基于ORB自然特征的AR实时系统实现

针对当前基于自然特征的增强现实效率低,提出一种新的基于ORB自然特征的实时注册方法。提取视频帧与基准图像中的ORB特征点,使用汉明距离匹配,利用RANSAC算法筛选得到最佳匹配点对,确定摄像机的位姿。将三维虚拟物体叠加到真实场景中,达到虚实结合的效果。实验表明,在不同尺度角度、一定环境光变化、复杂背景和基准图像部分遮挡的情况下,该AR系统都具有良好的性能,跟踪定位准确度高,速度基本达到实时要求。

Real-time augmented reality based on ORB

Now augmented reality based on natural features is not efficient. This paper proposed a new method of real-time AR registration based on ORB features. First it extracted ORB natural features from video frames and a reference image, Selected matching features using appropriate Hamming distance and RANSAC algorithm. Then it calculated and determined the camera posed. Finally, superimpose the computer-generated three-dimensional virtual objects on the real-life scene to achieve combination of virtual and real. Experiments show that, in the case of different scales, different angles, some changes in ambient light, complex backgrounds or part of the reference image blocked, the AR system performs well to track and position accurately. And the processing speed of the system approximately meets the requirement of real-time system.