摘要
为找出无人机实测数据高程转换拟合的方法,本文基于不同激活函数激活极限学习机模型得到ELMS、ELMR、ELMH3种模型,并将计算结果与广义回归神经网络模型和BP神经网络模型对比,结果表明:基于5种模型可对无人机实测高程数据的异常点进行筛选并剔除,3种ELM模型对于高程点的筛选结果基本一致,共筛选出了21个高程异常点,GRNN模型精度次之,BP神经网络模型精度最低,ELM模型在不同激活函数下的计算精度有所不同,其中ELMS模型在高程点剔除和高程数据拟合中精度最高,RMSE仅为0.157m,而Ens和R2分别达到了0.944和0.968,可为无人机实测数据高程转换拟合的标准模型使用。
In order to find out the method of height transformation and fitting of UAV measured data,elms,elmr and elmh models are obtained based on different activation functions,and the results are compared with generalized regression neural network model and BP neural network model.The results show that the abnormal points of UAV measured elevation data can be screened and eliminated based on the five models,and three elm models are available The accuracy of GRNN model is the second,BP neural network model is the lowest.Elm model has different calculation accuracy under different activation functions.Among them,elms model has the highest accuracy in elevation point elimination and elevation data fitting,RMSE is only 0.157 m,while ens and R2 are 0.944 and 0.968,respectively The standard model of UAV measured data height conversion fitting is used.
作者
陈明玲
CHEN Mingling(Hainan Guoyuan Land and Mineral Exploration Planning and Design Institute,Haikou Hainan 570203,China)
出处
《北京测绘》
2020年第8期1057-1060,共4页
Beijing Surveying and Mapping
关键词
无人机(UAV)
高程拟合
激活函数
极限学习机模型
Unmanned Aerial Vehicle(UAV)
elevation fitting
activation function
limit learning machine model
