| 题名 | 变电站带电水冲洗机器人的绝缘子串识别与定位方法研究 |
| 作者 | 刘澈 |
| 学位类别 | 硕士 |
| 答辩日期 | 2017-05-24 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 授予地点 | 沈阳 |
| 导师 | 姜勇 |
| 关键词 | 绝缘子串 联合标定 特征提取 信息融合 识别与定位 |
| 其他题名 | Research on Identification and Location of Insulator Strings for Live Washing Robot in Substation |
| 学位专业 | 控制工程 |
| 中文摘要 | 变电站绝缘子串的污闪事故是影响电网设备安全运行的主要隐患，及时对绝缘子表面污秽进行带电水冲洗作业是防止污闪事故发生的最有效手段。研制一种新型的可替代人工完成变电站带电水冲洗作业的智能型机器人是相关领域的一个研究热点。带电水冲洗机器人在进入变电站作业区域进行清污作业时，需要机器人能够通过自身的视觉系统找到绝缘子串，即机器人需要具备自主识别和定位绝缘子串的能力。这是变电站带电水冲洗机器人研制过程中急待解决的核心关键技术之一，亦是本文重点研究的科学问题。本文的研究工作得到了南方电网公司重点科技项目“110-220kV变电站带电作业水冲洗机器人的应用与研究”的资金与技术支持，主要研究包括： 首先，通过文献调研和现场考察，了解了关于绝缘子串清洗机器人、绝缘子串识别方法等相关领域的国内外研究现状、研究热点及难点，明确了本课题的研究意义、研究重点和研究思路。 其次，介绍了机器人视觉系统的基本组成结构，分析了工业相机和激光雷达的单独标定方法；针对镂空标定板存在的混合像素点问题，研究设计了一种用于激光雷达与相机联合标定的专用栅形标定板；在此基础上，通过激光点云图像网格化和法向量颜色空间映射，研究提出了一种基于法向量图的激光雷达与相机联合标定方法，标定效果得到了像素级图像融合的验证。 然后，分析了工业相机光学图像和激光雷达深度图像中绝缘子串的表征特点；针对预处理后的光学图像和深度图像，研究建立了包含绝缘子串特征信息的灰度和深度特征曲线模型；在此基础上，定义了灰度周期性特征矩阵、灰度阈值特征矩阵、深度周期性特征矩阵和宽度特征矩阵，并提出了基于四种特征矩阵的绝缘子串特征提取方法。 接着，研究提出了基于SVM概率输出与证据相容度的基本概率赋值构造方法，改进了D-S证据理论，融合激光雷达与相机提取出的特征信息，实现了复杂室外环境中绝缘子串的准确识别。该方法能够充分利用多源信息互补性，通过SVM概率输出得到环境加权系数，在一定程度上消除了光照对判别结果的影响；利用特征证据相容度，对冲突证据的可信度做出判断，使识别结果具有更高准确性，增强了判别结果的容错能力；通过识别结果与特征证据的一致性判定，弥补了单一传感器提取信息可能不完整的不足，使定位结果更加可靠。 最后，利用物理样机在实际变电站环境中开展了绝缘子串识别与定位模拟实验，验证了基于激光雷达和工业相机的带电水冲洗机器人视觉系统的实际使用效果；分析了所提出的绝缘子串识别与定位方法的可行性和有效性，为后续的系统改进提供了实验依据。 |
| 英文摘要 | The pollution flashover accident of insulator strings in substation is the main hidden trouble that affects the safe operation of power grid equipment, it is the most effective way to prevent the pollution flashover accident in time to carry out the flushing water on the surface of insulator. It is a hot research field to develop a new type of intelligent robot, which can be used to replace the manual water washing operation. Live washing robot into the substation operation area clean-up operation, need the robot can through the visual system itself to find the insulator strings, namely robots need to possess the ability to identify and locate the insulator string. This is one of the key technologies to be solved in the process of the development of the live washing robot, it’s also a scientific issue which the paper focuses on. This work is based on“110-220kV substation live washing robot” by China Southern Power Grid Company Limited, and the main research contents are as follows. First of all, through literature research and field investigation, the research status, research hotspots and difficulties in the field of insulator string cleaning robot and insulator string identification methods are studied, the significance of this research, research priorities and research ideas are clarified. Secondly, this paper introduces the basic structure of the robot vision system, and analyzes the calibration method of industrial camera and laser radar. In order to solve the problem of the mixed pixel in the hollow calibration plate, a special calibration plate is designed to complete the joint calibration between the camera and laser radar. On the basis of this, a joint calibration method of laser radar and camera is proposed based on the normal vector image, which is based on the grid point cloud image and the normal vector color space mapping, the calibration results are verified by pixel level image fusion. Then, the feature of the insulator strings in the optical image and the depth image of lidar are analyzed; The model of gray characteristic curve and depth characteristic curve is established. On this basis, the gray periodic feature matrix, the gray threshold feature matrix, the depth periodic feature matrix and the width feature matrix are defined, a method to extract feature of insulator strings based on four feature matrix is proposed. Next, this paper presents a method to construct the basic probability assignment based on SVM probability output and evidence compatibility degree, the D-S evidence theory is improved, and the feature information extracted from laser radar and camera is fused to realize the accurate identification of insulator strings in complex outdoor environment. This method can make full use of the complementary information of multi-source information, and obtain the environmental weighting coefficient by SVM probability output, to a certain extent, the effect of illumination on the results is eliminated. Making use of the compatibility degree of the evidence to judge the credibility of conflict evidence, it make the recognition results have higher accuracy, and the fault tolerance ability is enhanced. Based on the consistency of the identification results and the characteristic evidence, it can make up the deficiency that the single sensor cannot complete the information, and make the result more reliable. Finally, the simulation and experiment of insulator string identification and location are carried out by using the physical prototype. The actual effect of the visual system of the water washing robot based on laser radar and industrial camera is verified. The feasibility and effectiveness of the proposed method for identification and location of insulator strings are analyzed, and it provides the experimental basis for the following system improvement. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 内容类型 | 学位论文 |
| 源URL | [http://ir.sia.cn/handle/173321/20529]  |
| 专题 | 沈阳自动化研究所_空间自动化技术研究室 |
| 推荐引用方式 GB/T 7714 | 刘澈. 变电站带电水冲洗机器人的绝缘子串识别与定位方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所. 2017. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论