A Hardware–Software Cooperative Method for Transient Electromagnetic Data Noise Suppression

In the shallow target detection of urban underground space, transient electromagnetic (TEM) measurement data are easily affected by various noises such as power frequency noise and Gaussian noise, leading to a low signal-to-noise ratio (SNR) in late-stage data, which impact the identification of targets. To address this issue, this article proposes a hardware--software collaborative method for TEM data noise suppression. The method involves adding an optional band-stop filter (BSF) and an antialiasing filter in the hardware sensor circuit to filter out power frequency and high-frequency power noise interference. In the software, the optimized variational mode decomposition (VMD) using the northern goshawk optimization (NGO) algorithm is employed for noise recognition and elimination. First, a power frequency notch filter and an antialiasing filter are set up in the hardware reception conditioning circuit. Then, the collected TEM signals are processed using the NGO to obtain the optimal parameter combination (K, $\alpha$ ) for VMD. Based on the optimized parameters, the signal undergoes adaptive VMD decomposition into K modal components, which are then divided into effective signal components and noise components. To further improve the SNR, the effective signal components are reconstructed to produce the denoised TEM signal. Simulation experiments demonstrate that this method performs well in noise processing for TEM signals, while field experiments confirmed its effectiveness in real-world applications, successfully removing noise from collected data.