多尺度全覆盖视觉导航方法

MULTISCALE AND FULLVIEW VISUAL NAVIGATION

本文提出面向任务的多尺度、全覆盖视野的综合视觉建模方法，为自然环境下的视觉导航问题提供了一个系统的解决方案：将时域大尺度的时空远视，空域大尺度的全局环视，和小尺度的双目注视相结合，综合完成机器人行驶方向确定、实时路面障碍物检测和机器人全局定位等视觉任务，其目标是建立面向任务的高效的环境模型，提高视觉处理的效率和鲁棒性．

A taskoriented, multiscale and fullview visual modeling strategy is proposed for the natural environment. It combines panoramic vision for scene modeling, omnidirectional vision for road understanding and binocular vision for obstacle detection into an integrated system. This approach overcomes the drawbacks of traditional visual navigation methods that mainly depended on local and/or single view visual information. In this paper integrated sensor design, data processing and fusion, and visual navigation strategy are closely explored.