无人机导航中的立体视觉技术分析-analysis of stereo vision technology in uav navigation.docxVIP

南京航空航天大学硕士学位论文图5.3实验一中视差图计算和障碍物检测的结果·········································································50图5.4实验二中视差图计算和障碍物检测的结果·········································································51表清单表2.1摄像机MV-1300UM的性能参数指标9表2.2镜头M0814-MP的性能参数指标10表2.3立体匹配中标准测试样本的视差放大系数·········································································22表3.1本章方法与其它方法视差精度的量化对比·········································································32表3.2本章方法与其它方法的计算效率对比（计算teddy）························································32表4.1本章方法与其它方法视差精度的量化对比·········································································41表4.2实验中的不同光照和曝光条件·····························································································42表5.1障碍物检测实验中的参数设置·····························································································49表5.2障碍物三维深度信息的计算结果·························································································51vii注释表序号缩写英文全称中文全称1CSIROCommonwealthScientificandIndustrialResearchOrganization澳大利亚联邦科学与工业研究组织2GPSGlobalPositioningSystem全球定位系统3GPUGraphicsProcessingUnit图形处理单元4IMUInertialMeasurementUnit惯性测量装置5UAVUnmannedAerialVehicle无人驾驶飞机6VTOLVerticalTake-OffandLanding垂直起降承诺书本人声明所呈交的硕士学位论文是本人在导师指导下进行的研究工作及取得的研究成果。除了文中特别加以标注和致谢的地方外，论文中不包含其他人已经发表或撰写过的研究成果，也不包含为获得南京航空航天大学或其他教育机构的学位或证书而使用过的材料。本人授权南京航空航天大学可以将学位论文的全部或部分内容编入有关数据库进行检索，可以采用影印、缩印或扫描等复制手段保存、汇编学位论文。（保密的学位论文在解密后适用本承诺书）作者签名：日期：第一章绪论1.1课题研究背景和意义无人机(UAV)是无人驾驶飞机(UnmannedAerialVehicle)的简称，它广泛应用于军事、民用和科研三大领域：在军事上，可用于侦察监视、通信中继、电子对抗、火力制导、战果评估、骚扰诱惑、对地(海攻击)、目标模拟和早期预警等；在民用上，可用于大地测量、城市环境检测、地球资源勘测和森林防火、农业勘测、交通、民用导航、环境保护、边境巡逻与控制、自然灾害的监视与救援等；在科研上，可用于大气研究、气象观测、对核生化污染区的采样与监控、新技术新设备和新飞行器的试验验证等[1]。无人机自主导航是指无人机依赖机载的航空电子系统、飞行控制系统等设备实现无人机的状态和位置估计、路径规划以及控制到达或着陆的过程。传统的无人机导航研究主要是针对高空飞行状态下的无人机，但随着无人机在低空场景下的应用不断增加，如自然灾害条件下的人员搜救及低空勘察等，基于运动状态估计、障碍物检测和规避等关键技术的低空无人机自主导航技术逐渐成为了无人机导航领域的一个研究热点。无人机在低空条件下飞行存在各种干扰，难以获得可靠的GPS位置信号，为此需要增强无人机对未知动态环境的感知能力为合理的路径规划和准确的导航提供可靠丰富的环境信息。立体视觉技术作为计算机视觉的重要分支，不仅隐蔽性好、功耗低、信息量丰富，而且