基于EKF的神经网络在GPS高程转换中的应用

Application of neural network method with the extended Kalman filter to conversion of GPS height

GPS测量技术作为一种快速、高精度的测量手段已广泛地应用于各种建设工程中,但GPS提供的高精度的大地高,不是工程中需要的正常高,需要进行拟合转换成正常高才能应用于工程建设。基于神经网络的GPS高程转换方法能减少人为构建数学模型所导致的误差,是一种较好的方法,但也存在一些不足。将扩展Kalman滤波与神经网络结合起来,发挥两者的优势,克服不足,建立了基于EKF(扩展Kalman滤波)的神经网络GPS高程转换模型,并通过实例应用,结果表明该模型用于GPS高程转换精度有较大提高。

GPS as the fast and high accuracy survey method has widely applied in all construction projects.The GPS height is geodetic height,not normal height,which need transform to normal height.The conversion of GPS height based on neural network is a good method which reduces the error of the artificial mathematical model.But it also has some insufficiencies.With the superiority combination of the extended Kalman filter and neural network,the GPS height conversion method of neural network with the extended Kalman filter is proposed.Experiments in the GPS height conversion indicate that the accuracy of height conversion has been greatly improved.