| Summary: | 碩士 === 國立政治大學 === 資訊科學學系 === 90 === Walkthrough in a virtual environment enables a user to have realistic experiences in an unreachable scene. However, as the complexity and applications of virtual environments increase, it becomes an emerging issue to maintain the quality of a walkthrough experience. In this thesis, we propose an effective data management scheme to address this issue. First, we propose to use real-time scene management to reduce the amount of geometry data transmission and to manage the computing resources on the client side. Second, we propose a prioritized MLM(Most Likelihood Movement) model to prefetch possible future objects, based on users’ current motion intentions. In addition, this prefetching model has accounted for mouse control and influence of obstacles in walkthrough to increase prefetching accuracy. Lastly, a hybrid coherence cache model is proposed to take advantages of both the temporal and spatial localities of walkthrough. We have implemented the front-end and back-end systems for the proposed scheme and done extensive experiments to demonstrate how these techniques can improve the effectiveness of walkthrough in a large virtual environment.
|
|---|
