图像识别技术在机组运行状态判断及监控对点中的应用

Application of Image Recognition Technology to Judgement for Operation Condition and Monitoring Signals of Hydropower Unit

摘要:

水电站主要是通过人工查看数字量输入（DI）板信号指示灯、监控画面、事件表和报警表查看机组运行状态正常与否，花费时间较多且存在遗漏的可能。文中提出了一种将图像识别技术应用到DI板信号指示灯识别的方法，通过图像识别自动生成信号点表，并将图像处理算法（模板匹配、动态规划、几何定位与区域选择算法）集成到手机应用软件（APP）。使用该APP的拍照和识别功能，实现对机组DI板信号指示灯的实时、快速识别并输出结果，识别正确率可达到98%以上。现场人员根据识别结果可快速查看机组运行时信号点状态是否正常，有针对性地检查和对点，提高了工作效率，特别是对于新建水电站机组投运和已投产水电站机组检修作业的优化效果更为明显，同时所提出的图像处理算法可推广到更多工业场景下的指示灯及仪表智能识别应用中。

Abstract:

A hydropower station is always characterized by manual work involved in reading the indicators of digital input (DI) board, monitoring images, event lists and alarm apparatus to ensure the validity of the unit, which costs a lot of time and may ignore something by the staff. So the application of image recognition technology is proposed in reading the indicators of DI board. The signal location list is created automatically in image recognition, while the algorithms of image processing, such as template matching, dynamic programming, geometrical positioning and area selection, are integrated into a mobile APP. By taking photographs and recognizing the images, this APP can enable the user to read the indicators of DI board and output the result immediately at any time with an accuracy above 98%. According to the results, the staff can quickly know the validity of signal locations, and then inspect the locations purposefully. It is a great improvement in working efficiency, especially for the initiation of a new hydropower station or the maintenance of a running one. At the same time, the proposed image processing algorithm can be extended to more intelligent recognition applications of indicator lights and instruments in industrial scenes.