整区拟合似大地水准面的BP神经网络方法  

Study of whole fitting quasi-geoid using BP neural network method

作　　者：宋雷[1,2] 胡伍生[1] 

机构地区：[1]东南大学交通学院,江苏南京210096 [2]山东交通学院土木工程学院,山东济南250357

出　　处：《辽宁工程技术大学学报（自然科学版）》2013年第2期161-165,共5页Journal of Liaoning Technical University (Natural Science)

基　　金：国家高技术研究发展计划(863)资助项目(2007AA12Z228);山东省优秀中青年科学家科研奖励基金资助项目(BS2010SF019)

摘　　要：针对组合法似大地水准面精化过程中,传统的分区曲面拟合法存在模型代表性误差和及分区间的平滑连接问题,提出整区拟合似大地水准面的BP神经网络方法.利用某区域的重力似大地水准面模型和GPS/水准数据,将BP神经网络似大地水准面整区拟合法与整区曲面拟合法和分区曲面拟合法进行比较.研究结果表明:在较大区域和两类似大地水准面差别呈不规则的情况下,BP神经网络方法有效地减小了拟合模型的代表性误差,整区BP神经网络拟合法明显地较整区曲面拟合法和分区曲面拟合法提高了拟合结果的内、外符合精度.为区域高精度、高分辨率似大地水准面精化过程中似大地水准面整区拟合提供了有效的方法.Due to the model error and smooth joint problem of fitting quasi-geoid using polynomial fitting method, a new method of whole fitting quasi-geoid using BP neural network was proposed in this study. Using the gravimetric quasi-geoid and GPS/leveling data in an area, the BP neural network method was compared with the whole and divisional polynomial surface fitting method. The results show that the new method could reduce the model error in the case of a bigger area and the anomalous difference between two kinds of quasi-geoids. The new method can obviously improve the inner and outer precision of fitting two kinds of quasi-geoids than the whole and divisional polynomial surface fitting method. The BP neural network is a feasible fitting method for finishing local high precise and high resolution quasi-geoid.

关 键 词：整区拟合 似大地水准面 BP神经网络 曲面拟合法 高程异常 GPS 水准 大地高 正常高 

分 类 号：P227[天文地球—大地测量学与测量工程]