Abstract:Existing noise removal processes for airborne electromagnetic (AEM) data generally consist of several steps, with each using a specific method to remove a specific type of noise. To improve the efficiency of AEM denoising and reduce the impact of the subjective judgment of the operators on the processing results, we have adopted a deep learning method based on a denoising autoencoder (DAE), which enables in one single processing step the removal of multisource noise. The most common noise sources in AEM data, including motion-induced noise, nearby or moderately distant sferics noise, power-line noise, and background electromagnetic noise, will be combined with a large number of simulation responses to build a training set. The data in the training set will be used to train the deep learning DAE neural network so that the neural network could fully learn the respective characteristics of the signal and noise and further effectively distinguish the AEM response signal (useful signal) from the above noise. The field data were processed using this method, and the processing results were compared with those obtained using traditional methods. The comparison test revealed that this method is helpful to reduce the influence of subjective factors on the quality of data results and compress the entire AEM data processing time.

What problem does this paper attempt to address?

This paper aims to solve the problem of multi - source noise removal in airborne electromagnetic data. Specifically, the existing airborne electromagnetic (AEM) data denoising process usually consists of multiple steps, and each step uses a specific method to remove a specific type of noise. This method is not only inefficient but also easily influenced by the subjective judgment of the operator. In order to improve the efficiency of AEM data denoising and reduce the influence of the operator's subjective factors on the processing results, this paper proposes a denoising auto - encoder (DAE) method based on deep learning, which can remove multi - source noise in a single processing step. ### Main problems 1. **Complexity of multi - source noise**: - Common noise sources in airborne electromagnetic data include motion - induced noise (MIN), nearby or medium - distance sferics noise, power line noise, and background electromagnetic noise. These noise types are different, and traditional denoising methods need to be processed separately, resulting in a complex processing flow and low efficiency. 2. **Subjective influence of the operator**: - In many cases, existing denoising methods require the operator to manually set key parameters, which not only increases the requirement for the operator's personal experience but also makes the processing results easily influenced by subjective factors. ### Solutions - **Deep learning method**: - This paper proposes a deep learning method based on the denoising auto - encoder (DAE). By training a neural network to learn the characteristics of signals and noise, it can effectively separate useful signals from noise. - **Construction of training set**: - Combine a large amount of simulated responses and actual noise data to construct a training set for training the DAE neural network. - **Network structure**: - The number of nodes in both the input layer and the output layer is 500, corresponding to the selected signal length (500 sampling points within about 4 milliseconds). - The number and number of nodes in the hidden layer are set according to the signal characteristics and noise characteristics, and usually, the hidden layer does not exceed four layers. - **Activation function**: - The tanh function is selected as the activation function because it is symmetric near the origin and the output range is between - 1 and 1, which is suitable for processing continuous signal data. - **Loss function**: - The mean square error (MSE) is selected as the loss function because the output is a noise - free AEM signal, which is a real number and has an unbounded value. ### Experimental verification - **Experimental results**: - By processing actual airborne electromagnetic data and comparing with the results of traditional methods, the effectiveness and reliability of this method are verified. - The experimental results show that the deep learning method based on DAE can effectively reduce the influence of subjective factors on data quality and significantly compress the time of the entire AEM data processing. In conclusion, by introducing deep learning technology, this paper proposes an efficient and reliable multi - source noise removal method, which lays a good data foundation for subsequent inversion and interpretation work.

Removal of multisource noise in airborne electromagnetic data based on deep learning

CG-DAE: A noise suppression method for two-dimensional transient electromagnetic data based on deep learning

Deep Denoising Autoencoder for Seismic Random Noise Attenuation

Comparative Analysis of Deep Learning and Traditional Airborne Electromagnetic Data Processing: A Case Study

Distributed Acoustic Sensing Vertical Seismic Profile Data Denoiser Based on Convolutional Neural Network

Efficient Seismic Data Denoising via Deep Learning With Improved MCA-SCUNet

De-noising of Transient Electromagnetic Data Based on the Long Short-Term Memory-Autoencoder

Denoising distributed acoustic sensing (DAS) data using unsupervised deep learning

Denoising Distributed Acoustic Sensing Data Using Unsupervised Deep Learning

Noise reduction in the spectral domain of hyperspectral images using denoising autoencoder methods

A Denoising Method Based on DDPM for Radar Emitter Signal Intra-Pulse Modulation Classification

Removing Instrumental Noise in Distributed Acoustic Sensing Data: A Comparison Between Two Deep Learning Approaches

Removing multiple types of noise of distributed acoustic sensing seismic data using attention-guided denoising convolutional neural network

Denoising of Transient Electromagnetic Data Based on the Minimum Noise Fraction-Deep Neural Network

Electrostatic Signal Self-Adaptive Denoising Method Combined with CEEMDAN and Wavelet Threshold

A Deep-Learning-Based Denoising Method for Multiarea Surface Seismic Data

Unpaired training: Optimize the seismic data denoising model without paired training data

A multi-task learning network based on transformer network for airborne electromagnetic detection imaging and denoising

Denoising Controlled-Source Electromagnetic Data Using Least-Squares Inversion

A diffusion model-based framework USDDM for random noise elimination of seismic signal

Multi-Scale Progressive Fusion Attention Network Based on Small Sample Training for DAS Noise Suppression