基于无人机航拍输电线路杆塔的三维重建

发布时间：2018-06-29 05:17

本文选题：无人机 + 聚簇　；参考：《广西大学》2017年硕士论文

【摘要】：输电线路杆塔的三维重建研究是通过无人机采集输电线路杆塔的图像数据,然后利用计算机视觉技术进行三维重建,重建出表面致密、纹理丰富、可视化强的三维杆塔结构模型。本研究课题是基于电网的工程应用背景而设立,具体研究内容为:基于图像质量约束的无序图像关键帧提取和采用去抖动模糊算法的稠密三维重建。本课题研究不仅具有理论研究意义而且还具有实际应用价值:在理论研究层面,无人机采集的数字图像是无序图像,存在数目众多、信息冗余、质量参差不齐的问题。无人机采集大场景对象的图像进行三维重建时因抖动产生的模糊现象以及二维图像序列经运动恢复结构SFM(Structure From Motion)后得到的三维点云稀疏、可视化差的问题,都是我们急需解决的问题;在实际应用层面,无人机以轻便、灵巧、快捷的优势对输电线路进行巡检,可以弥补人工目测有死角、盲点的缺点,同时可以进入地形复杂、环境恶劣、巡检人员不宜进入的地方巡检,巡检效率高、安全系数高。巡检员可不去杆塔处去目测,直接在电脑面前对采集的杆塔图像信息进行立体视觉三维重建,使重建的三维杆塔效果逼真,细节丰富,可以满足巡检的视觉要求,本课题将有效的节省人力、物力,为电网的安全运行保驾护航。本课题包含基于图像质量约束的无序图像关键帧提取和采用去抖动模糊算法的稠密三维重建。具体研究内容如下:(1)本课题根据无人机巡检、图像采集的需求,建立四旋翼无人机图像采集系统。根据各子系统的结构功能对器件进行选型,搭建无人机输电线路图像采集平台,测试、验证平台的稳定性。(2)分析、对比无序图像的聚簇和关键帧提取算法:本课题首先采用不预设K-均值的聚簇算法对无序图像进行自动聚簇;其次根据相似距离从每簇中提取出离聚簇中心最近的一帧作为关键帧;最后采用二次模糊处理算法对提取的关键帧进行质量评价,其评价值若满足质量要求则保留,不满足则返回原来的簇中重新进行关键帧的提取与评价,直到提取的关键帧满足质量要求为止。(3)分析、对比航拍图像去抖动模糊算法和三维重建算法:本文首先采用去抖动模糊算法恢复模糊图像的原始图像信息,然后在运动恢复结构的基础上进行基于点云的稠密三维重建,最后对稠密重建后的点云进行泊松表面重建以得到表面致密、均匀、色彩丰富的三维杆塔模型。综上所述:基于无人机航拍输电线路杆塔的三维重建课题,提出基于图像质量约束的无序图像关键帧提取方法,提取出质量高、信息丰富的关键帧。提出采用去抖动模糊的稠密三维重建算法,重建出表面致密、纹理丰富、效果逼真的三维结构模型,因此本课题兼具理论研究意义与实际应用价值。
[Abstract]:The research of 3D reconstruction of transmission line tower is to collect the image data of transmission line tower by UAV, and then use computer vision technology to reconstruct the three dimensional transmission line tower. The reconstructed surface is compact and rich in texture. Strong visualization of three-dimensional tower structure model. This research is based on the engineering application background of power grid. The research contents are as follows: key frame extraction of disordered image based on image quality constraint and dense 3D reconstruction using jitter removing fuzzy algorithm. The research of this subject is not only of theoretical significance but also of practical application value: at the level of theoretical research, the digital images collected by UAV are disordered images, there are many problems, such as large number, information redundancy and uneven quality. The fuzzy phenomenon caused by jitter and the problem of sparse point cloud and poor visualization of 2D image sequence obtained by SFM (structure from Motion), which are used for 3D reconstruction of large scene object by UAV, are discussed in this paper. At the practical application level, unmanned aerial vehicles patrol transmission lines with the advantages of portability, dexterity, and speed, which can make up for the shortcomings of manual visual measurement of dead corners and blind spots, and at the same time, they can enter the complex terrain. The environment is abominable, the patrol personnel should not enter the place to patrol, the patrol inspection efficiency is high, the safety factor is high. Instead of going to the pole tower for visual measurement, the inspector can directly carry out three-dimensional visual reconstruction of the collected tower image information in front of the computer, making the reconstructed three-dimensional tower effect realistic and rich in detail, which can meet the visual requirements of the inspection tour. This subject will save manpower and material resources effectively, and guarantee the safe operation of power grid. This thesis includes key frame extraction based on image quality constraint and dense 3D reconstruction using jitter-removing fuzzy algorithm. The specific research contents are as follows: (1) according to the requirements of UAV inspection and image acquisition, a four-rotor UAV image acquisition system is established. According to the structure and function of each subsystem, the device is selected, the UAV transmission line image acquisition platform is set up, the stability of the platform is tested and verified. (2) Analysis, Compare the clustering algorithm of unordered image and the key frame extraction algorithm: firstly, we use the clustering algorithm without presupposing K- mean to cluster the unordered image automatically; Secondly, according to the similarity distance, the nearest frame of the cluster center is extracted from each cluster as the key frame. Finally, the second fuzzy algorithm is used to evaluate the quality of the extracted key frame, and the evaluation value is retained if the quality requirement is satisfied. If the key frame is not satisfied, the key frame will be extracted and evaluated again in the original cluster until the extracted key frame meets the quality requirement. (3) Analysis, Compared with the dejitter blur algorithm and 3D reconstruction algorithm of aerial image: firstly, the original image information of blur image is restored by de-jitter blur algorithm, and then dense 3D reconstruction based on point cloud is carried out on the basis of motion recovery structure. Finally, Poisson surface reconstruction is carried out on the dense point cloud to obtain a dense, uniform and colorful three-dimensional tower model. To sum up: based on the three-dimensional reconstruction of aerial photography transmission line tower, an image quality constraint based unordered image key frame extraction method is proposed to extract high-quality and informative key frames. A dense 3D reconstruction algorithm with dejitter ambiguity is proposed to reconstruct 3D structure model with dense surface rich texture and realistic effect. Therefore this paper has both theoretical research and practical application value.
【学位授予单位】：广西大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TM754

【参考文献】