雾天环境下的指针式仪表检测与读数识别  

The Method for Pointer Instrument Detection and Reading Recognition in Foggy Conditions

作　　者：吴攀超 杨鹏 孙电洋 杨明昊 WU Pan-chao;YANG Peng;SUN Dian-yang;YANG Ming-hao(School of Electrical and Information Engineering,Northeast Petroleum University)

机构地区：[1]东北石油大学电气信息工程学院

出　　处：《化工自动化及仪表》2025年第1期101-110,共10页Control and Instruments in Chemical Industry

基　　金：黑龙江省省属高校基本科研业务费(批准号:2022TSTD-04)资助的课题。

摘　　要：针对雾天油田巡检机器人对指针式仪表漏检及读数识别准确率低的问题,提出一种结合FFA-Net去雾算法和Yolov5检测网络的方法。首先,对于雾天图像,改进了FFA-Net算法,通过多尺度结构、特征融合残差块和优化模块,有效地提升了算法在去雾任务中的性能表现。其次,对于仪表检测与读数识别,采用较小的检测头来提升Yolov5对小目标的检测能力,并引入空间转换模块将检测到的表盘图像转换为更符合人眼观感的正视图像。最后,创建了一个端到端的框架,紧密耦合仪表成分检索和仪表读数识别,提高了仪表读数的准确性。实验结果表明,所提方法在油田雾天环境下展现出了良好的鲁棒性,提升了雾天环境下指针式仪表检测与读数识别的准确性。Aiming at the oilfield inspection robots’missed detection and low recognition accuracy of the pointer instrument in foggy weather,a method which combining FFA-Net dehazing network and Yolov5 detection algorithm was proposed.Firstly,it has a FFA-Net algorithm improved for the foggy images and has the multi-scale structure and the feature fusion residual block and optimization module based to effectively enhance the algorithm′s performance in dehazing operation;and then,as for the instrument detection and reading recognition,it has a smaller detector head adopted to improve Yolov5’s ability in detecting small targets,and has a spatial transformation module introduced to convert the dial’s images detected into a front elevation in line with human perception;finally,it has an end-to-end framework created to tightly couple the meter component retrieval and meter reading recognition so as to improve the accuracy of meter readings.The experimental results show that,the proposed method boasts good robustness in the oilfield foggy environment,and it improves both accuracy of detection and reading recognition of the pointer meter in foggy environment.

关 键 词：图像去雾 Yolov5 仪表检测 仪表识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]