针对用小波框架表示GPS速度场可能会产生病态问题,该文提出了采用吉洪诺夫正则化方法及相应的3种正则化参数选择方法(广义交叉检验法、L曲线法和留一交叉验证法)进行模型求解。该方法通过引入合适的正则化参数及正则化矩阵的方式,来克服GPS速度场球面小波模型难于得到惟一解的问题。基于中国地壳运动观测网络两个局部区域的GPS速度场数据和亚洲太平洋地区地球动力学计划局部区域的GPS速度场数据的实验结果表明:使用该方法可以得到模型的稳定解,且以外部检核均方误差最小为准则时,3种正则化参数选择方法获得的解的精度水平相当。
Using wavelet modeling for the GPS velocity field may lead to an ill-posed problem.The Tikhonov regularization and regularization parameter selection methods(GCV,L-curve and OCV)were introduced in the paper.The Tikhonov regularization was used to impose a unique solution of the Spherical Wavelet Model of GPS Velocity Fields by introducing aproper parameter and matrix.The effect of regularization parameter selection methods applied in the spherical wavelet model of GPS velocity fields with CMONC and APRGP GPS velocity fields data sets were analyzed based on the Tikhonov regularization.The results showed that the Tikhonov regularization imposed a unique solution,and the accuracy of the optimized solutions(selected by minimizing MSE)with 3types of regularization parameter selection methods was at the same level.