Study On Aeromagnetic Total Field Compensation Technology Of Rotor UAV

Aeromagnetic exploration is an important part of airborne geophysical exploration.It is a geophysical method for detecting changes in the geomagnetic field and investigating underground geological structures through a magnetometer equipped on an aircraft.Traditional aeromagnetic detection platforms are mostly manned large fixed-wing aircraft or helicopters.With the emergence of small high-precision aviation magnetometers and development of UAV technology recently,it is possible to implement aeromagnetic survey based on the UAV platform.UAV-based aeromagnetic detection has the advantages of high efficiency,low cost,and few flight restrictions.However,the UAV platform will cause magnetic interference to the magnetic sensor during the data collection process.Therefore,the research on the compensation algorithm suitable for the UAV platform is of great significance for improving the quality of UAV aeromagnetic data.In this paper,our research subject is rotor UAV platform.First,we build an aeromagnetic total field measurement system for the rotor UAV combining a highprecision aviation magnetometer.Then,according to features such as unstable flying attitude and special rotor noise of the UAV platform,and classical filtering model,we complete the modeling for magnetic interference of rotor UAV aeromagnetic detection platform.On this basis,this paper proposes two improved methods of aeromagnetic total field compensation for rotor UAV to solve the multicollinearity problem of interference model,thereby improve the accuracy and robustness of model.Analyzing UAV aeromagnetic model with unsupervised learning theory,we propose a compensation algorithm based on deep autoencoder(DAE).DAE uses gradient descent back-propagation algorithm to search the most changing direction in interference data to find features.Then,we extract features considering the coefficient matrix compressed by the special network structure so as to reduce the multicollinearity caused by the model itself and outlier noise,and improve the performance of the regression algorithm.We propose a generalized regression neural network(GRNN)method with strong generalization ability combining linear regression and neural network.GRNN is derived from the probability density function,it does not require back-propagation iterative operation,and the calculation is less time-consuming.The appropriate smoothing factor in the probability density function and the effective distribution of the training set will effectively improve the accuracy and generalization ability of the rotor UAV aeromagnetic compensation model.The aerial magnetic survey results of the mining area show that the improved compensation algorithm proposed in this paper can effectively eliminate and retain the magnetic anomaly signal,and the range of the magnetic anomaly after various corrections is clearer.On the basis of algorithm verification,the author develops compensation application software for the compensation processing of actual aeromagnetic measurement data.