虚拟自主汽车的动态几何视觉模型研究及实现

Research and Implement on Dynamic Geometry Vision Model of Virtual Autonomous Vehicle

虚拟自主汽车是驾驶员和汽车封装起来的“人-机系统”的几何、感知、行为等特性在虚拟空间中的数字化表示,研究虚拟自主汽车视觉感知模型对建立真实的自主汽车智能行为起到关键的作用。通过分析驾驶员的视觉特点和传统几何视觉模型不足的基础上,给出一种分层的、动态的虚拟自主汽车双眼立体几何视觉模型:引用车速过滤机制和注意力过滤机制对自主汽车的视觉局限性进行分层,引用注意熵对虚拟自主汽车的注意力进行自主定量分配,利用假记忆库模拟短时记忆信息沉淀到长时记忆信息中的学习记忆过程。然后利用假视野缩小视觉模型感知环境的范围。最后用EONStudio虚拟现实软件将此模型三维可视化。

Virtual Autonomous Vehicle is a digital representation of various characteristics （including geometry, perception, behavior etc.） of the man-machine system in virtual space. The man-machine system consists of both the driver and the automobile. Research on the visual model of Virtual Autonomous Vehicle can play an important role in the reality of intelligent behavior. Through analyzing the visual characteristics of drivers and the disadvantages of traditional geometry vision, a dynamic double eyes solid geometry vision model of VAV with hiberarchy was developed. The vision limitation was carried on with the velocity filter mechanism and attention filter mechanism. The attention of virtual autonomous vehicle was hypothesized with attention entropy. The study process of short-term memory information precipitated long-term memory information was simulated with false memory database. The false field of vision was used to reduce the field of the virtual environment. The three dimensional visible vision model of virtual autonomous vehicle was developed with EONStudio virtual reality software.