Research Of Virtual Walkthrough Technology For Image-based Modeling

With the development of information technology, the research of Virtual Reality technology (VR) is more and more mature. It gets more and more attention. VR is used to reproduce reality scene, show theoretical principles and concept image that the objective world does not exist so that people can achieve immersive feeling. The key of VR is that 3D virtual models with strong feeling of reality can be constructed fast and effectively. Its realization can be divided into two categories: traditional geometry-based modeling and rendering techniques and image-based modeling and rendering techniques (IBMR). With traditional geometry-based modeling and rendering techniques, the modeling process is more simple and faster. And the operation is also easier. This algorithm not only has high speed, but also ensures that the models are high-reality. The key technologies of IBMR in this paper are deeply studied and improved that are the image-based modeling process, the texture extraction, the correction of virtual walkthrough video and seamless splicing.TIP is a technology of IBMR that is used extensively at present. This paper is based on TIP. The method can be used to rebuild the 3D models of objects and scenes. Then virtual walkthrough video is received. In order to meet the realistic requirement of virtual walkthrough, the follow-up treatment on virtual walkthrough video is needed that include the correction of virtual walkthrough video and seamless splicing. For the realistic and real-time requirements of virtual walkthrough system the algorithm what is used to seamless splice virtual walkthrough video images must increase the splicing speed as high as possible under the assurance of splicing quality.The two issues in this paper are improved. For the correction of virtual walkthrough video, the correction of arc projection is improved after some careful study and experimentation which is a viewpoint-centered perspective projection. With other correction methods its advantages include simple correction, small computational amount, convenient operation and so on. An adaptive threshold video splicing algorithm based on distance is provided in this paper. The algorithm which is used to splice virtual walkthrough video images summarizes the two uppermost splicing algorithms at present that are a splicing algorithm based on feature points and a splicing algorithm based on image features. By extracting distance feature, it uses adaptive threshold SSDA (sequential similarity detection algorithms) algorithm and matches in estimating image, then estimating the beginning column of the overlapping part. Experimental results show the algorithm is wonderful and stable. It can also reduce workload and accelerate the splicing speed.