Research On Virtual And Real Occlusion Technology In Augmented Reality

Augmented reality(AR)is to embed the virtual object generated by computer into the real scene to realize virtual and real fusion,which enhances the experience of reality visually,and is the upgrade and development of virtual reality(VR).In the AR system,virtual and real occlusion is the key to the effect of virtual and real fusion,which includes two steps:virtual real occlusion relationship determination and rendering.In this thesis,we mainly research the judgment of virtual and real occlusion relationship and rendering.Aiming at the problem of occlusion relationship,we use the structure from motion(SFM)algorithm to calculate the depth on the basis of camera translation scale.Through coordinate transformation,the depth values of virtual object and real object are converted to the world coordinate system for comparison,and this method realizes the virtual real occlusion judgment under monocular condition.To solve the problem of virtual and real rendering,we use the frame difference method(dynamic scene)and Canny algorithm(static scene)to extract the contour,and directly build the occlusion mask based on the contour,then virtual and real fusion rendering can be realized through Open GL template and depth cache.The improved occlusion algorithm is tested on two kinds of augmented reality scenes,marker and markerless.The results show that the algorithm can effectively solve the problem of virtual and real occlusion under monocular condition,and has good scene adaptability.Compared with other mask generation methods(filter method and direct depth method),the method of extracting contour directly can gain smaller MSE and higher SSIM,which shows that this method can obtain a higher accuracy mask,thus ensuring that the occlusion processing algorithm in this thesis can obtain a higher edge clarity occlusion effect.Compared with other occlusion processing algorithms(dynamic transformation frame method,etc.)in two kinds of AR scenes,the results show that the algorithm in this thesis has higher edge definition in occlusion effect and output frame rate in performance parameters.