配电线路维护机器人目标识别和定位研究

发布时间：2018-03-10 19:33

本文选题：非结构化环境　切入点：预处理　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着技术的进步,在配电线路带电维护作业中使用机器人代替人进行作业具有重要的现实意义。在室外非结构化作业环境中,机器人视觉系统采集到的图像易受复杂背景、天气和光照等各种环境因素的影响,大大增加了识别和定位作业目标的难度。为了使得机器人能够自主完成带电作业任务,本文结合某配电线路维护机器人的研发项目,研究了作业目标识别和定位方法,论文主要完成以下工作:(1)对配电线路维护机器人的视觉系统进行需求分析,详细描述视觉系统应具备的功能。在此基础上,设计了视觉系统中目标识别与定位子系统、监控子系统和防碰撞检测子系统的技术方案。(2)研究配电线路环境中的图像预处理方法。针对复杂背景下作业目标提取困难的问题,提出了基于颜色直方图和形态学的图像分割方法。针对弱光、局部高光和全局高光的图像,分别采用CLAHE、灰度非线性变换以及同态滤波等方法进行处理,突出了弱光、局部高光和全局高光图像中的作业目标,为提取作业目标奠定了良好的基础。(3)复杂背景下作业目标识别方法研究。为识别跌落式熔断器中的熔丝管,提出了一种基于几何约束的直线拟合法,采用概率Hough变换,并利用直线平行和距离约束等条件,对熔丝管管体进行识别。针对操作环识别困难的问题,利用熔丝管管体和操作环的几何位置关系,提出基于位置和大小约束的椭圆拟合法识别操作环。针对复杂背景中其它物体对熔丝管识别造成的干扰,提出基于熔丝管管体和操作环互相约束的熔丝管识别方法,识别复杂背景中的熔丝管。实验结果表明,所提方法可以识别不同场景中的熔丝管,准确度高,且鲁棒性较好。(4)基于双目视觉的作业目标空间位置测量方法研究。在对双目摄像机进行标定的基础上,利用Bouguet算法对图像进行立体校正。利用熔丝管的几何参数确定匹配窗口的大小,使用极线约束降低匹配维度,再采用归一化的相关系数作为匹配测度函数,对左右图像中的熔丝管进行立体匹配,具有匹配速度快且精度高的优点。最后给出了利用双目立体匹配的视差计算熔丝管空间位置的方法。
[Abstract]:With the development of technology, it is very important to use robot instead of human in the maintenance of distribution line. In the outdoor unstructured working environment, the images collected by robot vision system are vulnerable to complex background. The influence of various environmental factors, such as weather and light, greatly increases the difficulty of identifying and locating the operation target. In order to enable the robot to accomplish the task of live operation independently, this paper combines the research and development project of a power distribution line maintenance robot. This paper studies the method of target identification and location. The main work of this paper is as follows: 1) analyzing the requirements of the vision system of the distribution line maintenance robot, and describing in detail the functions that the vision system should have. The technology scheme of target recognition and location subsystem, monitoring subsystem and anti-collision detection subsystem in visual system is designed to study the image preprocessing method in distribution line environment, aiming at the difficult problem of target extraction in complex background. A method of image segmentation based on color histogram and morphology is proposed. For the images with weak light, local highlights and global highlights, the methods of Clare, gray nonlinear transformation and homomorphic filtering are used to deal with them, respectively, and the weak light is highlighted. In order to identify the fuse tube in the drop fuse, the operation target in the local highlight and the global high light image is studied, which lays a good foundation for the extraction of the operation target under the complex background. In this paper, a linear fitting method based on geometric constraints is proposed. Using probabilistic Hough transform, linear parallelism and distance constraint are used to identify the fuse tube body. Based on the geometric position relationship between the fuse tube body and the operating ring, an elliptical fitting method based on position and size constraints is proposed to identify the operation ring. A method of identifying fuse tube based on mutual constraint between fuse tube body and operation ring is proposed to identify fuse tube in complex background. The experimental results show that the proposed method can identify fuse tube in different scenes with high accuracy. Moreover, the method of spatial position measurement based on binocular vision is studied. Based on the calibration of binocular camera, The size of matching window is determined by the geometric parameters of fuse tube, the matching dimension is reduced by pole line constraint, and the normalized correlation coefficient is used as the matching measure function. The stereo matching of fuse tubes in left and right images has the advantages of fast matching speed and high precision. Finally, a method for calculating the space position of fuse tubes by using the parallax of binocular stereo matching is presented.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】