为实现高速鲁棒的户外增强现实系统,提出了一种新的基于图像匹配算法的解决方案,同时描述了系统框架和工作原理,并在景深远小于物距的近似条件下,提出了一种基于全局视频信息的注册方法,即将环境注册问题通过基于快速傅里叶变换的图像匹配算法来实现。该方法先对当前摄像机摄入的图像和系统预存图像进行匹配;再将计算出的匹配信息对虚拟模型进行变换后叠加到当前的摄入图像上;最后输出到头盔显示设备,以实现户外增强现实。与已有方法相比,该方法具有运算量小、精度高、速度快的优点,试验结果证明了理论推导的正确性和该算法应用于数字圆明园重建的可行性。

A new solution for outdoor augmented reality is introduced in this paper and its architecture and principle are described.Under the condition that the depth of field is much shorter than the distance between camera and scene,a registration method based on panoramic-video information is presented,which is an FFT-based image matching method.The FFT-based method will work to match the camera-captured image with the pre-treatment image,and then according to the match information,the virtual model will be transformed and merged to the real scene.Finally the merged scene will be sent to Head-Mounted-Display to achieve augmented reality.Compared to the existing registration method,the proposed method possesses the advantages of low computing-cost,high accuracy and high speed.Experimental results show the correctness of theoretical analysis and the feasibility of its application in the digital reconstruction of Yuanmingyuan garden.