几种改进的BP神经网络在GPS高程拟合中的应用比较

Comparison of Several Improved BP Neural Network in GPS Height Fitting

为了提高GPS高程拟合的精度和效率,本文分别利用几种常见的算法对标准BP神经网络进行了改进,结合控制网测量实例,分别利用外符精度、内符精度、测试步长以及运行时间4个指标对拟合结果进行评价。结果表明,不同改进算法的设计分别能有效提高拟合的精度和效率,在实际的工程实践中,可以根据具体拟合要求采取合适的算法对BP神经网络进行改进。

In order to improve the efficiency and accuracy of GPS height fitting.In this paper,the use of several common algorithms of standard BP neural network was improved,combined with the control network measurement examples,respectively using the match precision and match accuracy,the testing step size and running time four indicators,evaluates the results of fitting.The results show that the improved algorithm design can effectively improve the fitting precision and efficiency,in the actual engineering practice,the project can according to the specific fitting to take appropriate algorithm to improve the BP neural network.