Research On Recovery Method Of Occlusion Consistency For Augmented Reality Based On RGB-D Camera

Occlusion consistency analysis is the key step to realize augmented reality system.It ensures that virtual and real occlusion effect can make users feel the seamless integration of virtual objects and real world.It gives users the feeling that virtual content and reality are integral in nature.Calculating the depth of virtual objects and real scenes is the key factor ofocclusion.It is very important to obtain accurate three-dimensional information of real scene and compare it with the depth of virtual object In the existing occlusion analysis methods,it is difficult to restore the complete scene three-dimensional information based on image analysis.Scene-based reconstruction method requires high modeling conditions and has hardware dependence.In order to solve these problems,based on the analysis and research of the current occlusion consistency processing methods,and combined with the theory of stereo vision and three-dimensional reconstruction,an augmented reality occlusion consistency analysis method based on RGB-D information is proposed.Firstly,Kinect sensor is used to acquire color and depth information,calibrate the color camera and depth camera,repair the missing information of depth image,and build the mutual mapping mode between color camera coordinate system and depth camera coordinate system,as well as the alignment relationship between color image and depth image.Secondly,according to the mapping mode of depth camera and color camera,in order to obtain the exact position of virtual objects superimposed in real scene,a registration method of point cloud registration combined with color image feature matching is designed,which can obtain stable and accurate camera position and pose under the restriction of Kinect depth camera,and carry out virtual and real superposition.Then,accurate scene information is acquired by dynamic reconstruction of three-dimensional scene,scene depth is calculated by template caching technology,depth comparison and occlusion judgment are made between virtual object and real scene,and occlusion area is eliminated.Experiments show that this method has good real-time performance and accurate occlusion effect.