基于无人机倾斜摄影的溢流坝缺陷监测  

Defect Monitoring of Overflow Dam Based on UAV Tilt Photography

作　　者：蒋红辉 Jiang Honghui(Guangdong Hydropower Yunnan Investment Co.,Ltd,Honghe 661014,China)

机构地区：[1]广东水电云南投资有限公司,云南红河661014

出　　处：《云南电力技术》2023年第1期74-77,共4页Yunnan Electric Power

摘　　要：云南地区地质条件复杂,地震、泥石流等自然灾害多发、频发,对水电站溢流坝造成严重的安全威胁。本文针对传统的依赖人工对溢流坝进行裂缝巡检存在成本高、安全风险大、耗时长等问题,提出了一种基于无人机倾斜摄影的溢流坝缺陷监测技术,构建了基于优化pix2seq的溢流坝裂缝智能识别模型,实现了溢流坝裂缝的主动识别,针对贴近摄影时图像容易产生退化的问题,本文讨论了退化图像复原模型,并完成了基于无人机倾斜摄影的溢流坝裂缝长度自主监测,并于同类别其它方法进行了对比,通过实验证明了所提方法的有效性,预计可为水电站大坝安全运维的信息化、智能化发展提供技术支撑。The geological conditions in Yunnan are complex,and natural disasters such as earthquakes and debris flows are frequent,which pose a serious security threat to the overflow dam of hydropower stations.In view of the problems of high cost,high safety risk and long time consuming in the traditional manual inspection of overflow dam cracks,this paper proposes a kind of overflow dam defect monitoring technology based on UAV tilt photography,constructs an intelligent identification model of overflow dam cracks based on optimized pix2seq,and realizes the active identification of overflow dam cracks.Aiming at the problem of image degradation when close to photography,this paper discusses the degradation image restoration model,The independent monitoring of overflow dam crack length based on UAV tilt photography has been completed and compared with other methods of the same kind.The effectiveness of the proposed method has been proved by experiments.This paper is expected to provide a certain technical support for the informatization and intelligent development of the safe operation and maintenance of hydropower stations.

关 键 词：神经网络 缺陷检测 倾斜摄影 退化图像 人工智能 

分 类 号：TM74[电气工程—电力系统及自动化]