Denoising method and system based on geophysical signals

Abstract

The invention relates to the technical field of signal processing, in particular to a denoising method and system based on geophysical signals, and the method comprises the following steps: carrying out the correlation analysis of features between adjacent samples based on the geophysical signal sample data, constructing the dependency relationship between each sample and the adjacent samples, extracting the signal features, and carrying out the correlation analysis of the features between the adjacent samples; and constructing a noise structure dependency relationship analysis result. According to the invention, the method achieves the precise modeling of the noise structure in the seismic signal through the building of the dependency matrix between samples, effectively recognizes the spatial distribution and dependency features of high-frequency noise, and avoids the interference of the noise remaining in the reconstructed signal. In the signal processing process, smooth transition of a signal connection area is ensured through similarity weight smoothing processing of overlapped signal blocks, and high-frequency noise is effectively isolated and suppressed by decomposing signal components of each frequency band, setting adaptive thresholds for different frequencies and dynamically adjusting in combination with energy distribution of a local time-frequency spectrum.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to a denoising method and system based on geophysical signals. Background Art

[0002] Signal processing is the technical field of analyzing, modifying and synthesizing signals, optimizing the usability and accuracy of data by extracting useful information from the original signal. In this process, signal processing involves a variety of algorithms and mathematical tools, such as Fourier transform, filtering, time-frequency analysis and wavelet transform, in order to effectively process data in different dimensions such as time, frequency and space.

[0003] Among them, the denoising method based on geophysical signals improves the reliability and resolution of signals by eliminating the noise associated with geophysical data. It is usually used in fields such as seismic exploration, groundwater monitoring and mineral resource exploration, aiming to extract valuable information from complex geophysical signals, such as stratigraphic structure, fluid distribution and geological characteristics. Through effective denoising technology, researchers can analyze the underground environment more clearly and provide support for resource development and environmental monitoring.

[0004] In seismic exploration, the existing technology has the following deficiencies in the processing of residual signals after denoising. For example, in actual operation, in the reconstruction of seismic signals after denoising, the high-frequency noise components in the residual signal are prone to form sudden peaks, which are difficult to effectively eliminate through existing smoothing methods, resulting in poor overall smoothness of the signal and even local sharp noise. The superposition of noise peaks and real seismic reflection waves blurs the clarity of the stratum reflection interface, making it difficult for exploration data to accurately present the detailed features of the stratum interface. The direct impact is the distortion of the reflection interface morphology, especially in areas with complex stratum boundaries or many faults. Noise peaks may be mistaken for geological structural features, resulting in errors in the interpretation of geological horizons, affecting subsequent geological analysis and resource evaluation. The resolution and reliability of signal analysis are limited. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a denoising method and system based on geophysical signals.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a denoising method based on geophysical signals, comprising the following steps: S1: Based on the geophysical signal sample data, the dependency relationship between each sample and adjacent samples is constructed by correlation analysis of the features between adjacent samples, and the signal features are extracted to construct the noise structure dependency analysis results; S2: Based on the noise structure dependency analysis result, the geophysical signal is divided into a plurality of overlapping signal blocks, similarity weights between the signal blocks are determined, the overlapping areas are smoothed, and a signal smoothing reconstruction result is generated; S3: Based on the signal smoothing reconstruction result, decompose the signal layer by layer, obtain the decomposition coefficient of each layer, identify the stationary part and the non-stationary part in the signal, mark the non-stationary part as noise, and generate a stationary signal separation result; S4: Based on the stationary signal separation result, thresholds are set according to frequency characteristics, the thresholds are dynamically adjusted in combination with the energy distribution of the local time-frequency spectrum, and the remaining components are reconstructed and converted back to the time domain to obtain a multi-band denoising result; S5: based on the multi-band denoising result, obtaining spectrum information of the residual signal, determining the key noise frequency band of the residual signal, applying smoothing filtering operations to the frequency bands of the residual signal in sequence, merging all processed frequency bands, and obtaining a residual signal smoothing result; S6: Based on the residual signal smoothing result, the optimized residual signal is integrated with the stationary signal component, and the integrated signal is smoothed in the time domain to generate a geophysical denoised signal.

[0007] The present invention has the following improvements: the noise structure dependency analysis result includes a dependency matrix between samples, characteristic correlation parameters of samples and distribution characteristics of noise structure; the signal smoothing reconstruction result includes overlapping signal blocks, similarity scores between signal blocks and smooth transition areas after processing; the stationary signal separation result includes a decomposition level of the signal, a stationary characteristic signal part and an isolated non-stationary noise part; the multi-band denoising result includes a denoising threshold assigned to each band, a local time-frequency spectrum adjustment coefficient and a denoised signal after band reconstruction; the residual signal smoothing result includes an identified key noise band, smoothing filtering parameters of the band and a merged smoothed signal; and the geophysical denoising signal includes an optimized residual signal, stationary signal characteristic components and an integrated time-domain smoothed signal.

[0008] The present invention is improved in that, based on the geophysical signal sample data, by performing correlation analysis on the features between adjacent samples, constructing the dependency relationship between each sample and the adjacent samples, and extracting the signal features, the specific steps of constructing the noise structure dependency relationship analysis result are as follows: S101: Based on the geophysical signal sample data, the amplitude, frequency and phase values ​​of each sample are set, the variables are initialized and the state parameters of each sample are stored, and a potential state variable set is generated by data marking and storing the variable distribution; S102: Based on the potential state variable set, analyze the features between adjacent samples, perform parameterization processing on the dependencies of each sample node and its adjacent nodes through the feature matching degree and dependency strength between samples, and establish a dependency matrix; S103: Based on the dependency matrix, extract the characteristics of the geophysical signals between nodes, optimize the distribution of the characteristic weights of the samples, integrate the dependency strength and characteristic weights of each pair of nodes, and generate the noise structure dependency analysis results.

[0009] The present invention is improved in that, based on the noise structure dependency analysis result, the geophysical signal is divided into a plurality of overlapping signal blocks, the similarity weights between the signal blocks are determined, and the overlapping areas are smoothed to generate the signal smoothing reconstruction result in the following specific steps: S201: Based on the noise structure dependency analysis result, the geophysical signal is regionally divided into blocks, the boundary position and overlap ratio of each signal block are determined, the characteristics and overlap metrics of the signal blocks are stored, and the boundary position and overlap data are generated; S202: based on the boundary position and the overlapping data, obtaining the Euclidean distance between adjacent signal blocks, and constructing a signal block similarity distance matrix according to the distance data and the distance weights between the signal blocks; S203: Based on the signal block similarity distance matrix, a weighted average value of adjacent signal blocks is calculated, the connection of the overlapping areas is smoothed, signal block connection information and smooth transition parameters are stored, and a signal smoothing reconstruction result is generated.

[0010] The present invention is improved in that, based on the signal smoothing reconstruction result, the signal is decomposed layer by layer, the decomposition coefficient of each layer is obtained, the stationary part and the non-stationary part in the signal are identified, and the non-stationary part is marked as noise. The specific steps of generating the stationary signal separation result are as follows: S301: Based on the signal smoothing reconstruction result, the geophysical signal is decomposed layer by layer, and the signal features of each scale are extracted according to the details of each layer, and the decomposition coefficient of each layer is recorded to generate a decomposition coefficient set; S302: Based on the decomposition coefficient set, distinguish the low-frequency or high-frequency part according to the set first threshold frequency, extract the decomposition coefficient of the low-frequency part as the stationary signal component, compare the low-frequency coefficient with the high-frequency coefficient one by one, identify the stationary part and the non-stationary part in the signal, and generate stationary and non-stationary part identifiers; S303: Based on the stationary and non-stationary part identifiers, mark the non-stationary part in the high-frequency component, record the characteristics of the noise component, and generate a stationary signal separation result.

[0011] The present invention has the following improvements: based on the stationary signal separation result, thresholds are set according to frequency characteristics, the thresholds are dynamically adjusted in combination with the energy distribution of the local time-frequency spectrum, the remaining components are reconstructed and converted back to the time domain, and the specific steps of obtaining the multi-band denoising result are as follows: S401: based on the stationary signal separation result, extract low-frequency and high-frequency component information therefrom, analyze the characteristics of each frequency component in turn, set respective second thresholds, adjust the denoising and reconstruction of the geophysical signal according to the second thresholds, and generate a frequency characteristic threshold set; S402: Based on the frequency characteristic threshold set and in combination with the energy distribution of the local time-frequency spectrum, dynamically adjust the threshold of each frequency band, and generate an adjusted frequency band signal by processing the geophysical signal to match the new second threshold; S403: Based on the adjusted frequency band signal, reconstruct the remaining high-frequency part, convert the geophysical signal back to the time domain, save the reconstructed data, and generate a multi-band denoising result.

[0012] The present invention is improved in that, based on the multi-band denoising result, the spectrum information of the residual signal is obtained, the key noise frequency band of the residual signal is determined, the smoothing filtering operation is applied to the frequency bands of the residual signal in sequence, and all the processed frequency bands are merged to obtain the residual signal smoothing result. The specific steps are as follows: S501: Based on the multi-band denoising result, calculate the point-by-point difference between the reconstructed signal and the original geophysical signal, extract the corresponding residual signal, separate the residual signal and classify and store it, and generate residual signal data; S502: extracting spectrum information of the residual signal based on the residual signal data, analyzing the energy concentration distribution of the frequency bands on the spectrum, identifying key noise frequency bands according to the energy size, and generating key noise frequency band information; S503: Based on the key noise frequency band information, apply smoothing filtering operations to the key noise frequency bands of the residual signal in sequence, adjust the spectrum characteristics, and merge all processed frequency bands item by item to obtain a residual signal smoothing result.

[0013] The present invention is improved in that, based on the residual signal smoothing result, the optimized residual signal is integrated with the stationary signal component, and the integrated signal is smoothed in the time domain to generate the geophysical denoised signal in the following specific steps: S601: Based on the residual signal smoothing result, the residual signal and the stationary signal components are accumulated item by item, the amplitude matching and structure adjustment are performed on each component, the frequency characteristics of the key components are retained, including the frequency range, amplitude change, phase shift and energy concentration, the combination result is recorded, and the combined signal is generated; S602: smoothing the combined signal item by item in the time domain, smoothing and correcting the mutation between each data item, adjusting and recording the processed time series data, and generating a smoothed combined signal; S603: Based on the smoothed combined signal, verify and correct the global consistency of the time series, determine the overall continuity of the signal in the time domain, preserve the signal structure, and generate a geophysical denoised signal.

[0014] A denoising system based on geophysical signals, the system comprising: The signal correlation analysis module performs correlation analysis on the features between adjacent samples based on the geophysical signal sample data. By constructing the dependency relationship between each sample and the adjacent samples, the signal features are extracted and the noise structure dependency analysis results are generated. The signal reconstruction module divides the geophysical signal into a plurality of overlapping signal blocks based on the noise structure dependency analysis result, determines the similarity weights between the signal blocks, performs smoothing processing on the overlapping areas, and generates a signal smoothing reconstruction result; The signal decomposition module decomposes the signal layer by layer based on the signal smoothing reconstruction result, obtains the decomposition coefficient of each layer, determines the stable part and the non-stationary part, selects to mark the non-stationary part as noise, and generates a stable signal separation result; The signal threshold adjustment module sets the thresholds according to the frequency characteristics based on the stationary signal separation result, dynamically adjusts the thresholds based on the energy distribution of the local time-frequency spectrum, reconstructs the remaining components and converts them back to the time domain to obtain a multi-band denoising result; Based on the multi-band denoising results, the signal integration module obtains the spectrum information of the residual signal, determines the key noise frequency band, applies smoothing filtering operations to the frequency bands of the residual signal in sequence, merges all processed frequency bands, integrates the optimized residual signal with the stationary signal component, and performs time domain smoothing to generate a geophysical denoised signal.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by establishing a dependency matrix between samples, the noise structure in the seismic signal is accurately modeled, the spatial distribution and dependency characteristics of high-frequency noise are effectively identified, and the interference of noise residues in the reconstructed signal is avoided. In the signal processing process, the similarity weight smoothing of overlapping signal blocks is used to ensure the smooth transition of the signal connection area, thereby reducing the appearance of sudden peaks in the signal, and by decomposing the signal components of each frequency band, setting adaptive thresholds for different frequencies, and dynamically adjusting the energy distribution of the local time-frequency spectrum, the high-frequency noise is effectively isolated and suppressed, ensuring the continuity of the spectrum and the accuracy of denoising. In addition, the residual signal is smoothly integrated with the stationary signal characteristics, and the residual signal in the time domain is dynamically adjusted to avoid the superposition of high-frequency noise peaks, so that the formation reflection interface presents clearer details. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a work flow chart of the present invention; Figure 2 This is a detailed flow chart of step S1 of the present invention; Figure 3 This is a detailed flow chart of step S2 of the present invention; Figure 4 This is a detailed flow chart of step S3 of the present invention; Figure 5 This is a detailed flow chart of step S4 of the present invention; Figure 6 This is a detailed flow chart of step S5 of the present invention; Figure 7 This is a detailed flow chart of step S6 of the present invention; Figure 8 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0019] See also Figure 1 The present invention provides a technical solution: a denoising method based on geophysical signals, comprising the following steps: S1: Based on the geophysical signal sample data, the dependency relationship between each sample and adjacent samples is constructed by correlation analysis of the features between adjacent samples, and the signal features are extracted to construct the noise structure dependency analysis results; S2: Based on the noise structure dependency analysis results, the geophysical signal is divided into multiple overlapping signal blocks, the similarity weights between the signal blocks are determined, the overlapping areas are smoothed, and the signal smoothing reconstruction results are generated; S3: Based on the signal smoothing reconstruction result, the signal is decomposed layer by layer, the decomposition coefficient of each layer is obtained, the stationary part and the non-stationary part in the signal are identified, and the non-stationary part is marked as noise to generate a stationary signal separation result; S4: Based on the stationary signal separation results, the thresholds are set according to the frequency characteristics, and the thresholds are dynamically adjusted in combination with the energy distribution of the local time-frequency spectrum. The remaining components are reconstructed and converted back to the time domain to obtain the multi-band denoising results; S5: Based on the multi-band denoising result, the spectrum information of the residual signal is obtained, the key noise frequency band of the residual signal is determined, smoothing filtering operations are applied to the frequency bands of the residual signal in sequence, and all processed frequency bands are merged to obtain a smoothing result of the residual signal; S6: Based on the residual signal smoothing result, the optimized residual signal is integrated with the stationary signal component, and the integrated signal is smoothed in the time domain to generate a geophysical denoised signal.

[0020] The results of noise structure dependency analysis include the dependency matrix between samples, the characteristic correlation parameters of samples and the distribution characteristics of noise structure. The results of signal smoothing reconstruction include overlapping signal blocks, similarity scores between signal blocks and smooth transition areas after processing. The results of stationary signal separation include the decomposition level of the signal, the stationary characteristic signal part and the isolated non-stationary noise part. The results of multi-band denoising include the denoising threshold assigned to each band, the local time-frequency spectrum adjustment coefficient and the denoised signal after band reconstruction. The residual signal smoothing results include the identified key noise bands, the smoothing filter parameters of the bands and the merged smoothed signals. The geophysical denoising signal includes the optimized residual signal, the characteristic components of the stationary signal and the integrated time domain smoothed signal.

[0021] See also Figure 2 Based on the geophysical signal sample data, the dependence relationship between each sample and the adjacent samples is constructed by correlation analysis of the features between adjacent samples, and the signal features are extracted. The specific steps for constructing the noise structure dependence analysis results are as follows: S101: Based on the geophysical signal sample data, the amplitude, frequency and phase values ​​of each sample are set, the variables are initialized and the state parameters of each sample are stored, and a potential state variable set is generated by data marking and storing the variable distribution; First, the amplitude, frequency and phase values ​​of each sample are obtained and set to establish the initial state parameter table, and the signal characteristics of each sample are classified in detail. The specific process of classifying the variable categories is as follows: in terms of amplitude division, the amplitude value range is divided into low amplitude (0-0.3), medium amplitude (0.3-0.7) and high amplitude (0.7-1.0); frequency division is divided into low frequency band (0-30Hz), medium frequency band (30-60Hz) and high frequency band (above 60Hz) by calculating the main component of signal frequency; the phase classification is based on the offset value of the sample signal on the time axis, and the phase value is divided into small offset (0°-90°), medium offset (90°-180°) and large offset (180°-360°). The divided data of all categories are marked and stored one by one in the potential state variable set. Each variable category contains a specific value range and classification criteria to complete the generation of the potential state variable set.

[0022] S102: Based on the potential state variable set, the features between adjacent samples are analyzed, and the dependency between each sample node and its adjacent nodes is parameterized through the feature matching degree and dependency strength between the samples to establish a dependency matrix; First, the amplitude difference, frequency difference and phase difference of the sample pair are obtained, and the similarity or dependence of the two samples is judged based on these difference values. Nodes with strong dependence are determined by difference thresholds. Specifically, if the amplitude difference between samples is less than 0.2, the frequency difference is less than 10Hz, and the phase difference is less than 45°, the sample pair is considered to have a strong dependence. At this time, the dependence strength values ​​of the two nodes are set to a higher value (such as 1) in the dependency matrix, and the node pairs that do not meet the above conditions have their dependence strength set to a lower value (such as 0.2). The dependence strengths of all sample pairs form a dependency matrix, in which each row and column represent a sample node, and the value of each matrix element represents the dependence strength between sample nodes.

[0023] S103: Based on the dependency matrix, extract the characteristics of the geophysical signals between nodes, optimize the characteristic weights of the samples, integrate the dependency strength and characteristic weights of each pair of nodes, and generate the noise structure dependency analysis results; The dependency strength data between each pair of nodes is extracted from the dependency matrix, and the feature weights of each node are adjusted and optimized. The node features with stronger dependencies are given higher weight values ​​during the optimization process, which is completed through the following steps: First, the initial weight values ​​are assigned according to the dependency strength of each pair of nodes in the matrix, and the weights of the strong dependency node pairs are increased to between 0.8 and 1, while the weights of the weak dependency nodes are assigned between 0.3 and 0.5. Then, the weights of the entire matrix are normalized to ensure that the total weight distribution meets the unified standard. Finally, the optimized weight values ​​of each pair of nodes are summarized into the dependency matrix to generate the noise structure dependency analysis results.

[0024] See also Figure 3 ,Based on the results of noise structure dependency analysis, the geophysical signal is divided into multiple overlapping signal blocks, the similarity weights between the signal blocks are determined, the overlapping areas are smoothed, and the specific steps of generating signal smoothing reconstruction results are as follows: S201: Based on the noise structure dependency analysis result, the geophysical signal is regionally divided into blocks, the boundary position and overlap ratio of each signal block are determined, the characteristics and overlap metrics of the signal blocks are stored, and the boundary position and overlap data are generated; First, the boundary position of each signal block is determined, and the overlapping area is confirmed by comparing the boundaries of adjacent signal blocks one by one. The specific operation is that the boundary of the signal block is marked by analyzing the changing position of the signal strength in the spatial coordinate system, and then the boundary data of adjacent signal blocks are paired, and the position of the boundaries of each pair of adjacent signal blocks is compared to find the start and end coordinate points of the overlapping area, and then the specific position and length of the overlapping area are determined. After that, the overlap metric is determined by the ratio of the length of the overlapping area to the total length of the signal block, and the overlap ratio of the two signal blocks is expressed in numerical form. Finally, the boundary position, overlap metric and overlapping area of ​​each signal block are stored in the signal block data.

[0025] S202: based on the boundary position and the overlapping data, obtaining the Euclidean distance between adjacent signal blocks, and constructing a signal block similarity distance matrix according to the distance data and the distance weights between the signal blocks; By obtaining the distance information of each pair of adjacent signal blocks under a specific spatial mapping, the spatial distance between signal blocks is calculated. This distance is measured in the signal feature space. The spatial mapping is based on a three-dimensional coordinate system. The eigenvalues ​​of the signal blocks are mapped to the corresponding points in the three-dimensional space. Matlab is used to automatically calculate the Euclidean distance of the three-dimensional coordinates of adjacent signal blocks in the mapping space. In the process of constructing the signal block similarity distance matrix, the distance data of each pair of signal blocks is first filled into the corresponding position as a matrix element to form the basic structure of the matrix. Each row and column represents a signal block, and each element in the matrix represents the similarity weight between two signal blocks. By converting the distance value into a weight and assigning it to the matrix, the matrix expression of the similarity between signal blocks is realized.

[0026] S203: Calculate the weighted average of adjacent signal blocks based on the signal block similarity distance matrix, perform smoothing on the connection of the overlapping area, store the signal block connection information and smooth transition parameters, and generate a signal smoothing reconstruction result; Smooth the joints of overlapping areas using the formula:

[0027] Calculate the weighted average of adjacent signal blocks ; in, is a weighted average, which represents the smoothed value of adjacent signal blocks in the overlapping area. is the signal block index, is the number of signal blocks in the overlapping area, determined by the block setting, is the signal stability factor, which is used to describe the The signal stability of a signal block is obtained based on the volatility analysis of the signal block. The standard deviation and mean of the signal block are usually calculated using statistical analysis in MATLAB or Python. When the standard deviation is relatively small, a higher value is set. The value is (0.9 to 1.2), and vice versa is lower (0.7 to 0.9), reflecting the stability of the signal block. is the distance weight, which describes the similarity between adjacent signal blocks. The distance weight is obtained by mapping the signal blocks to a three-dimensional coordinate system and measuring the Euclidean distance, and then converting the distance value into a weight. The K nearest neighbor algorithm or the distance transformation function is used to calculate and convert the weight using the scikit-learn library of MATLAB or Python. The closer the signal blocks are, the higher the weight value (such as 0.8 to 1.0) is assigned, and the farther the signal blocks are, the lower the weight value (such as 0.3 to 0.5). It is The characteristic strength value of a signal block is the average signal strength of the signal block, which is determined according to the energy or amplitude level of the signal block.

[0028] For example, there are three signal blocks, whose characteristic values ​​are , the distance weight is , the signal stability coefficient is . Calculate according to the formula: In order to obtain the smoothed weighted average After that, it is assigned to the connection point of the overlapping area and applied to the transition section connecting the two signals. Corresponding to the midpoint position of the overlapping area of ​​the signal block, the midpoint is used as a reference, extending to both sides to the overlapping boundary, and gradually adjusting the value of the connection segment signal to the smooth value to make the connection transition smoother. In addition, the smooth transition parameters include the boundary point position of the signal block, the range of the overlapping area, the interpolation method of the smooth transition and the weighted average value. All parameters are stored in the signal block connection information table, recording the connection method of each pair of signal blocks and the metric value of the smoothing process, generating the smooth transition parameters, and completing the signal smooth reconstruction.

[0029] Assuming the expected benchmark value is 10.5, the smoothed value is compared with the benchmark value, and the result 10.43 is close to the expected benchmark value 10.5. This shows that the generated smooth transition parameters and weighted smoothing of the overlapping areas have achieved the expected effect, successfully removing noise and smoothing signal connections, ensuring the continuity and integrity of the signal at the regional connection, and providing a stable basic signal for subsequent geophysical signal processing.

[0030] See also Figure 4 Based on the signal smoothing reconstruction result, the signal is decomposed layer by layer, the decomposition coefficient of each layer is obtained, the stationary part and the non-stationary part in the signal are identified, and the non-stationary part is marked as noise. The specific steps to generate the stationary signal separation result are as follows: S301: Based on the signal smoothing reconstruction result, the geophysical signal is decomposed layer by layer, and the signal features of each scale are extracted according to the details of each layer, and the decomposition coefficient of each layer is recorded to generate a decomposition coefficient set; First, use the wavelet decomposition tool in MATLAB to select a mother wavelet suitable for the signal type (such as db4 or sym5), and determine the number of decomposition layers according to the signal sampling frequency to ensure that the decomposed components can fully characterize the signal characteristics of different frequency ranges. At the beginning of decomposition, import the original signal data into the wavelet decomposition toolbox, set the number of wavelet decomposition layers, and obtain the decomposed detail coefficients and approximate coefficients based on the low-frequency approximate components and high-frequency detail components of each layer. After the decomposition is completed, the system will generate multi-level wavelet coefficients, and the decomposition coefficients of each layer will be stored as a matrix file through the data export function, and a decomposition coefficient set will be established to provide a structured data basis for subsequent multi-level signal analysis.

[0031] S302: Based on the decomposition coefficient set, distinguish the low-frequency or high-frequency part according to the set first threshold frequency, extract the decomposition coefficient of the low-frequency part as the stationary signal component, compare the low-frequency coefficient with the high-frequency coefficient one by one, identify the stationary part and the non-stationary part in the signal, and generate stationary and non-stationary part identifiers; The low-frequency components are extracted by bandpass filters. The bandpass filter design tool in MATLAB is used to set the low-frequency bandpass filter range (such as 1Hz to 10Hz) to distinguish the stationary components according to the frequency components determined by the number of decomposition layers. The specific operation is to select the coefficients of each layer in the decomposition coefficient set, apply the bandpass filter layer by layer, extract the low-frequency coefficients separately, apply the high-frequency filter to extract the high-frequency part, and further extract the stationary characteristics of the signal by comparing the low-frequency and high-frequency components of each layer. After the filtering operation is completed, the low-frequency components and high-frequency components of each layer are stored as two data sets respectively, and the mean, variance and other statistical data of the low-frequency and high-frequency components of each layer are calculated through the MATLAB data analysis function, and the stationary changes of high and low-frequency components are compared. The frequency range of the stationary and non-stationary parts is marked, and the stationary and non-stationary part identification is generated.

[0032] S303: Based on the identification of the stationary and non-stationary parts, the non-stationary part in the high-frequency component is marked, the characteristics of the noise component are recorded, and a stationary signal separation result is generated; Select MATLAB's spectrum analysis, import high-frequency component data marked as non-stationary, set the spectrum analysis range (such as 10Hz to 100Hz), use the Fourier transform method to obtain the spectrum density of the noise component, and record the frequency distribution, amplitude range and phase information of the noise. After the spectrum analysis is completed, the noise frequency component is further segmented, and the frequency range and noise amplitude intensity are stored as independent noise feature data tables to form a noise component feature set. In subsequent operations, by comparing with the stable part of the original signal, the frequency influence range of the noise is clarified, the noise separation processing of the stable signal is completed, and the stable signal separation result containing the noise spectrum characteristics is generated.

[0033] See also Figure 5 Based on the stationary signal separation results, the thresholds are set according to the frequency characteristics, and the thresholds are dynamically adjusted in combination with the energy distribution of the local time-frequency spectrum. The remaining components are reconstructed and converted back to the time domain. The specific steps to obtain the multi-band denoising results are as follows: S401: based on the stationary signal separation result, extract low-frequency and high-frequency component information therefrom, analyze the characteristics of each frequency component in turn, set respective second thresholds, adjust the denoising and reconstruction of the geophysical signal according to the second thresholds, and generate a frequency characteristic threshold set; First, the separated signal data is imported, and the frequency feature extraction is performed on the low-frequency component and the high-frequency component respectively. The time domain signal is converted into the frequency domain signal through the FFT (Fast Fourier Transform) tool of MATLAB, the spectrum density and power spectrum information of each frequency band are obtained, and the amplitude, phase and energy density parameters of each frequency component are calculated respectively. The threshold range is selected based on the distribution of the frequency components on the spectrum graph. The signal power distribution curve of each frequency band is obtained through the signal analysis tool of MATLAB, and the peak frequency and average energy of the signal in different frequency bands are analyzed to determine the frequency range of the noise component. When setting the second threshold, the noise area with lower frequency amplitude is selected and set as the noise threshold of the frequency component. Combined with the frequency component distribution of the signal, the threshold setting range is adjusted to achieve the control effect of denoising and reconstruction of signals in different frequency bands. Through the MATLAB data saving function, the threshold value information of different frequency bands is stored in sequence to form a frequency feature threshold set.

[0034] S402: Based on the frequency characteristic threshold set and in combination with the energy distribution of the local time-frequency spectrum, the threshold of each frequency band is dynamically adjusted, and the adjusted frequency band signal is generated by matching the new second threshold with the geophysical signal through processing; First, the short-time Fourier transform (STFT) is used to obtain the time-frequency spectrum of the signal, and the frequency distribution and energy density of the signal in each time period are converted into a two-dimensional image. In this process, the sampled signal data is divided into several time windows, and the short-time spectrum of the signal is generated by STFT. For each time window, the MATLAB tool calculates its spectral density and extracts the instantaneous frequency component and energy density of the signal. Based on the data in the frequency feature threshold set, the threshold is set segment by segment according to the frequency range. When setting the threshold, the part of the low-frequency energy of each frequency band that is higher than the second threshold is marked as the stable signal part, and the component below the threshold is marked as the noise component. In order to improve the processing accuracy, multi-threshold segmentation is used for the frequency bands with more complex signal spectral characteristics, and the noise component is separated on the time-frequency spectrum of the signal through adaptive adjustment. Finally, the signal spectrum data after denoising in each time period is reintegrated.

[0035] S403: reconstructing the remaining high-frequency part based on the adjusted frequency band signal, converting the geophysical signal back to the time domain, saving the reconstructed data, and generating a multi-band denoising result; After the denoising process is completed, the inverse fast Fourier transform (iFFT) is used to convert the frequency data into a time domain signal to reconstruct the time domain characteristics of the original signal. The specific steps include superimposing each adjusted frequency band signal in sequence to ensure that the amplitude and phase of each frequency band are consistent during reconstruction. During the reconstruction process, by setting time synchronization, the time domain signals of different frequency bands are restored layer by layer, the overlapping areas between each frequency band are checked, and the phase consistency of the transition points is adjusted to avoid phase jumps caused by discontinuities between frequency bands. The reconstructed signal data is exported to the time domain waveform analysis function of MATLAB to calculate the amplitude change and phase characteristics of the reconstructed signal to ensure that the reconstructed signal is synchronized in time. Finally, the complete time domain reconstruction data is saved to generate a denoised multi-band time domain signal.

[0036] See also Figure 6 , based on the multi-band denoising results, the spectrum information of the residual signal is obtained, the key noise frequency band of the residual signal is determined, the smoothing filter operation is applied to the frequency bands of the residual signal in turn, and all the processed frequency bands are merged to obtain the residual signal smoothing result. The specific steps are as follows: S501: Based on the multi-band denoising result, calculate the point-by-point difference between the reconstructed signal and the original geophysical signal, extract the corresponding residual signal, separate the residual signal and classify and store it, and generate residual signal data; Calculate the point-by-point difference between the reconstructed signal and the original signal according to the formula: , and obtain the point-by-point residual signal , Indicated in The residual between the reconstructed signal and the original signal at sampling points is Represents the original geophysical signal at The value of the sampling point, Indicates that the denoised reconstructed signal is For example, for a 1000-point signal ( ), calculated from arrive At the 10th sampling point, if the original signal , and the reconstructed signal ,but: Point-wise residual signal The data reflects the denoising difference between the reconstructed signal and the original signal at each sampling point. According to the preset frequency segmentation range, the residual signal data is separated into independent data sets segment by segment, and the residual signal characteristics of the low-frequency band and the high-frequency band are saved respectively. Subsequently, the residual signal is classified and stored according to the characteristics and frequency distribution of different frequency bands, and the frequency band data sets are saved with frequency identification numbers to ensure that the subsequent noise analysis can carry out targeted processing on specific frequency bands. Finally, the separated residual signal data is stored in the specified file directory according to the classification standard to generate residual signal data.

[0037] S502: extracting spectrum information of the residual signal based on the residual signal data, analyzing the energy concentration distribution of the frequency bands on the spectrum, identifying key noise frequency bands according to the energy size, and generating key noise frequency band information; First, the residual signal data in the MATLAB environment is imported, and the fast Fourier transform (FFT) is applied to obtain the spectrum information of the signal, and the residual signal is converted from the time domain to the frequency domain. By analyzing the frequency distribution and energy density of each frequency band, the frequency band with high spectrum energy is set as the key noise band. The specific operation includes calculating the spectrum density function value segment by segment to observe the concentrated distribution of energy on each frequency band. In this process, an appropriate frequency band energy threshold is selected, and the spectrum energy distribution is statistically analyzed through the MATLAB analysis toolbox. The energy density of different frequency bands is compared, and the frequency bands exceeding the set energy threshold are marked as key noise bands, and the information of key noise bands is generated and recorded. Finally, the frequency range and energy value of each noise band are output to the data set to form the key noise band information.

[0038] S503: Based on the key noise frequency band information, apply smoothing filtering operations to the key noise frequency bands of the residual signal in sequence, adjust the spectrum characteristics, and merge all processed frequency bands one by one to obtain a residual signal smoothing result; First, load the data set of the key noise frequency band in MATLAB, and select a smoothing filter (such as Butterworth low-pass filter) to process the high-energy noise frequency band. During the smoothing process, set the cutoff frequency of the filter, perform smoothing segment by segment according to the frequency distribution of the residual signal, and dynamically adjust the filter parameters according to the noise characteristics of different frequency bands to ensure the smooth transition of the noise components of the frequency band. In the specific operation, generate a smoothing filter through the MATLAB filter, and call the smoothing operation for each noise frequency band in turn, adjust the spectrum peak segment by segment, reduce the noise amplitude, and make the spectrum of different frequency bands smoothly connected. After the smoothing process is completed in all noise frequency bands, merge the processed spectrum of each frequency band and save it as the smoothed residual signal data file to obtain the residual signal smoothing result.

[0039] See also Figure 7Based on the residual signal smoothing result, the optimized residual signal is integrated with the stationary signal component, and the integrated signal is smoothed in the time domain to generate the geophysical denoising signal. The specific steps are as follows: S601: Based on the residual signal smoothing result, the residual signal and the stationary signal components are accumulated item by item, the amplitude matching and structure adjustment are performed on each component, the frequency characteristics of the key components are retained, including the frequency range, amplitude change, phase shift and energy concentration, the combination result is recorded, and the combined signal is generated; In the MATLAB environment, the stationary signal and residual signal data are imported separately, and the combined signal is generated by adding item by item using the point-by-point accumulation method. During the accumulation process, the signal amplitude is matched, and the amplitude difference between the stationary signal and the residual signal at each frequency component is calculated, and the amplitude adjustment operation is applied to make the amplitude value of each frequency component consistent, so as to reduce the amplitude difference between different components and retain the frequency characteristics of the key components. In addition, the phase offset is calculated item by item, and the phase correction operation is applied to eliminate the phase jump phenomenon that occurs during the signal superposition process. During the adjustment process, the frequency range, phase, amplitude change and energy concentration of the combined signal are confirmed and recorded one by one using the MATLAB spectrum analysis tool to ensure that the combined signal achieves the fusion effect of the stationary and residual signals in terms of spectrum characteristics. The combined signal is stored as a data file with detailed frequency component annotations.

[0040] S602: smoothing the combined signal item by item in the time domain, smoothing and correcting the mutation between each data item, adjusting and recording the processed time series data, and generating a smoothed combined signal; First, import the combined signal data, apply the moving average filter function of MATLAB, and smooth each point in the combined signal point by point in the form of a time window. Select an appropriate smoothing window size, usually within the range of 3 to 7 sampling points, and remove outliers with large local fluctuations by calculating the mean of the data in the smoothing window. For mutation points with obvious amplitude fluctuations, the smoothing effect is further optimized by setting a smoothing threshold. For data points exceeding the smoothing threshold, an interpolation method is used to compensate for mutations to smooth the amplitude difference between adjacent data points and reduce the impact of signal mutations. To ensure the smoothing effect, the time series of the entire combined signal is monitored section by section, and the numerical changes after smoothing in each time period are recorded. Finally, the combined signal after time series smoothing is stored as a continuous time series file to generate smoothed combined signal data.

[0041] S603: Based on the smoothed combined signal, verify and correct the global consistency of the time series, determine the overall continuity of the signal in the time domain, preserve the signal structure, and generate a geophysical denoised signal; Load the smoothed combined signal data in MATLAB, check the continuity of the signal at each time point, and verify the consistency of the signal in amplitude, phase and frequency characteristics through adjacent point difference analysis. Set the global threshold of amplitude difference and phase difference, and apply the interpolation method to correct the time points that do not meet the threshold. During the verification operation, calculate the difference between adjacent data points point by point, identify and record the period when the mutation is greater than the global threshold. For the mutation segment whose amplitude exceeds the threshold range, apply the interpolation method to smooth the mutation point, and use the weighted interpolation method to ensure the smooth transition of the previous and next data points. Finally, save the corrected data as a continuous denoising signal file to generate a geophysical denoising signal that meets the time domain consistency requirements.

[0042] See also Figure 8 , a denoising system based on geophysical signals, the system comprising: The signal correlation analysis module performs correlation analysis on the features between adjacent samples based on the geophysical signal sample data. By constructing the dependency relationship between each sample and the adjacent samples, the signal features are extracted and the noise structure dependency analysis results are generated. The signal reconstruction module divides the geophysical signal into multiple overlapping signal blocks based on the noise structure dependency analysis results, determines the similarity weights between the signal blocks, performs smoothing processing on the overlapping areas, and generates signal smoothing reconstruction results; The signal decomposition module decomposes the signal layer by layer based on the signal smoothing reconstruction result, obtains the decomposition coefficient of each layer, determines the stable part and the non-stationary part, selects to mark the non-stationary part as noise, and generates a stable signal separation result; The signal threshold adjustment module sets the thresholds based on the stationary signal separation results according to the frequency characteristics, dynamically adjusts the thresholds based on the energy distribution of the local time-frequency spectrum, reconstructs the remaining components and converts them back to the time domain to obtain multi-band denoising results; Based on the multi-band denoising results, the signal integration module obtains the spectrum information of the residual signal, determines the key noise frequency band, applies smoothing filtering operations to the frequency bands of the residual signal in turn, merges all processed frequency bands, integrates the optimized residual signal with the stationary signal component, and performs time domain smoothing to generate a geophysical denoised signal.

[0043] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A denoising method based on geophysical signals, characterized in that: The following steps are involved: Based on the geophysical signal sample data, the dependency relationship between each sample and adjacent samples is constructed by correlation analysis of the features between adjacent samples, and the signal features are extracted to construct the noise structure dependency analysis results; Based on the noise structure dependency analysis result, the geophysical signal is divided into a plurality of overlapping signal blocks, similarity weights between the signal blocks are determined, the overlapping areas are smoothed, and a signal smoothing reconstruction result is generated; Based on the signal smoothing reconstruction result, the signal is decomposed layer by layer, the decomposition coefficient of each layer is obtained, the stable part and the non-stationary part in the signal are identified, and the non-stationary part is marked as noise to generate a stable signal separation result; Based on the stationary signal separation result, thresholds are set according to frequency characteristics, the thresholds are dynamically adjusted in combination with the energy distribution of the local time-frequency spectrum, and the remaining components are reconstructed and converted back to the time domain to obtain a multi-band denoising result; Based on the multi-band denoising result, spectrum information of the residual signal is obtained, key noise frequency bands of the residual signal are determined, smoothing filtering operations are applied to the frequency bands of the residual signal in sequence, and all processed frequency bands are combined to obtain a residual signal smoothing result; Based on the residual signal smoothing result, the optimized residual signal is integrated with the stationary signal component, and the integrated signal is smoothed in the time domain to generate a geophysical denoised signal.

2. The denoising method based on geophysical signals according to claim 1, characterized in that: The noise structure dependency analysis result includes the dependency matrix between samples, the characteristic correlation parameters of the samples and the distribution characteristics of the noise structure. The signal smoothing reconstruction result includes overlapping signal blocks, similarity scores between signal blocks and the processed smooth transition area. The stationary signal separation result includes the decomposition level of the signal, the stationary characteristic signal part and the isolated non-stationary noise part. The multi-band denoising result includes the denoising threshold assigned to each band, the local time-frequency spectrum adjustment coefficient and the denoised signal after band reconstruction. The residual signal smoothing result includes the identified key noise bands, the smoothing filter parameters of the bands and the merged smoothed signal. The geophysical denoising signal includes the optimized residual signal, the stationary signal characteristic components and the integrated time domain smoothed signal.

3. The denoising method based on geophysical signals according to claim 1, characterized in that: Based on the geophysical signal sample data, the dependency relationship between each sample and adjacent samples is constructed by correlation analysis of the features between adjacent samples, and the signal features are extracted. The specific steps to construct the noise structure dependency analysis results are as follows: Based on the geophysical signal sample data, the amplitude, frequency and phase values ​​of each sample are set, the variables are initialized and the state parameters of each sample are stored, and the potential state variable set is generated by data marking and storing the variable distribution; Based on the potential state variable set, the features between adjacent samples are analyzed, and the dependency of each sample node and its adjacent nodes is parameterized through the feature matching degree and dependency strength between the samples to establish a dependency matrix; Based on the dependency matrix, the characteristics of geophysical signals between nodes are extracted, the characteristic weights of samples are optimally allocated, the dependency strength and characteristic weights of each pair of nodes are integrated, and the noise structure dependency analysis results are generated.

4. The denoising method based on geophysical signals according to claim 1, characterized in that: Based on the noise structure dependency analysis results, the geophysical signal is divided into multiple overlapping signal blocks, the similarity weights between the signal blocks are determined, and the overlapping areas are smoothed to generate the signal smoothing reconstruction results. The specific steps are as follows: Based on the noise structure dependency analysis results, geophysical signals are regionally divided into blocks, the boundary position and overlap ratio of each signal block are determined, the characteristics and overlap metrics of the signal blocks are stored, and the boundary position and overlap data are generated; Based on the boundary positions and overlapping data, obtaining the Euclidean distances between adjacent signal blocks, and constructing a signal block similarity distance matrix according to the distance data and the distance weights between the signal blocks; Based on the signal block similarity distance matrix, a weighted average value of adjacent signal blocks is calculated, smoothing is performed on the connection of the overlapping areas, signal block connection information and smooth transition parameters are stored, and a signal smoothing reconstruction result is generated.

5. The denoising method based on geophysical signals according to claim 4, characterized in that: Smooth the joints of overlapping areas using the formula: Calculate the weighted average of adjacent signal blocks ; in, is the signal block index, is the number of signal blocks in the overlapping area, is the signal stability factor, is the distance weight, It is The characteristic intensity value of a signal block.

6. The denoising method based on geophysical signals according to claim 1, characterized in that: Based on the signal smoothing reconstruction result, the signal is decomposed layer by layer, the decomposition coefficient of each layer is obtained, the stationary part and the non-stationary part in the signal are identified, and the non-stationary part is marked as noise. The specific steps of generating the stationary signal separation result are as follows: Based on the signal smoothing reconstruction result, the geophysical signal is decomposed layer by layer, and the signal characteristics of each scale are extracted according to the details of each layer, and the decomposition coefficient of each layer is recorded to generate a decomposition coefficient set; Based on the decomposition coefficient set, distinguish low-frequency or high-frequency parts according to a set first threshold frequency, extract the decomposition coefficients of the low-frequency parts as stationary signal components, compare the low-frequency coefficients with the high-frequency coefficients one by one, identify the stationary part and the non-stationary part in the signal, and generate stationary and non-stationary part identifiers; Based on the stationary and non-stationary part identifications, the non-stationary part in the high-frequency component is marked, the characteristics of the noise component are recorded, and a stationary signal separation result is generated.

7. The denoising method based on geophysical signals according to claim 1, characterized in that: Based on the stationary signal separation result, the thresholds are set according to the frequency characteristics, the thresholds are dynamically adjusted in combination with the energy distribution of the local time-frequency spectrum, the remaining components are reconstructed and converted back to the time domain, and the specific steps for obtaining the multi-band denoising result are as follows: Based on the stationary signal separation result, extract low-frequency and high-frequency component information therefrom, analyze the characteristics of each frequency component in turn, set respective second thresholds, adjust the denoising and reconstruction of the geophysical signal according to the second thresholds, and generate a frequency characteristic threshold set; Based on the frequency characteristic threshold set and in combination with the energy distribution of the local time-frequency spectrum, the threshold of each frequency band is dynamically adjusted, and the adjusted frequency band signal is generated by matching the new second threshold with the geophysical signal through processing; Based on the adjusted frequency band signal, the remaining high frequency part is reconstructed, the geophysical signal is converted back to the time domain, the reconstructed data is saved, and a multi-band denoising result is generated.

8. The denoising method based on geophysical signals according to claim 1, characterized in that: Based on the multi-band denoising result, the frequency spectrum information of the residual signal is obtained, the key noise frequency band of the residual signal is determined, the smoothing filtering operation is applied to the frequency bands of the residual signal in sequence, and all the processed frequency bands are merged to obtain the residual signal smoothing result. The specific steps are as follows: Based on the multi-band denoising result, calculating the point-by-point difference between the reconstructed signal and the original geophysical signal, extracting the corresponding residual signal, separating the residual signal and classifying and storing it, and generating residual signal data; Based on the residual signal data, extract the spectrum information of the residual signal, analyze the energy concentration distribution of the frequency band on the spectrum, identify the key noise frequency band according to the energy size, and generate the key noise frequency band information; Based on the key noise frequency band information, smoothing filtering operations are applied to the key noise frequency bands of the residual signal in sequence, the spectrum characteristics are adjusted, and all processed frequency bands are merged item by item to obtain a residual signal smoothing result.

9. The denoising method based on geophysical signals according to claim 1, characterized in that: Based on the residual signal smoothing result, the optimized residual signal is integrated with the stationary signal component, and the integrated signal is smoothed in the time domain to generate the geophysical denoised signal. The specific steps are as follows: Based on the residual signal smoothing result, the residual signal and the stationary signal components are accumulated item by item, the amplitude matching and structure adjustment are performed on each component, the frequency characteristics of the key components are retained, including the frequency range, amplitude change, phase shift and energy concentration, the combination result is recorded, and the combined signal is generated; Smoothing the combined signal item by item in the time domain, smoothing and correcting the mutation between each data item, adjusting and recording the processed time series data, and generating a smoothed combined signal; Based on the smoothed combined signal, the global consistency of the time series is verified and corrected, the overall continuity of the signal in the time domain is determined, the signal structure is preserved, and a geophysical denoised signal is generated.

10. A denoising system based on geophysical signals, characterized in that: According to the denoising method based on geophysical signals according to any one of claims 1 to 9, the system comprises: The signal correlation analysis module performs correlation analysis on the features between adjacent samples based on the geophysical signal sample data. By constructing the dependency relationship between each sample and the adjacent samples, the signal features are extracted and the noise structure dependency analysis results are generated. The signal reconstruction module divides the geophysical signal into a plurality of overlapping signal blocks based on the noise structure dependency analysis result, determines the similarity weights between the signal blocks, performs smoothing processing on the overlapping areas, and generates a signal smoothing reconstruction result; The signal decomposition module decomposes the signal layer by layer based on the signal smoothing reconstruction result, obtains the decomposition coefficient of each layer, determines the stable part and the non-stationary part, selects to mark the non-stationary part as noise, and generates a stable signal separation result; The signal threshold adjustment module sets the thresholds according to the frequency characteristics based on the stationary signal separation result, dynamically adjusts the thresholds based on the energy distribution of the local time-frequency spectrum, reconstructs the remaining components and converts them back to the time domain to obtain a multi-band denoising result; Based on the multi-band denoising results, the signal integration module obtains the spectrum information of the residual signal, determines the key noise frequency band, applies smoothing filtering operations to the frequency bands of the residual signal in sequence, merges all processed frequency bands, integrates the optimized residual signal with the stationary signal component, and performs time domain smoothing to generate a geophysical denoised signal.

Images