【摘要】 提出基于高程异常差的GPS跨海水准新方法,给出新方法的基本原理和数据处理方案,对用垂线偏差、神经网络法、支持向量机求解高程异常差方法的精度和实用范围作计算分析。结果表明:基于垂线偏差的高程异常差拟合方法适用于距离约5 km的高程传递。基于支持向量机的高程异常差拟合方法,在平原地区,当GPS和水准重合点的点数较多、点位分布较合理时,跨距可以达到30 km。而基于神经网络进行高程异常差拟合适用于距离约10 km的高程传递。更多还原

【Abstract】 The new methods of GPS cross-ocean leveling based on height anomaly difference are put forward.The theory and data processing design of new methods obtaining elevation anomaly difference using vertical deflection,the SVM(support vector machine),and the neural network respectively are given.Through the practical analysis,the height anomaly difference fitting method based on vertical deflection is suitable for the elevation transmission within the distance of 5km or so.As for the height anomaly difference fitting method based on modeling,if used in plain area,and taken the SVM(support vector machine) as tools to fit the height anomaly difference,its cross-distance can extend as long as 30 km if there are enough points both used as GPS and leveling points,and the distribution is reasonable.Adopting neural network to fit the height anomaly difference,it would only be suitable for the elevation transmission of 10km or so.更多还原