基于立体视觉的无人机感知与规避研究

发布时间：2018-10-17 17:24

【摘要】：无人机(Unmanned Aerial Vehicle,UAV)是目前国内外的研究热点。随着无人机的广泛使用,无人机的自主安全飞行提出了要求,无人机感知与规避(Sense and Avoid,SAA)能力是保证无人机进入空域飞行、保障无人机飞行安全的先决条件。轻小型无人机系统因载荷、任务、成本等限制,无法搭载大型感知设备如雷达、激光雷达等,基于视觉传感器的SAA系统在质量、功耗、成本方面的优势使其更易在轻小型系统集成和应用。本文针对轻小型无人机的视觉感觉与规避进行研究。考虑单目视觉只能得到目标的二维平面运动信息,本文采用基于立体视觉的感知与规避技术,实现目标精确感知、可靠威胁评估和障碍物规避,克服了单目视觉难以获取目标的深度信息的缺点。本文主要研究工作如下:1.从体系方案、政策标准、技术研究等方面分析了SAA。系统全面分析了SAA技术发展史,SAA技术研究史。分析了无人机SAA的意义以及各国对无人机SAA技术研究情况。对无人机SAA进行层次分解和技术分解,分析了无人机SAA方法。根据SAA技术100年发展史以及基于无人机SAA的研究现状,提出SAA技术下一步发展并对SAA的未来关键技术进行分析。2.围绕轻小型无人机的应用特点和能力属性出发,分析并搭建了一套低成本、低功耗感知与规避系统。针对小型无人机系统载荷小、飞行速度低等约束,从环境数据感知、机载数据处理、障碍规避控制等方面分别进行等硬件系统设计;针对感知与规避系统能力要求,对立体视觉感知系统进行建模和标定。3.针对低空无人机环境复杂多变的特点,设计了多模式无人机感知与规避算法。针对无人机飞行环境的不同,无人机在不同空域环境下不同的遭遇环境进行相应的建模。在低空环境,通过重构三维环境,采用UV视差法进行障碍物检测;在高空环境,背景相对简单、空域障碍物多为运动的、微小的、特征点少,采用背景分割与弱小目标检测算法进行障碍物检测并通过三角定位法对障碍物进行测距。采用安全包络法对运动目标进行威胁评估,根据威胁评估结果,采用目标规避点方法对障碍物进行规避。实验结果表明,基于立体视觉的无人机感知与规避系统能够实现不同空域环境下的障碍物避撞,保障无人机的飞行安全。
[Abstract]:Unmanned aerial vehicle (Unmanned Aerial Vehicle,UAV) is a research hotspot at home and abroad. With the wide use of UAV, the autonomous safety flight of UAV is required. The ability of UAV to perceive and evade (Sense and Avoid,SAA is the prerequisite to ensure the UAV to fly into airspace and to ensure the UAV flight safety. Because of the limitation of load, task, cost and so on, the light and small UAV system can not carry large sensing equipment such as radar, lidar and so on. The SAA system based on vision sensor has high quality and power consumption. Cost advantages make it easier to integrate and apply in light and small systems. In this paper, the visual perception and avoidance of light and small UAV are studied. Considering that monocular vision can only obtain two-dimensional planar motion information of the target, this paper uses stereo vision based perception and circumvention technology to realize target accurate perception, reliable threat assessment and obstacle avoidance. It overcomes the disadvantage that monocular vision is difficult to obtain depth information of target. The main work of this paper is as follows: 1. This paper analyzes SAA. from the aspects of system scheme, policy standard, technology research and so on. The development history of SAA technology and the history of SAA technology research are systematically analyzed. The significance of UAV SAA and the research of UAV SAA technology are analyzed. The SAA of UAV is decomposed by hierarchical decomposition and technical decomposition, and the SAA method of UAV is analyzed. According to the 100-year history of SAA technology and the research status of UAV SAA, this paper puts forward the next development of SAA technology and analyzes the key technology of SAA in the future. 2. Based on the application characteristics and capability attributes of light and small UAV, a set of low cost, low power sensing and evading system is analyzed and built. Aiming at the small UAV system with small load and low flight speed, the hardware system is designed from the aspects of environmental data perception, airborne data processing, obstacle avoidance control, etc. Modeling and calibration of stereo vision perception system. 3. In view of the complex and changeable environment of low altitude UAV, a multi-mode UAV perception and avoidance algorithm is designed. According to the different flight environment of UAV, the UAV has different encounter environment in different airspace environment. In low altitude environment, by reconstruction of 3D environment, UV parallax method is used to detect obstacles. In high altitude environment, the background is relatively simple, and the obstacles in airspace are mostly moving, small and few feature points. Background segmentation and small and weak target detection algorithms are used to detect obstacles and the distance of obstacles is measured by triangulation. The method of security envelope is used to evaluate the threat of moving target, and according to the result of threat assessment, the method of avoidance point is used to circumvent the obstacle. The experimental results show that the UAV perception and avoidance system based on stereo vision can avoid obstacles in different airspace environment and ensure the flight safety of UAV.
【学位授予单位】：西北工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：V279

【参考文献】