基于ORB特征的复杂场景下的增强现实

ORB-feature Based Augmented Reality under Complex Scenes

针对增强现实开发工具ARtoolkit利用特定编码的卡片来获取目标,并且利用标记卡片角点来推算出相机的空间位置,准确率不高,有严格的观测角度限制,且不能在复杂场景使用的问题。在采用即时定位与地图构建和随机抽样一致算法的基础上,研究了如何将利用即时空间定位算法准确的计算出相机的外参和图像序列ORB特征点的空间坐标集,并利用随机抽样一致和poster帧得到二维感兴趣区域特征点坐标集,实现了感兴趣区域的空间坐标定位。实验结果表明,该算法解决了标记卡片以及观测角度限制,提高了相机位置的准确率,提高程度达90%。为后期渲染3D目标实现增强现实提供了有力的保障。

Aiming to the development tools for Augmented Reality,ARtooll^it,and to obtain a particular coding region of interest,Marker Cards and the comer feature are used to calculate the position of the camera.But the accuracy rate is low,and viewing angle and scenes are stirctly limited.Based on the simultaneous localization and mapping and random sample consensus algorithm,this paper studies howto use instant accurately calculate the spatial coordinates of camera external parameters and image sequence ORB(omented fast and rotated brief)feature point set.Then random sampling and poster frame are used to obtain two dimensional coordinates of region of interest to locate.For these restrictions,a new method is proposed in this paper by integrating SLAM and RANSAC to obtain a particular coding region of interest and calculate the position of the camera.Experiments showthat the method can solve the limitation of viewing accuracy of calculate the position of the camera.The degree of improvement is up to 90%,providing a strong guarantee render 3D augmented reality.