基于BP神经网络的DEM粗差探测法分析  

Analysis of DEM Gross Error Detection Method based on BP neural network

作　　者：陈百了 周访滨[1] CHEN Bailiao;ZHOU Fangbin(Changsha University of Technology,Hunan 410114,China)

机构地区：[1]长沙理工大学,湖南410114

出　　处：《电子技术（上海）》2024年第1期405-407,共3页Electronic Technology

基　　金：长沙理工大学研究生科研创新项目(SJCX2021121)。

摘　　要：阐述一种以BP神经网络为基础的DEM粗差探测方法,为了减少该算法的误判率,加入怀疑率这一参数,这样某一点的粗差判断可以根据多个区域的检测结果来进行。为了确定怀疑率的阈值,分别采用冒泡法、均值法、中位数法。其中中位数法的实验效果更优。以中位数确定的怀疑率阈值进行实验,检测率达到94.1%,误判率为31.5%。This paper expounds a DEM gross error detection method based on BP neural network.In order to reduce the misjudgment rate of the algorithm,the parameter of suspicion rate is added,so that the gross error judgment of a certain point can be based on the detection results of multiple regions.To determine the threshold of suspicion rate,bubble method,mean method,and median method were used respectively.The median method has better experimental results.The experiment was conducted with a suspicion rate threshold determined by the median,and the detection rate reached 94.1%,with a misjudgment rate of 31.5%.

关 键 词：深度学习 BP神经网络 数字高程模型 粗差探测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]