基于北斗的低成本无人机倾斜摄影系统的设计与实现

发布时间：2018-06-26 05:39

本文选题：低成本无人机 + 北斗定位　；参考：《上海海洋大学》2017年硕士论文

【摘要】：低成本、微小型无人机由于其可定制度高,容易实现拆卸折叠运输、容易实现弹射起飞加伞降或垂直起降,不依赖机场跑道或受场地局限等特点,越来越多地应用于市政规划、农业用地测量等需要即时获取地图及地理信息的工作中。传统的航空摄影遥感通过垂直安装在机身上的相机,进行拍摄,摄影过程中相机拍摄的图像基本控制在与地面垂直,该拍摄模式获取的影像进行几何纠正可以获取较高的几何定位,从而能够生成高精度的正摄影像。但与此同时,在对地物进行三维重建时,正射影像图仅拍摄到了地物的顶面纹理信息,然而侧面由于缺乏侧面的纹理信息,难以建立完整的地物三维地表模型,特别是由于部分遮挡影响生成的三模模型存在空洞。无人机倾斜摄影是一种通过在同一无人机平台搭载两个或多个不同倾角组合的摄影平台,可以同时获取地物的顶部纹理和侧边纹理信息,通过严格的多视几何关系以及立体匹配算法,可以自动重建地物的三维地表模型,降低三维重建的外业工作量,提升作业速度,并且能在较大范围区域推广使用。但是低成本无人机由于自身运载能力、电池容量、结构强度等条件的限制,在进行倾斜摄影时需要进行取舍。问题主要归纳为:首先是GNSS定位精度方面,无人机上搭载的GNSS设备在一定间隔时间生成的POS数据包含了无人机的三维坐标和姿态数据,该数据对地物三维重建的精度具有重要的影响,但在复杂环境如山区中,无人机航拍作业的飞行高度上会有信号遮挡,单一 GNSS定位精度不足。因此,单一的GNSS信号源难以满足无人机倾斜摄影在复杂环境下的适用性,通过组合北斗、GPS等不同GNSS定位信号,可以提升无人机倾斜摄影的导航定位精度,并且更好的进行航线规划。其次,倾斜摄影产生大量的影像,如何进行快速准确的地物三维重建需要构建无地面控制条件下基于POS数据的无人机倾斜摄影的密集匹配方法。最后是无人机续航能力方面,由于低成本无人机载重有限,在需要搭载成像载荷的条件下难以搭载很大的电池,因此其续航能力往往难以满足无人机倾斜摄影需要的长航时需求。针对上述问题,本文设计并实现了了一种基于低成本材料、续航能力较强、三维重建精度较高的无人机倾斜摄影系统。主要工作如下:1)研究了无人机姿态稳定性问题,提出利用扩展卡尔曼滤波(EKF)方法对两组IMU单元和两组地磁传感获取的惯性数据进行降噪并融合处理,通过航姿航向解算,提升了无人机的稳定性,获取了高精度的姿态数据,为后续的定位精度提高提供了保障。2)研究了无人机在线北斗/GPS组合导航技术,该方法通过优选北斗与GPS的可视卫星来实现组合定位,该方法在信号不佳的情况下也能达到较高的定位精度。3)研究了基于无人机倾斜摄影的地物三维重建技术,该方法首先通过SIFT特征提取方法生成大量特征点,在此基础上通过KD-TREE匹配方法对立体相对进行匹配从而自动建立相对的连接点,自动获取倾斜影像的连接点,利用多视几何关系重建相机内方位元素和外方参数。然后利用散列图像聚簇CMVS (cluster multi-view stereo)对序列图像组进行聚簇,以减少重建过程的数据量,提高运算速度,最后通过基于贴片模型的密集匹配PMVS (patch-based multi-view stereo)通过匹配、膨胀、过滤三个主要步骤完成最终的密集匹配。4)设计以及开发了一款低成本固定翼无人机,机身以泡沫为基底材料辅以碳纤维杆加强,动力采用电机驱动的单个螺旋桨。搭载了一个双相机的倾斜摄影的载荷,以及用于控制曝光同步的单片机控制器。利用固定翼机型较高的飞行效率以获取更长航时。5)通过两次试飞调整飞控参数与机身重心,提升了无人机的稳定性和续航能力。整机成本控制在一万以内,地面站部分的其它器材成本在五千以内;根据试验航程时间及试验后电池剩余电量的估算,机体作业飞行时间约1小时30分钟,作业航程约70公里。达到了预期的设计目标。本研究开发的无人机倾斜摄影系统,可以在较大范围内实现对地物进行倾斜摄影,并利用该摄影结果进行三维重建,可以推广应用在土木建设监管、市政规划、农业用地测量等需要即时获取地图及地理信息同时可以接受一定的建模、建图误差的地理信息相关工作中。
[Abstract]:Low cost, miniaturized UAV is easy to realize disassembly and folding transportation because of its high fixed system. It is easy to realize ejection takeoff plus parachute drop or vertical landing, and is not dependent on airport runway or site limitation. It is more and more applied to municipal planning, agricultural land measurement and so on. It is necessary to get maps and geographic information in time. Aerial photographic remote sensing is photographed by a camera mounted vertically on the fuselage. The image captured by the camera is basically controlled perpendicular to the ground during the process of photography. The image obtained by the shooting mode can be corrected by geometric correction to obtain high geometric positioning, thus generating a positive image of high degree of precision, but at the same time, the object is in the ground. In 3D reconstruction, the Orthophoto Image only takes the top surface texture information of the ground objects. However, the side due to the lack of side texture information, it is difficult to establish a complete 3D surface model of ground objects, especially the three model model, which is generated by the partial occlusion, is empty. The top texture and side texture information of the ground objects can be obtained at the same time with two or more different dip angles. The 3D surface model of the ground objects can be reconstructed automatically by the strict multi view geometric relationship and the stereo matching algorithm, which can reduce the work volume of the 3D reconstruction, improve the operation speed, and can be in a large scope. But low cost UAV, due to its own carrying capacity, battery capacity, structural strength and other conditions, needs to be taken out during the tilt photography. The main problem is: first, the GNSS positioning accuracy, the POS data generated by the GNSS equipment on the UAV at a certain interval contains the UAV. The 3D coordinate and attitude data have important influence on the accuracy of the 3D reconstruction of the ground objects, but in the complex environment such as the mountain area, the flight height of the UAV aerial aerial work will have signal occlusion and the single GNSS positioning accuracy is insufficient. Therefore, the single GNSS signal source is difficult to meet the suitability of the unmanned aerial vehicle in the complex environment. By combining the different GNSS positioning signals such as the dipper and GPS, the navigation positioning accuracy of the unmanned aerial vehicle can be improved and the route planning is better. Secondly, the tilt photography produces a large number of images, and how to make rapid and accurate 3D reconstruction of the ground needs to build the unmanned aerial vehicle (UAV) based on the POS data under the condition of no ground control. In the end, the compact matching method of photography. Finally, because of the limited load of the low cost unmanned aerial vehicle (UAV), it is difficult to carry large batteries under the condition of carrying the imaging load, so the endurance of the unmanned aerial vehicle is often difficult to meet the long flight needs of the unmanned aerial vehicle (UAV). An unmanned aerial vehicle (UAV) tilt camera system based on low cost material, strong endurance and high precision of 3D reconstruction. The main work is as follows: 1) the attitude stability of unmanned aerial vehicle (UAV) is studied. The extended Calman filter (EKF) method is used to denoise and fuse the inertial data obtained from two groups of IMU units and two groups of geomagnetic sensing. Through the flight direction solution, the stability of the UAV is improved, the high precision attitude data is obtained, and the following positioning precision is guaranteed by.2). The integrated navigation technology of the unmanned aerial vehicle (UAV) online Beidou /GPS is studied. The method is based on the optimal selection of the visible satellite of the Beidou and GPS, and the method is under the condition of poor signal. It can also achieve high positioning accuracy.3). The 3D reconstruction technology based on unmanned aerial vehicle (UAV) tilt photography is studied. First, a large number of feature points are generated by the SIFT feature extraction method. On this basis, the stereo relative is matched by the KD-TREE matching method and the relative connection points are built automatically, and the inclined image is automatically obtained. In order to reduce the amount of data in the reconstruction process and improve the computing speed, a dense matching PMVS (patch-based multi-view stere) based on the patch model (patch-based multi-view stere) is used to reconstruct the azimuth and external parameters of the camera by using the multi view geometric relationship. Then the sequence image group is clustered with the hash image clustering CMVS (cluster multi-view stereo). O) a low cost fixed wing UAV is designed and developed by matching, expanding, filtering three main steps to complete the final dense matching.4). The fuselage is reinforced with a carbon fiber rod with foam as the base material and a single propeller driven by the motor. The load of a two camera tilt camera is carried and the aeration is used to control the aeration. Optical synchronous microcontroller controller. Using the higher flight efficiency of the fixed wing to obtain longer flight time.5), the flight control parameters and the body center of gravity are adjusted by two flight tests. The stability and endurance of the UAV are improved. The cost control of the whole machine is within ten thousand, the cost of other equipment in the ground station is less than five thousand. The time and the residual quantity of the battery after the test are estimated. The flight time of the body is about 1 hours and 30 minutes, and the operating range is about 70 kilometers. The designed target is achieved. The tilt photography system of the unmanned aerial vehicle (UAV) developed in this study can carry out the tilt photography in a large range and make use of the photographic results for three-dimensional reconstruction. The promotion and application in civil construction supervision, municipal planning, agricultural land use measurement and so on need immediate access to maps and geographic information at the same time to accept certain modeling, and to build mapping errors in the geographical information related work.
【学位授予单位】：上海海洋大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P231;TN967.1

【参考文献】