Magnetic anomaly detection based on fast convergence wavelet artificial neural network in the aeromagnetic field

•FC-W-ANN and OBF combined detector realizes low SNR ratio MAD.•The error quotient updates the learning rate to realize high-speed convergence.•The error quotient gradient can ensure the stability of the network.•FC-W-ANN denoises signals without studying noise frequency. The orthogonal basis functions (OBFs) detector is a detection method widely used in the aerial magnetic measurement. However, OBFs detector works ineffectively under non-Gaussian noise and colored noise. This paper proposes an OBFs detector based on fast convergence wavelet artificial neural network (FC-W-ANN), which can detect abnormal magnetic signals under low SNR. First, the magnetic anomaly signal is modelled. Then, the learning rate is corrected by the iterative error convergence rate under the stability of the network. Finally, the improved network is combined with the OBFs detector to detect magnetic abnormal signals. From the simulation and experimental results, the reconstructed signal of the new method has a higher SNR (SNR = 10.01) compared with OBFs (SNR = -0.21) and OBFs based on the wavelet artificial neural network (W-ANN; SNR = 9.59). Furthermore, the statistical methods are used to analyze FC-W-ANN and W-ANN, showing that FC-W-ANN has higher training accuracy and better stability.