基于感知组织的输电线路结构识别方法  

Method on recognizing the structure of transmission line based on perceptual organization

作　　者：韩军[1] 张晶晶[2] 王滨海[2] 

机构地区：[1]上海大学通信与信息工程学院,上海200072 [2]国家电网公司电力机器人重点实验室,山东济南250002

出　　处：《红外与激光工程》2013年第12期3458-3463,共6页Infrared and Laser Engineering

基　　金：国网山东省电力公司重点科技项目(2012A-17);基于无人机的输电线路异常和缺陷自动诊断技术

摘　　要：为了提高输电线路缺陷诊断正确率,有效降低各种复杂背景纹理及光线对识别输电线路结构的影响,从Gestalt感知理论着手,研究一种多感知识别输电线路结构的方法。在图像识别的底层,提取不同方向、不同宽度的线段,研究了一种融合计算Gestalt定律的近似性、连续性、共线性的多级搜索算法,获得显著的、完整的输电线路人造对象轮廓;在图像识别的中层,研究一种基于分块与合并的计算方法能视觉感知近平行线、近对称交叉的结构,设计了一个三级分类器感知聚类平行线组;在图像识别的高层,研究输电线路的知识模型,建立识别输电线路组成结构的约束机制,进而从语义上唯一地识别输电线路的结构。通过无人机巡检采集的输电线路图像,验证这种方法能有效识别输电线路组成的杆塔、导线、地线及绝缘子所在区域。In order to improve the diagnosis accuracy of the transmission line defect, and reduce the influence on identifying the structure of the transmission line made by complicated background texture and light. Starting from Gestalt perception theory, a multiple perceptual identification method was developed to identify transmission line structure. Firstly, line with different directions and different width was extracted and sorted. Through a kind of multilevel searching algorithm with similarity, continuity and colinearity of Gestalt Law, the contour of transmission line was obtained accurately and completely. Secondly, a method based on block partition was developed, which can visually perceive near parallel lines and near symmetrical cross structure. A three level classifier for clustered parallel line group was designed. Lastly, combined with prior knowledge of the transmission line model, constraint mechanism was built to recognize the structure of transmission lines, and then uniquely identify the semantically structure of transmission lines. Experimental results show that the method can effectively identify the transmission line consisting of the tower, conductor, earth wire and the insulator region through recognition of the UAV inspection acquisition transmission line image.

关 键 词：输电线路结构 人造设施识别 人造设施缺陷诊断 感知组织 

分 类 号：TN911.73[电子电信—通信与信息系统]