AR High Precision Real-time Registration Core Technology And Application System Development

Along With the development of information technology,Augmented Reality(AR)has become a hotspot in the field of computer vision and computer graphics in recent years.Stable and accurate registration is crucial in markerless Augmented Reality.In this thesis,for the crucial technologies in markerless Augmented Reality,we introduce Visual Simultaneous Localization and Mapping(V-SLAM)technology,which will directly match the objects in the real scene and get more accurate position.The main contents are as follows:(1)In order to solve the problem of tracking instability based on Augmented Reality of two-dimensional image markers,this thesis presents an Augmented Reality system based on sparse map by using V-SLAM method.The Augmented Reality system has two key parameters: One is the pose of the camera and the other is the position of the virtual object.The former can be completed by V-SLAM,the latter by using 3D point cloud and the method based on Principal Component Analysis(PCA)to detect the plane.This method does not need to set the threshold.Therefore,it is better than method which is based on Random Sample Consensus(RANSAC),which can make the system run faster.In the end,the whole system is experimentation,which proves that the method based on the three-dimensional environment is more stable and effective.(2)For augmented reality based on sparse maps,the shortcomings of the plane can only be identified.In this thesis,we use Kinect camera to acquire RGB-D data and reconstruct the dense scene by V-SLAM system.Then use Hough Voting algorithm to identify any object in the scene,which can get the location of the virtual object.After that,this thesis proposes an augmented reality system based on dense map.The experimental results show that the system can effectively identify objects and the virtual can superimpose in the real environment.