架空输电线路巡线机器人视觉检测与导航控制研究
文献类型：学位论文
| 作者 | 张运楚 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2006-06-05 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 谭民 ; 梁自泽 |
| 关键词 | 巡线机器人 视觉导航 物体识别 视觉伺服 视觉检测 Power Line Inspection Robot Visual Navigation Object Recognition Visual Servo Visual Inspection |
| 其他题名 | Research on Visual Inspection and Navigation for Overhead Power Line Inspection Robot |
| 学位专业 | 控制理论与控制工程 |
| 中文摘要 | 巡线机器人能够沿输电线路爬行，利用携带的检测仪器代替人对杆塔、导线、绝缘子、线路金具等实施接近检查。这种作业方式，将大大减轻输电线路维护人员的劳动强度，提高巡检效率与质量，对保证输电系统的安全可靠运行具有重要意义。巡线机器人多采用轮式行走机构，架空输电线路相线和地线上安装的防震锤、绝缘子、悬垂线夹、耐张线夹等线路附件是影响巡线机器人行走的障碍。因此，巡线机器人必须借助传感器探测并识别这些障碍，然后根据障碍类型规划越障行为，跨越杆塔，才能沿输电线路大范围自主行走。 本文围绕巡线机器人视觉导航、输电线路附件缺陷视觉检测等任务，针对障碍目标的检测、识别、空间定位、越障视觉伺服和绝缘子破损检测等问题进行了研究。本文工作的主要贡献体现在： 针对220kV架空输电线路的结构特点，提出了一种基于存在概率图的圆/椭圆检测方法，在此基础上设计了基于结构约束的障碍识别算法。并采用Kalman预估器跟踪障碍物，融合帧间信息以提高障碍物识别质量。 采用双目立体视觉测量障碍物与巡线机器人之间的距离。对基于区域的立体匹配方法中区域相似性测度、匹配窗口尺度和匹配搜索策略等进行了研究。针对障碍物表面亮度较为均匀的问题，提出了一种自适应多匹配窗的搜索策略，提高了对应点的匹配精度。 根据双臂巡线机器人越障操作的特点，设计了基于图像的越障视觉伺服方案。通过对傅立叶描述子性能的分析，构造了具有平移、旋转、尺度缩放及起始点不变性的轮廓形状特征向量，提出了基于傅立叶描述子的驱动轮识别、图像特征抽取和驱动轮―相线图像空间中相对位姿估计等算法，在此基础上设计了带有死区的比例控制器，实现了驱动轮―相线对中视觉伺服控制。 提出了一种基于绝缘子图像窗口纵向剖线模式的表面破损检测算法。算法根据瓷绝缘子的结构特点，从CCD摄像机获取的图像中抽取瓷绝缘子局部区域特征,利用三层BP神经元网络从复杂背景中识别出目标绝缘子并定位，然后采用扫描分析绝缘子图像窗口纵向剖线模式的方法判断绝缘子的完好状态，实现瓷绝缘子破损的视觉检测。 |
| 英文摘要 | Power line inspection robot can crawl along the power line and inspect the line subsidiary equipments such as power tower, power line, isolator etc instead of people with devices, which aims at the relief of maintenance workload, the improvement of work efficient, the decrease in cost and the guarantee of the steady operation of the power line system. The inspection robot is wheel-driven, when it crawls along the power transmission line, the line subsidiary equipments will act as obstacles to block its way. In order to fulfill the inspection task, a navigation system is needed to recognize and locate the obstacles with its sensors; then the control system of inspection robot will plan its motions according to the obstacle information to negotiate these obstacles autonomously. The dissertation focuses on visual inspection and visual navigation technologies for power line inspection robot, the key problems such as obstacles detection and recognition, 3D reconstruction, visual servo control of obstacle negotiation and isolator visual inspection are studied. The main contribution of the dissertation is as follows: An existence-probability-map based circle/ellipse detection method is put forward; according to the characteristic of a 220kV power line network, the structure-constrained obstacle recognition algorithm is designed by using the above method. In order to improve the recognition accuracy, a Kalman prediction filter is used to track the detected obstacle. The distance from inspection robot to obstacle is measured by binocular vision. Region-based stereo matching method is used; and the metric of region similarity, the selection of matching window size and the match searching strategy are discussed in detail. Because of homogeneous brightness of the obstacle surface, an adaptive multi-matching windows stereo matching algorithm is put forward to improve the matching accuracy. An image-based visual servo control scheme is designed to guide the inspection robot to negotiate the obstacles. Based on the Fourier descriptors, the feature vector for the driving-wheel contour is constructed, which is an invariant to rotation, scaling, translation and starting point. Then the algorithms for driving-wheel recognition, image feature extraction and driving-wheel-powerline relative pose estimation are put forward. A proportional control law with dead zone is given to achieve the driving-wheel-powerline alignment operation. A vertical-section-line-mode based isolator fault visual inspection method is put forward. First, based on its part region feature, the isolator being inspected is recognized and located from its background by using a 3-layer BP neural network; then by scanning the cropped image window contained the isolator and analyzing the mode of vertical section line, the status of the isolator is determined. |
| 语种 | 中文 |
| 其他标识符 | 200318014602995 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/5937]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 张运楚. 架空输电线路巡线机器人视觉检测与导航控制研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2006. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
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。