年第6期407-411,共5页
Computer Simulation
基金
北京市属高等学校创新团队建设与教师职业发展计划项目(IDHT20130519)
关键词
增强现实
特征匹配
东巴古迹
Augmented reality
Feature matching
Dongba monuments
