ȫ���������ݶ������к��и�Ƶ�ɷ֣�Ϊ����ȥ����Ƶ������ͬʱ������dz��С�쳣����Ϣ������������С����ԭ����������ݶȴ�ֱ������������������Լ���µij��������顣������������ż��㴰�ڵ�����ȥ��Ч��Խ��Խ�ã����ڱ�������Ƶ�ģ�ͣ������㴰������ 32 × 32 ʱ������������99%���ϵ�������ȥ������������ V xz �� V yz �У��ֱ��д�Լ40%��50% ��������ȥ��; ͬʱ��dz��С�쳣����Ϣ����ȫ�����档

There are high- frequency contents in full tensor gravity gradient data�� To preserve shallow small- scale features while removing the high-frequency noise��we solve the overdetermined equation by least squares on constraints of gravity data and vertical components of full tensor gravity gradient data�� The results of the experiment demonstrate that noise reduction results become better when calculation windows increase�� When calculation win- dows are 32 ×32��more than 99% noise can be removed from the gravity data��respectively about 40% and 50% noise can be removed from the data V xz �� V yz �� At the same time��the method do not lose high frequency signal��