【作者】 张勇；

【导师】 袁修孝；

【作者基本信息】 武汉大学 ， 摄影测量与遥感， 2017， 博士

【摘要】 超高压输电网络中的电力线没有绝缘层,以空气作为绝缘体,为了保证输电线路的安全运行,需要确保电力线周边一定范围为纯净空间,不能存在导电物体。输电电网投入运行后,电力线通道内的植被会自然生长,当植被与电力线之间的距离小于安全阈值时,可能会引起放电,危及电网的安全运行。因此,关于大范围、高效率的电力线通道障碍物巡检方法的研究受到了广泛关注。借助无人机成本低廉、操作方便的优势,通过在无人机上安装可见光数码相机获取电力线通道内的影像,进而检测潜在的危险物体,正在成为一种新的电力线通道障碍物巡检作业方法。无人机在执行电力线通道障碍物巡检作业时,沿电力线方向飞行,按照一定重叠度拍摄通道内的立体影像,通过摄影测量方法可以完成电力线和地表的三维重建,在已知地表形态和电力线空间位置的前提下,计算地面点云和导线的空间距离可自动识别和定位障碍物。电力线通道地表的三维重建、导线的三维重建以及障碍物自动检测是利用无人机影像进行线路通道障碍物巡检的关键问题。对于电力线通道地表三维重建问题,采用影像稠密匹配算法提取三维点云是一种切实可行的方法,但是由于电力线通道地表经常被密集植被所覆盖,给影像匹配带来了较大的困难,并且电力线通道内也可能存在违章建筑,因此影像匹配方法需要同时满足树冠和屋顶等典型匹配困难目标的三维重建需要;对于电力线三维重建问题,由于电力线在影像上表现为一条细线,并且纹理单一,采用常规的影像匹配方法很难获取电力线上的同名像点,无法实现电力线三维坐标的自动测量,采用人工立体测量的方法虽然能够实现电力线三维重建,但是费时费力;利用数字摄影测量方法完成电力线通道三维重建之后,障碍物的检测问题可以通过计算导线和地物间的空间距离来实现。因此,电力线通道障碍物的检测和定量描述的关键在于电力线通道的三维重建,有必要开展关于电力线通道以及导线三维重建方法的研究,实现基于无人机影像的电力线通道障碍物自动化定量巡检。本文使用无人机影像,以电力线通道障碍物的定位和定量评价为目标,研究了电力线通道地表稠密点云提取方法、电力线三维坐标自动测量方法以及障碍物检测方法。具体研究内容和创新性成果如下:(1)基于SPMEC稠密匹配算法的电力线通道地表三维重建方法研究在研究影像匹配算法的基础上,提出了 SPMEC(Semi-Patch Matching based on Epipolar Constrains)影像密集匹配方法。首先,将立体像对中的相对定向点变换到核线影像上,构建初始视差三角网。然后,使用较大的匹配窗口沿核线进行一维影像搜索,对初始视差三角网进行精化,生成等间距的密集视差栅格。接着,在影像分割结果的基础上,构建匹配面元,进而获取稠密同名像点。最后,使用粗差剔除算法过滤误匹配点,并输出匹配结果。实验结果表明,SPMEC算法能够提取到稠密的三维点云,实现了通道地表的三维重建,特别是对树冠、屋顶等电力线通道内的常见地物具有较好的三维重建能力。当匹配窗口在搜索窗口内滑动过程时,随着匹配窗口接近和远离同名像点,相关系数会呈现从小到大再从大到小的变化规律,相关系数曲线上存在明显的孤峰,且峰值两侧近似对称。SPMEC算法充分利用了这一现象,将匹配窗口和搜索窗口间的归一化相关系数作为匹配测度,提出了顾及相关系数曲线形状的相似性判据。与传统相关系数阈值法判据相比,明显提高了匹配窗口位于阴影、树冠等匹配困难区域时的匹配成功率和可靠性。待匹配点视差与其邻域内已匹配点视差间的相容性是影像稠密算法的另一重要方面。影像上纹理相似的区域所对应的地面物体一般同属一个类别,视差变化通常也是连续的。SPMEC假定同一影像分割对象内,视差是连续的,通过影像分割算法,为匹配过程引入邻域相容性约束条件,同时顾及了地面高度突变引起的视差不连续问题,提高了影像匹配算法在视差不连续区域的稳定性。(2)基于PLAMEC电力线自动测量算法的导线三维重建方法研究基于立体模型中同名像点必然位于同名核线的原理,提出了电力线自动测量方法 PLAMEC(Power Line Automatic Measurement method based on Epipolar Constrains)。首先,从无人机拍摄的两条航线中选取电力线两侧对应的影像构成立体像对,与同一条航线内相邻影像构成的立体像对不同,电力线在由航线间影像构成的立体像对中,其方向与核线方向近似垂直,无论对电力线人工立体测量还是对电力线自动测量,这样的立体模型构成方式都是十分必要的。然后,研究了电力线在影像上的灰度特征和几何特征,提出了基于灰度比值的电力线特征检测算子,结合数学形态学和电力线先验空间拓扑关系,分别从立体像对的左右核线影像上提取电力线二维矢量。接着,在核线影像的上下视差方向,每隔一定距离取一对同名核线,计算它们与电力线二维矢量中同名电力线的交点,这对交点就是位于电力线上的同名像点。最后,使用抛物线拟合算法,将交点坐标由单像素精度提高至亚像元精度后,经前方交会可得这对同名像点对应的物方点坐标。实验结果表明,电力线自动测量的成功率为93.2%,测量精度与人工测量结果一致,中误差优于±0.15m,可以替代人工测量,完成电力线的三维重建。(3)电力线通道障碍物检测与分类评定方法研究通过计算导线与电力线通道地表三维点云之间的距离,识别和定位通道内的障碍物;通过对障碍物点云的空间分布特征和纹理特征进行分析,对障碍物类别进行分类,评估它们的危险程度。选取长度为3.9km的典型线路作为实验区,发现6.5m有效障碍物8处,经野外实地复核,障碍物与电力线之间距离测量结果与现行人工巡检方法相比,差值优于0.5m,满足障碍物巡检对距离测量精度的要求(优于±2.0m)。更多还原

【Abstract】 The inspection of power lines is a critical factor for the safe operation of power transmission grids.To ensure the safe operation of power lines,it is necessary to ensure that a range of empty space without conductive objects exists around an extra high voltage(EHV)power line.Trees or tree branches that are too close to power lines should be trimmed,because vegetation is conductitve,which will lead to electric arcs.However,vegetation within the power line corridor will naturally grow after a power transmission grid becomes operational.A discharge may be generated when the distance between the vegetation and the power line is less than the safety threshold,thereby endangering the safe operation of the power transmission grid.As a result,the research of power line inspection across a large area has become a hot spot in recent years,whose main purpose is to detect possible obstacles within the corridor.Of all the possible means to carry out power line inspection,UAVs show a great potential due to the ability to perform 3D reconstructions for the power line corridor at low costs when equipped with lightweight digital cameras.Once stereo images of the power lines and the ground are taken during flights,they can be subjected to photogrammetric methods to give a 3D reconstruction result,which can then be used to automatically detect and measure possible obstacles through a spatial distance calculation.To use UAV to detect the obstacles within the power line corridor,three key problems need to be solved;3D reconstruction of the ground of the power line corridor,automatic power line measurement,and automatic recognition of obstacles.For ground reconstruction of the power line corridor,dense image matching techniques can be adopted to extract dense point clouds of the power line corridor.In UAV power line inspection,image-matching algorithms must be adapted to ground conditions with dense vegetation.At the same time,algorithms must be capable of extracting tree canopies and buildings stably and reliably.It is also difficult to use image-matching methods to find the corresponding points along a power line because of the small diameter of power lines and the complexity of their backgrounds.A 3D reconstruction of power lines can be conducted by the traditional manual stereo measurement method,which restricts the level of automation of inspection by using UAV images and which needs further improvement.In this paper,we propose an automatic inspection method for power lines using UAV images.This method,known as the power line automatic measurement method based on epipolar constraints(PLAMEC),acquires the spatial position of the power lines.Then,the semi patch matching based on epipolar constraints(SPMEC)dense matching method is applied to automatically extract dense point clouds within the power line corridor.Obstacles can then be automatically detected by calculating the spatial distance between a power line and the point cloud representing the ground.The main researchcontents and innovative contributions are as follows:(1)Semi Patch Matching Algorithm Based on Epipolar ConstraintsIn our proposed method,SPMEC,epipolar images are the processing unit and a coarse to fine image matching strategy was adopted,under the initial parallax constraint,a large matching window searches a one-dimensional image along the epipolar line and is used to extract the coarse matching seed points.If a parallax is continuous within an object and the same object has a consistent texture in the image,then the coarse matching seed point is considered the center of the fine matching window.In the segmented image,based on the segmented object of the seed point,a patch matching constraint is constructed.Then,the initial parallax of the points to be matched within the semi patch is determined according to the geometric conditions of the semi patch.A one-dimensional search is conducted within a smaller search range.Finally,an outlier detection algorithm is applied to eliminate the mismatched points.The experimental results showed that dense point clouds of canopies and buildings with regular outlined can be extracted by SPMEC in rural areas with lush vegetation.A major improvement of the proposed method over most matching algorithms that based on a normalized CC(correlation coefficient)is that the value of the normalized CC as well as its curve characteristics are both used.Once the corresponding image point is found in the search region of the right image when sliding the matching window in the image correlating process,the CC will change following a law from small to large as the matching window gets closer to the corresponding point;and from large to small as the matching window moves far away.Isolated peaks are evident in the data;two sides of the peak value are approximately symmetrical.It’s demonstrated by the experimental results that to get more dense matching points,successful matching rates of the curve characteristics of the CC can be improved without introducing too many mismatch points if the proposed method is used instead of simply lowering the threshold.Another crucial issue of dense image matching is the disparity compatibility between a matching point and its neighbor’s.The left epipolar imageis segmented using Graph-Based image segmentation method.The experimental result shows in the road and rooftop of the point cloud obtained using semi patch,there are fewer holes,for which the initial parallax of the surrounding objects and homogenous textures in the match window are to be blamed,compared with the point cloud obtained without semi path.By determining the initial parallax of the match windowusing the corresponding segmentafter semi patch is applied,the successful match rate are increased in roads,rooftop areas as well as vegetation areas.(2)Power Line Automatic Measurement method based on Epipolar ConstrainsPLAMEC(Power Line Automatic Measurement method based on Epipolar Constrains)is proposed for automatically measuring power lines because of the necessity to locate ground points in the image pairs along the corresponding epipolar lines.The first step is to select a pair of images from two adjacent strips as the stereo image pair,which must be selected carefully to ensure a perpendicular relationship between the epipolar line and the direction of the power line.The coplanarity condition is then applied when generating the epipolar images while the relative orientation algorithm is used to derive the relative orientation parameters of the stereo image pair.After that,the 2D vectors of the power lines are obtained from the left and right epipolar images using the power line automatic extraction algorithm before finally extracting pairs of corresponding epipolar lines at a certain interval in the direction of the y parallax of the epipolar image.A pair of corresponding points on the power line is eventually extracted by intersecting the epipolar line with the two power line vectors in the left and right epipolar images respectively.It was showed by the results that automatic measurement of power lines was done at a success rate of 93.2%,which is as good as that of manual measurement.Besides,the method is further proved to be an appropriate substitution of manual measurements since the RMS error of the elevation differences is better than ±0.15 m between the two methods.(3)Automatic Detection of Obstacles within a Power Line CorridorThe extracted power line is taken as the bus line to construct a spatial buffer around at a safe distance,after which obstacles can be detected by intersecting the 3D point clouds against the spatial buffer zone.Asection,whose approximate length is 3.9 km，of a 220 kV power line was selected with the elevation variation along it being about 200 m;nine towers were found within the corridor and the safe distance threshold was set as 6.5 m.The result shows that eight true obstacles were detected and their distances to the power lines turned to be approximately consistent to those measured by field surveying.Besides,the distance difference between the two methods was better than±0.5 m while the obstacle inspection requirement was better than ±2m,which means the requirement was meted by the proposed method.更多还原