水电站溢流坝表观裂缝损伤智能检测方法研究  

Research on Intelligent Detection Method for Crack Damage of Overflow Dam of Hydropower Station

作　　者：冯春成[1] 张华[1,2] 汪双[1] 李永龙 王皓冉 FENG Chun-cheng;ZHANG Hua;WANG Shuang;LI Yong-long;WANG Hao-ran(School of Information and Engineering,Southwest University of Science and Technology,Mianyang 621000,China;Sichuan Energy Internet Research Institute,Tsinghua University,Chengdu 610000,China)

机构地区：[1]西南科技大学信息工程学院,绵阳621000 [2]清华大学四川能源互联网研究院,成都610000

出　　处：《自动化与仪表》2021年第6期55-60,共6页Automation & Instrumentation

基　　金：国家重点研发计划项目(2019YFB1310505);四川省重点研发计划项目(2020YFSY0062,2021YFG0100)。

摘　　要：针对人工巡视获取坝面图像方式存在风险高、效率低及传统裂缝检测方法的检测精度偏低和实时性差等问题,搭建了具备长续航能力的系留无人机系统,通过搭载高清云台相机采集坝面图像;同时基于语义损伤检测网络(SDDNet)提出了一种改进的坝面裂缝损伤检测方法;在特征编码器与解码器之间引入注意力机制模块提升模型对感兴趣区域的关注度,并采用非对称卷积操作减小模型参数量。实验结果表明,所提方法的裂缝检测精度高于SDDNet,且模型参数量也明显减低,有效地提升了模型的实时性和准确率。Aiming at the problems of high risk and low efficiency in obtaining images of the dam surface by manual inspection,as well as the low detection accuracy and poor real-time performance of traditional crack detection methods,a sensing system of tethered UAV with long endurance capabilities is developed,and images of the dam are collected through the equipped high-definition gimbal camera.A crack detection method of dam is proposed based on improved semantic damage detection network,a attention module is added between the encoder and decoder of SDDNet to improve the attention of region of interest,the asymmetric convolution is used to reduce the amount of model parameters.The Experimental results show that the crack detection accuracy of the proposed method is higher than that of SDDNet,and the model parameters are also significantly reduced,which effectively improves the real-time performance and accuracy.

关 键 词：系留无人机 深度卷积 注意力机制 裂缝检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]