神经网络支持下的无人机山区DEM测量技术研究

Research on DEM Measurement Technology by UAV in Mountain Area Supported by Neural Network

无人机技术作为现如今测量的先进技术,已被逐渐广泛应用,为找出其在山区DEM测量中的处理方法。基于卷积神经网络模型,研究了无人机不同测点的DEM处理,并与滤波方法和样条函数法进行比较,结果表明:当卷积网络的卷积层数为3,卷积核数为3×3时可保证模型的最高精度;该模型对DEM处理的结果与实际情况相符,DEM图、等高线图、TIN图均与实际保持了较高的同步度;同时该模型与实测值的误差最低,一致性最高,表明该模型精度较高,因此卷积神经网络可用于无人机DEM测量中。

UAV technology has been widely used as an advanced technology in surveying and mapping nowadays.In order to find out the processing method of DEM measurement in mountainous area,this paper studied the DEM processing of different measuring points of UAV based on convolutional neural network model,and compared it with filtering method and spline function method.The result showed that when the layer of convolutional neural network was 3 and the convolution kernel was3 x3,the model results for DEM processing were precise and consistent with the actual situation.The DEM map,contour map,and TIN map all maintained a high degree of synchronization with the actual situation.The model had the lowest error and the highest consistency with the measured values.It indicated that the model had the highest precision and could be used in UAV DEM survey.