Wavelet noise reduction and gradient feature extraction method for buried pipeline welding seam positioning

Abstract

The invention discloses a wavelet noise reduction and gradient feature extraction method for buried pipeline welding seam positioning, and relates to the technical field of nondestructive testing. The method comprises: acquiring an original magnetic field signal of a pipeline area through a magnetic field acquisition device; wavelet transformation is carried out on the signal, wavelet reconstruction is carried out after a noise coefficient is suppressed through threshold processing, and a denoised signal is obtained; finally, gradient feature extraction is conducted, specifically, the gradient of the denoised signal is calculated and standardized, extreme points in the standardized gradient are recognized by setting a threshold value to serve as welding seam feature points, and therefore accurate positioning of the welding seam position is achieved. According to the method, non-periodic noise is effectively filtered through wavelet transformation, gradient analysis is combined to strengthen and extract the magnetic field abrupt change features of the welding seam, the problem that a traditional frequency domain method is poor in adaptability in buried pipeline welding seam positioning is solved, and the method has the advantages of being good in filtering effect, accurate in feature extraction and high in robustness.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology for special equipment, specifically to a method for magnetic field signal filtering and feature extraction for locating butt welds of buried pipelines. Background Technology

[0002] Currently, there is a lack of mature, precise positioning systems for buried pipeline weld seams on the market. In similar fields such as metal magnetic memory detection, signal processing typically employs frequency domain filtering and frequency domain feature extraction methods. However, in the buried pipeline weld seam detection scenario addressed by this invention, background interference signals and weld seam feature signals often exhibit non-periodic and highly uncertain characteristics, making it difficult for traditional frequency domain analysis methods to effectively separate and extract them.

[0003] Therefore, existing technologies are insufficient to meet the requirements for accurate noise reduction and effective feature extraction of magnetic field signals during the positioning of buried pipeline butt welds. There is an urgent need for a signal processing method that can adapt to non-periodic and random interference. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the above-mentioned background technology and provide a wavelet noise reduction and gradient feature extraction method for locating buried pipeline welds, so as to solve the problem that the existing technology cannot effectively handle non-periodic and uncertain interference signals and feature signals, and realize accurate filtering and feature extraction of magnetic field signals during the locating of buried pipeline butt welds.

[0005] The technical solution of the present invention is as follows: A wavelet noise reduction and gradient feature extraction method for locating weld seams in buried pipelines, which relies on a magnetic field acquisition device and a data processing device, includes the following steps: (1) Magnetic field signal acquisition steps: Use a magnetic field acquisition device to move along the pipeline axis to obtain the original magnetic field signal sequence Data_In[n] containing the weld area, where n is the number of sampling points.

[0006] (2) Wavelet denoising process: Perform wavelet transform on the original signal, suppress the wavelet coefficients corresponding to the noise through threshold processing, and then perform inverse wavelet reconstruction to obtain the denoised signal sequence Data_Out[n], n=0,1,...,N-1.

[0007] (3) Gradient feature extraction steps: Calculate the gradient sequence of the denoised signal and standardize it; set the feature extraction threshold, identify the points whose standardized gradient values ​​exceed the threshold as feature points (outliers), and thus determine the precise location of the differential magnetic field signal of the weld.

[0008] Compared with the prior art, the beneficial effects of the present invention are: Excellent non-periodic noise suppression capability: Employing wavelet denoising technology, it can adaptively separate signals and noise in the time and frequency domains, and is particularly adept at handling non-stationary and non-periodic noise that traditional frequency domain filtering cannot handle.

[0009] Accurate feature extraction and localization: Gradient transformation and standardization enhance the significance of magnetic field abrupt changes at the weld. Combined with threshold determination, feature points can be accurately located, providing a reliable basis for the subsequent physical localization of the weld.

[0010] The algorithm is robust: the method flow is clear, and the parameters (such as wavelet basis, noise reduction threshold, and feature extraction threshold) can be optimized according to the statistical characteristics of the actual signal, making it highly adaptable. Detailed Implementation

[0011] The present invention will now be described in detail with reference to the embodiments.

[0012] System components: Magnetic field acquisition device: used to acquire spatial magnetic field signals above buried pipelines, usually a magnetic sensor array or a single-point scanning sensor.

[0013] Data processing device: connected to the magnetic field acquisition device via wired or wireless connection, with embedded software program running the algorithm of this invention, responsible for signal noise reduction, feature extraction and result display.

[0014] The core algorithm of this method consists of two main stages: Phase 1: Wavelet Denoising Wavelet decomposition: The original magnetic field signal Data_In[n] is subjected to discrete wavelet transform to obtain the wavelet coefficients C[j, k] at each scale. The formula is as follows: C[j, k] = Σ (Data_In[n] · ψ_{j, k}(n)), where ψ_{j, k}(n) is the selected wavelet basis function, j is the scale, and k is the translation amount. The selection of the wavelet basis function ψ (such as the db series, sym series, etc.) can be optimized according to the characteristics of the magnetic field signal to maximize the signal-to-noise separation effect.

[0015] Thresholding: A threshold T is set, and the wavelet coefficients are subjected to thresholding to remove noise. Taking the hard thresholding method as an example: C_thresh[j, k] = { C[j, k], if |C[j, k]| > T; 0, otherwise} The threshold T is set based on the statistical characteristics of the magnetic field signal noise level (such as the general threshold method, Stein unbiased risk estimation method, etc.), and the optimal value can be determined through experiments or simulations.

[0016] Wavelet reconstruction: The inverse wavelet transform is performed using the processed wavelet coefficients C_thresh[j, k] to reconstruct the denoised time-domain signal Data_Out[n].

[0017] Phase Two: Gradient Feature Extraction Gradient calculation: Calculate the first-order difference (gradient) of the denoised signal Data_Out[n] to obtain the gradient sequence G[n], which characterizes the instantaneous rate of change of the signal. G[n] = Data_Out[n+1] - Data_Out[n], where n = 0, 1, ..., N-2.

[0018] Gradient standardization: Calculate the mean μ and standard deviation σ of the gradient sequence, and standardize it to obtain a sequence G_norm[n] with a mean of 0 and a standard deviation of 1. Calculate the mean: μ = (1 / (N-1)) * Σ G[n], summing over n from 0 to N-2.

[0019] Calculate the standard deviation: σ = sqrt( (1 / (N-1)) * Σ (G[n] - μ)^2 ), summing over n from 0 to N-2.

[0020] Standardization: G_norm[n] = (G[n] - μ) / σ Feature point determination: Set a feature extraction threshold θ, and determine points whose absolute value of the normalized gradient exceeds θ as weld feature points: Anomaly[n] = { 1, if |G_norm[n]| > τ; 0, otherwise} Here, Anomaly[n]=1 indicates that the nth position is identified as a weld feature point. The threshold θ is determined experimentally based on the statistical characteristics of the feature point distribution to balance the detection rate and the false alarm rate.

[0021] Results Output and Visualization: Based on the Anomaly[n] sequence, the positions of feature points are marked on the Data_Out[n] waveform and visualized to intuitively indicate the possible locations of buried pipeline butt welds.

[0022] Example To verify the effectiveness of this method, a set of simulated magnetic field data from pipe welds was processed. The original signal sequence (unit: microtesla, μT) is as follows: [10, 12, 11, 14, 13, 17, 20, 18, 25, 24, 22, 30, 28, 35, 33, 34, 40, 38, 37, 50] Wavelet denoising: Using the 'db4' wavelet and setting the denoising threshold T=5, the denoised signal Data_Out is obtained after decomposition, thresholding, and reconstruction. [9.8, 12.0, 12.5, 14.0, 13.5, 17.0, 20.2, 18.5, 24.0, 23.2, 21.8, 30.0, 28.5, 35.0, 33.0, 34.0, 40.0, 38.0, 37.0, 50.0] Gradient feature extraction: Calculate the gradient G: [2.2, 0.5, 1.5, -0.5, 3.5, 3.2, -1.7, 5.5, -0.8, -1.2, 8.2, -1.5, 6.5, -1.5, 1.0, 6.0, -2.0, -1.0, 13.0] Calculate the mean μ and standard deviation σ: μ ≈ 1.5 uT, σ ≈ 3.8 uT Normalized gradient G_norm: [0.18, -0.26, 0.0, -0.66, 0.53, 0.63, -0.77, 1.04, -0.22, -0.37, 1.75, -0.66, 1.29, -0.53, -0.13, 1.18, -1.30, -0.47, 3.02] Feature point determination: With θ=1.1, the feature labels are obtained as follows: [0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 1]. A total of 6 feature points were identified.

[0023] Setting a more stringent θ=1.2, the feature labels are obtained as follows: [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,1, 0, 0, 0, 1, 0, 1]. A total of 4 feature points were identified, corresponding to the locations where the gradient abrupt changes were most dramatic.

[0024] By adjusting the threshold θ, the sensitivity of feature extraction can be controlled to adapt to different detection requirements and noisy environments. The final marked feature point locations indicate the areas of magnetic field anomalies caused by the weld.

Claims

1. A wavelet noise reduction and gradient feature extraction method for locating weld seams in buried pipelines, characterized in that, Includes the following steps: (1) Signal acquisition steps: Obtain the original magnetic field strength sequence along the axis of the buried pipeline; (2) Wavelet domain adaptive noise reduction step: Perform discrete wavelet transform on the original magnetic field strength sequence to decompose it to different scales; Based on the statistical characteristics of wavelet coefficients at the aforementioned scale, adaptive thresholds for each scale are determined. The adaptive threshold is applied to perform nonlinear thresholding on the wavelet coefficients to suppress non-periodic noise components; the processed wavelet coefficients are then subjected to inverse transform to reconstruct the denoised magnetic field signal sequence. (3) Gradient statistical feature extraction steps: Calculate the first-order difference gradient sequence of the denoised magnetic field signal sequence; calculate the mean and standard deviation of the gradient sequence, and perform standardization processing on the gradient sequence based on the mean and standard deviation to obtain the standardized gradient sequence; (4) Weld feature location determination steps: Set a feature determination threshold; compare the absolute value of each point in the standardized gradient sequence with the feature determination threshold, and determine the point whose absolute value exceeds the feature determination threshold as the weld feature point; determine the magnetic field feature signal location of the buried pipeline butt weld according to the location of the weld feature point.

2. The method according to claim 1, characterized in that, In the wavelet domain adaptive noise reduction step, the adaptive threshold for each scale is determined based on the statistical characteristics of the wavelet coefficients at the scale, specifically using the following formula: T_j = σ_j *sqrt(2 * log(N)); where T_j is the threshold for the j-th scale, σ_j is the noise level estimate of the wavelet coefficients at the j-th scale, and N is the signal length.

3. The method according to claim 1 or 2, characterized in that, In the wavelet domain adaptive noise reduction step, the nonlinear thresholding is either a soft thresholding function or a hard thresholding function.

4. The method according to claim 1, characterized in that, In the gradient statistical feature extraction step, the specific formula for the standardization process is: G_norm[n] = (G[n] - μ) / σ; where G[n] is the original gradient value, μ is the mean of the gradient sequence, σ is the standard deviation of the gradient sequence, and G_norm[n] is the standardized gradient value.

5. The method according to claim 1, characterized in that, The feature determination threshold is set based on the statistical distribution of the standardized gradient sequence, and a predetermined multiple of the standard deviation is selected as the threshold.

6. A positioning system for butt welds of buried pipelines, characterized in that, The system for implementing the method as described in any one of claims 1-5 comprises: A magnetic sensor array is used to move along the pipeline and acquire the original magnetic field strength sequence. The signal processing unit, which is communicatively connected to the magnetic sensor array, is configured to perform the wavelet domain adaptive noise reduction step, the gradient statistical feature extraction step, and the weld feature location determination step. The positioning and marking unit is used to mark the position on the ground above the pipeline based on the weld feature signal position output by the signal processing unit.

7. The system according to claim 6, characterized in that, The signal processing unit further includes a visualization module for displaying the original magnetic field strength sequence, the noise-reduced magnetic field signal sequence, the normalized gradient sequence, and the marked positions of the weld feature points.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-5.