Analysis And Experiment Of Sampling Technology Of Gravity And Magnetic Anomalies

Gravity and magnetic exploration is an important geophysical exploration method for geological structure research and mineral resources investigation,and data collection and inversion are the basic work content of the above methods.In the process of inversion,if the amount of data will lead to long processing time and occupy a lot of memory space.Assuming that the number of data is n = 100 × 100 and the number of models is m = 100 × 100 × 50,the conventional inversion method needs to store a 20 GB kernel function matrix,which cannot be stored and calculated for ordinary computers.With the development of the collection technology of the gravity and magnetic exploration,the amount of data collected is often massive,which requires the data to be sampled before inversion,with the purpose of reducing the amount of data as much as possible while retaining the original data information.How to sample to achieve the best results requires sampling experiment research.This article first sorts out the sampling methods proposed in recent years,and selects uniform sampling method,based on discrete sampling method,and adaptive sampling method to compare and analyze.On this basis,the author proposes a new sampling method—the minimum error sampling method.Then the author designed one-dimensional and two-dimensional different original data,and carried out the sampling number test,the sampling position test,and the sampling method test in turn,solve the three basic questions of how many sampling points are appropriate,where to sample is suitable,and which sampling method works best.Finally,two special curves of straight line and semicircle curve are used to test the minimum error sampling method.Research shows that the sampling effect is best when the sampling frequency is 3 times the highest frequency of the original data,any point is taken on the entire curve or surface,and the minimum error sampling method is used,at this time,the sampling points are densely distributed in the high frequency abnormal area of the data,and sparsely distributed in the low frequency non-abnormal area of the data,basically retaining the original data information volume,which is in line with the ideal sampling situation.