基于神经网络的测量机器人大坝监测观测值折光修正  

Refraction Correction of Dam Monitoring Observations Using Robotic Total Station Based on Neural Network

作　　者：孙倩雯 周建国 刘冠兰[2] SUN Qian-wen;ZHOU Jian-guo;LIU Guan-lan(School of Civil Architecture and Environment,Hubei University of Technology,Wuhan 430068,Hubei Province,China;School of Surveying and Mapping,Wuhan University,Wuhan 420072,Hubei Province,China)

机构地区：[1]湖北工业大学土木建筑与环境学院,湖北武汉430068 [2]武汉大学测绘学院,湖北武汉420072

出　　处：《中国农村水利水电》2024年第4期187-192,共6页China Rural Water and Hydropower

基　　金：国家自然科学基金(41904170)。

摘　　要：测量机器人在水库大坝自动化监测中应用广泛,为确保监测结果的可靠性,需对原始观测值进行折光修正。结合测量机器人自动化监测可提高监测频率、提供大量观测数据的特点,在分析观测值折光影响因素的基础上,构建了神经网络折光修正模型来对原始观测值:垂直角、斜距、水平角进行修正。并利用某水电大坝自动化监测系统展开了试验与验证,选取评价指标对模型修正后观测值精度进行分析。结果表明,对于以垂直角为基础的垂直位移修正,基于神经网络的方法相对于传统的K值修正公式效果明显,最高可提升垂直位移的精度约2 mm。对于斜距修正,基于神经网络的方法有一定的效果,显示出距离越远效果越明显的趋势。对于水平角修正,数据结果表明水平角观测值受折光影响较小,使用神经网络进行修正并不能进一步提高观测值的精度。The robotic total station is widely used in automatic monitoring of reservoir dams.In order to ensure the reliability of the monitoring results,the refraction correction of the original observations is required.Based on the characteristics that automatic monitoring of robotic total station can improve the monitoring frequency and provide a large amount of observation data,this paper builds a neural network refraction correction model to correct the original observations including vertical angle,slope distance and horizontal angle.And an automatic monitor⁃ing system for a hydropower dam is used to carry out experiments,and the accuracy of the observations after correction is analyzed.The re⁃sults show that for the correction of vertical displacement based on vertical angle,the method based on neural network is more effective than the traditional K-value correction formula,and the maximum accuracy of vertical displacement can be improved by about 2 mm.For slope distance correction,the method based on neural network has a certain effect,showing a trend that the greater the distance,the more obvious the effect.As for the correction of horizontal angle,the results show that the horizontal angle is less affected by refraction,and the correction using neural network cannot further improve its accuracy.

关 键 词：大坝监测 测量机器人 观测值修正 神经网络 

分 类 号：P642.22[天文地球—工程地质学] TV698.1[天文地球—地质矿产勘探]