Search

CN-121995344-A - GPR echo signal denoising method based on improved WTD and VMD

CN121995344ACN 121995344 ACN121995344 ACN 121995344ACN-121995344-A

Abstract

The invention relates to the technical field of geophysical exploration signal processing, and discloses a GPR echo signal denoising method based on improved WTD and VMD. The method comprises the steps of adaptively optimizing decomposition mode numbers and penalty factors of VMDs by using an HHO algorithm, carrying out 2D-VMD decomposition on original GPR data by using optimal parameters to obtain a plurality of IMFs, dividing the IMFs into signal IMFs and noise IMFs by using correlation coefficients, optimizing and improving wavelet threshold parameters by using a PSO algorithm, processing the noise IMFs, and removing direct waves by using a mean value method after reconstruction. The invention realizes the self-adaptive high-precision denoising of the GPR signal through a dual parameter optimization mechanism, an improved threshold function with continuity and unbiasedness and a differential threshold processing strategy, effectively improves the signal-to-noise ratio, reserves weak target signals, remarkably inhibits direct wave interference, and has the advantages of strong parameter self-adaptation capability, high denoising precision, strong engineering practicability and the like.

Inventors

  • SONG HUAJUN
  • ZHAO JIAQI
  • LI SHIBAO

Assignees

  • 中国石油大学(华东)

Dates

Publication Date
20260508
Application Date
20260410

Claims (10)

  1. 1. The GPR echo signal denoising method based on improved WTD and VMD is characterized by comprising the following steps: step 1, acquiring two-dimensional noisy GPR data, and preprocessing the two-dimensional noisy GPR data to obtain one-dimensional noisy data for parameter optimization; Step 2, optimizing the one-dimensional noisy data by using an HHO algorithm and taking envelope entropy minimization as a target to obtain an optimal VMD parameter combination, wherein the optimal VMD parameter combination comprises an optimal decomposition mode number K and an optimal penalty factor alpha; Step 3, performing two-dimensional variation modal decomposition on the original two-dimensional noisy GPR data by adopting the optimal decomposition modal number K and the optimal penalty factor alpha to obtain K IMFs; Step 4, calculating correlation coefficients of each IMF and original two-dimensional noisy GPR data, and dividing K IMFs into signal IMFs and noise IMFs according to a preset threshold; Step 5, optimizing and improving parameters of WTD (WTD) by using a PSO (generalized cross validation) rule as an adaptability function for each noise IMF and utilizing the optimized parameters to perform improved wavelet threshold denoising treatment on the noise IMFs to obtain the treated noise IMFs; step 6, overlapping and reconstructing the processed noise IMF and the signal IMF to obtain a reconstructed signal; and 7, performing direct wave removal processing on the reconstructed signal to obtain a final denoised signal.
  2. 2. The GPR echo signal denoising method based on modified WTD and VMD according to claim 1, wherein step 2 comprises: Initializing parameters of an HHO algorithm, including population scale, maximum iteration times and search space boundaries, and randomly generating an initial population, wherein each individual represents a candidate [ K, alpha ] combination; 2.2, for each individual in the population, performing VMD decomposition on the one-dimensional noisy data by using the corresponding [ K, alpha ] to obtain K IMFs, calculating the envelope entropy of each IMF, and taking the minimum envelope entropy value as the fitness value of the individual; and 2.3, iteratively updating the population position according to the exploration and development strategy of the HHO algorithm until the maximum iteration times or the convergence of the fitness value are reached, and outputting the global optimal individual position, namely the optimal decomposition modal number K and the optimal penalty factor alpha.
  3. 3. The GPR echo signal denoising method based on the modified WTD and the VMD according to claim 1, wherein in the step 4, the correlation coefficient is calculated as follows: ; Wherein, the Is the first Correlation coefficients between the individual IMF components and the original two-dimensional noisy GPR data, Is raw two-dimensional noisy GPR data; mean value of raw two-dimensional noisy GPR data; Is the first The IMF components; Is the first An average of the IMF components; Is the number of sampling points.
  4. 4. The GPR echo signal denoising method based on improved WTD and VMD according to claim 1, wherein in step 4, the preset threshold value is 0.5, IMF with a phase relation greater than the threshold value is determined as signal IMF, and IMF with a phase relation less than the threshold value is determined as noise IMF.
  5. 5. The GPR echo signal denoising method based on improved WTD and VMD according to claim 2, wherein in step 2.2, the calculation formula of envelope entropy is as follows: ; Wherein, the In order to normalize the envelope signal, Is the first The envelope entropy of the individual IMFs, Is the length of the envelope sequence, i.e. the signal length.
  6. 6. The GPR echo signal denoising method based on modified WTD and VMD according to claim 2, wherein in step 2.3, the energy escapes The strategy for controlling the HHO algorithm switches, the values of which decay with iteration, are calculated as follows: ; Wherein, the For a random initial value of the value, For the current number of iterations, The maximum iteration number; When (when) When entering the exploration stage The utilization phase is entered.
  7. 7. The GPR echo signal denoising method based on the modified WTD and the VMD according to claim 1, wherein the method of step 5 specifically comprises: Step 5.1, carrying out multi-scale discrete wavelet decomposition on each noise IMF to obtain high-frequency coefficients of each layer; Initializing PSO algorithm parameters, and setting search dimensions, wherein the search dimensions correspond to thresholds in horizontal, vertical and diagonal directions of each layer and threshold function adjusting factors of each layer; iteratively updating the particle positions by taking the minimized GCV function as the fitness function, and optimizing to obtain the optimal threshold values of each layer in the horizontal, vertical and diagonal directions and the optimal adjustment factors of each layer; step 5.3 improvement of thresholding Substituting optimal threshold values of horizontal, vertical and diagonal directions of each layer obtained by PSO optimization into an improved wavelet threshold function and using the optimal adjustment factors of each layer in threshold processing of wavelet high-frequency coefficients of the corresponding layer; Step 5.4 wavelet reconstruction And finally, carrying out wavelet reconstruction on the high-frequency coefficient and the low-frequency coefficient after the threshold processing to obtain the processed noise IMF.
  8. 8. The GPR echo signal denoising method based on modified WTD and VMD of claim 7, wherein in step 5.2, the fitness function is as follows: ; Wherein, the Is a number of coefficients of the high-frequency wavelet, For the number of wavelet coefficients that are thresholded to be set to 0, As the high-frequency coefficients of the wavelet, As a result of the threshold value being set, To estimate wavelet coefficients.
  9. 9. The GPR echo signal denoising method based on modified WTD and VMD according to claim 7, wherein the modified wavelet threshold function in step 5.3 is expressed as follows: ; Wherein, the As the high-frequency coefficients of the wavelet, As a result of the threshold value being set, In order to estimate the wavelet coefficients, Is a regulatory factor, and 。
  10. 10. The GPR echo signal denoising method based on the modified WTD and the VMD according to claim 1, wherein the method of step 7 is as follows: ; Wherein, the Reconstructed signal in The pixel value at which it is located, Is that Pixel values of the points after denoising, For the number of lines, i.e. the number of temporal samples, The number of columns, i.e., the number of scan tracks.

Description

GPR echo signal denoising method based on improved WTD and VMD Technical Field The invention relates to the technical field of geophysical exploration signal processing, in particular to a GPR echo signal denoising method based on improved WTD and VMD. Background Ground penetrating radar (group PENETRATING RADAR, GPR) is a geophysical prospecting device for detecting an underground structure by using high-frequency electromagnetic waves, and is widely applied to the fields of road defect detection, tunnel lining evaluation, urban underground space detection and the like by virtue of the advantages of short detection period, high resolution, nondestructive detection and the like. However, in actual detection, the GPR echo signal is often interfered by various noises, mainly including direct waves (composed of antenna coupled waves and earth surface direct waves), system electronic noises (such as amplifier noises and sampling quantization noises), background clutter (caused by medium nonuniformity) and the like. These noise results in a reduced signal-to-noise ratio of the GPR signal, and the target reflection characteristics are overwhelmed, severely affecting subsequent data interpretation. In prior studies, wavelet threshold denoising (Wavelet Threshold Denoising, WTD) and variational modal decomposition (Variational Mode Decomposition, VMD) are two types of commonly used GPR signal denoising methods. The WTD suppresses noise by thresholding the high frequency coefficients of the signal using wavelet multi-resolution analysis and time-frequency localization characteristics, and can compromise noise suppression and signal detail retention to some extent. However, the conventional WTD method has the following defects that (1) the hard threshold function has the discontinuity which is easy to cause reconstruction oscillation, the soft threshold function is continuous but can cause excessive smoothing of signals, (2) the threshold setting is inaccurate, the high-frequency wavelet coefficient often contains effective signals, the conventional method carries out threshold calculation by regarding the high-frequency wavelet coefficient as noise, the threshold is higher or lower, the denoising effect is affected, and (3) the threshold processing strategy is single, the conventional method uses the same threshold for the high-frequency coefficients with different decomposition layers and directions, the characteristic that the noise decays along with the decomposition layers is not considered, and the phenomenon of excessive suppression is easy to occur. The VMD is used for decomposing the signal into a plurality of self-adaptive mode functions to realize efficient frequency component separation, so that the mode aliasing problem in Empirical Mode Decomposition (EMD) is effectively solved, and the VMD gradually becomes another important means for denoising the GPR signal. However, the decomposition quality of the VMD is highly dependent on the choice of two key parameters, the decomposition modality number and the penalty factor. The traditional method relies on manual experience or trial and error to determine parameters, so that decomposition effect is difficult to ensure, and problems of mode underdecomposition, redundant modes or mode aliasing and the like are easy to occur, so that noise suppression and signal detail reservation are difficult to be compatible. In summary, the existing GPR signal denoising method has not achieved an effective balance among parameter adaptability, noise suppression precision and signal detail retention, and a joint denoising method capable of achieving parameter adaptive optimization, giving consideration to noise suppression and signal fidelity, and effectively removing direct waves is needed. Disclosure of Invention In order to solve the technical problems, the invention provides a GPR echo signal denoising method based on improved WTD and VMD, so as to achieve the purposes of adaptively optimizing denoising parameters, improving signal-to-noise separation precision, retaining weak target signal details and effectively inhibiting direct wave interference. In order to achieve the above purpose, the technical scheme of the invention is as follows: A GPR echo signal denoising method based on improved WTD and VMD comprises the following steps: step 1, acquiring two-dimensional noisy GPR data, and preprocessing the two-dimensional noisy GPR data to obtain one-dimensional noisy data for parameter optimization; Step 2, optimizing the one-dimensional noisy data by using an HHO algorithm and taking envelope entropy minimization as a target to obtain an optimal VMD parameter combination, wherein the optimal VMD parameter combination comprises an optimal decomposition mode number K and an optimal penalty factor alpha; Step 3, performing two-dimensional variation modal decomposition on the original two-dimensional noisy GPR data by adopting the optimal decomposition modal number K and th