基于神经网络的新息自适应卡尔曼滤波在高速公路变形监测中的应用  

Application of innovation adaptive Kalman filter based on BP neural network in expressway deformation monitoring

作　　者：韩亚坤[1,2,3] 文鸿雁[1,2,3] 郭雷[1,2,3] 王清涛 谢劭峰[1,2,3] 孔令帅[1,2,3] HAN Ya-kun WEN Hong-yan GUO Lei WANG Qing-tao XIE Shao-feng KONG Ling-shuai(a. Guangxi Key Laboratory of Spatial Information and Geomatic b. College of Geomatics and Geoinformatio c. Guangxi Scientific Experiment Center of Mining, Metallurgy and Environment, Guilin University of Technology, Guilin 541004, Chin)

机构地区：[1]桂林理工大学广西空间信息与测绘重点实验室,广西桂林541004 [2]桂林理工大学测绘地理信息学院,广西桂林541004 [3]桂林理工大学广西矿冶与环境科学实验中心,广西桂林541004

出　　处：《桂林理工大学学报》2017年第1期136-139,共4页Journal of Guilin University of Technology

基　　金：国家自然科学基金项目(41461089);广西自然科学基金项目(2014GXNSFAA118288);广西“八桂学者”岗位专项经费项目;广西空间信息与测绘重点实验室项目(桂科能140452402;130511402);广西矿冶与环境科学实验中心项目(KH2012ZD004);广西研究生教育创新计划项目(YCSZ2014151;YCSZ2012083)

摘　　要：实际应用中卡尔曼滤波的动态噪声和观测噪声往往不是固定的,提出运用新息自适应卡尔曼滤波法则对噪声进行实时估计。同时,为了克服新息自适应卡尔曼滤波要求系统模型过分精确的缺点,基于神经网络的新息自适应卡尔曼滤波算法,通过使用神经网络进行误差补偿来提高滤波性能,并运用于高速公路变形监测中。结果表明:该模型与原始观测值的残差全部降到了0.25 mm以下,说明该模型精度很高,在实际应用中具有一定的可行性。As the dynamic noise and observation noise are non-fixed in practical application, innovation adap-tive Kalman filter is used to estimate these noises in real time, to cope with the innovation adaptive Kalman fil-ter which needs a precise model system in practical application. So this paper uses error compensation of BP neural network to improve the filter performance. The result shows that all residuals are not more than 0. 25 mm, and the precision of this model is much higher than before, a feasible effect in practical application.

关 键 词：新息 神经网络 补偿 KALMAN滤波 高速公路 变形监测 

分 类 号：P258[天文地球—测绘科学与技术]