地面无人平台视觉导航定位技术研究  

Research on Vision-based Navigation and Positioning Technology of Unmanned Ground Vehicle

作　　者：付梦印[1,2] 宋文杰[1] 杨毅[1] 王美玲[1] FU Meng-yin;SONG Wen-jie;YANG Yi;WANG Mei-ling(School of Automation,Beijing Institute of Technology,Beijing 100081,China;Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]北京理工大学,北京100081 [2]南京理工大学,南京210094

出　　处：《导航定位与授时》2019年第4期1-11,I0002,共12页Navigation Positioning and Timing

基　　金：国家自然科学基金(61105092,61473042);长江学者计划(IRT-16R06);教育部创新团队(T2014224)

摘　　要：地面无人平台实现自动驾驶等功能的核心难题在于如何感知其所处环境并获得自身在环境中的实时状态。机器视觉作为地面无人平台在复杂环境下实现自主导航及定位的重要手段,近年来得到了快速发展。系统性地分析了视觉定位与地图构建系统的基本架构,并对该架构包含的图像信息预处理、视觉里程计、回环检测、全局优化和地图构建模块分别进行了详细介绍。针对各模块所涉及的关键技术,总结了近些年来国内外主流的研究成果,对比分析了各个关键技术中主流方法的性能,并展望了地面无人平台视觉导航定位技术的发展方向。The core challenge for unmanned ground vehicle (UGV) to realize autonomous driving and other functions is how to perceive the environment and obtain its real-time status in the environment.Machine vision,as an important means of autonomous navigation and positioning for unmanned ground vehicle in complex environments,has developed rapidly in recent years.This paper systematically analyzes the basic architecture of visual positioning and mapping system,and introduces image information preprocessing,visual odometer,loop detection,global optimization,map building and other modules contained in this architecture in detail.Aiming at the key technologies involved in each module,the main research achievements at home and abroad in recent years are summarized.The advantages and disadvantages of the main methods in each key technology are compared and analyzed,and the development direction of vision-based navigation and positioning technology for unmanned ground vehicle is prospected.

关 键 词：地面无人平台 自动驾驶 机器视觉 自主导航及定位 环境感知 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]