基于神经网络的场景调度策略

Scene scheduling strategy based on neural network

虚拟场景的实时显示是虚拟现实技术的重要研究课题,优化场景的调度策略是解决实时显示的主要方法之一.根据虚拟装配的特点,建立视点状态变化的马尔可夫模型,并且使用径向基函数神经网络挖掘装配过程中视点变化的规律,提出基于神经网络的场景调度策略,利用该策略预测并提前调入装配过程中将要出现的场景.该策略可以使调度更加智能化,提高场景调度的效率,达到实时显示场景的目的.同时该调度策略也可以应用于虚拟现实的其他领域中,如虚拟场景漫游等.

Real time display of the virtual scene is the important research topic of virtual reality, and the optimal scene scheduling strategy is one of the methods to solve this problem. According to the characteristics of the virtual assembly, this article established a Markov model of the change of viewpoint state, used the RBF neural network to dig the rules of viewpoint change during the assembly process, proposed scene scheduling strategy based on the neural network, and employed this strategy to preview and access the scene ahead of time which was going to display in the assembly process. This strategy intellectualizes the scheduling and raises the efficiency of the scene scheduling, so as to achieve the goal to display scene in real time. At the same time this scheduling strategy can be also applied to other domains in the virtual reality , like the scene roaming etc.