CN113031059B - Visual cognition-based seismic data event detection method based on environment suppression and contour combination model - Google Patents
Visual cognition-based seismic data event detection method based on environment suppression and contour combination model Download PDFInfo
- Publication number
- CN113031059B CN113031059B CN202110249015.1A CN202110249015A CN113031059B CN 113031059 B CN113031059 B CN 113031059B CN 202110249015 A CN202110249015 A CN 202110249015A CN 113031059 B CN113031059 B CN 113031059B
- Authority
- CN
- China
- Prior art keywords
- seismic data
- suppression
- contour
- visual
- environment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
- G01V1/288—Event detection in seismic signals, e.g. microseismics
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
- G01V1/282—Application of seismic models, synthetic seismograms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Abstract
The invention discloses a visual cognition-based method for detecting a same phase axis of seismic data by using an environment suppression and contour combination model, which is based on pre-stack seismic data, wherein the pre-stack seismic data contains abundant information such as amplitude, frequency and the like and can reflect the micro structure of a stratum; firstly, preprocessing the seismic data, solving wavelet fusion of the instantaneous frequency of the envelope peak value and the instantaneous amplitude of the inclined superposition peak value, and providing the optimal quality data to the maximum extent. According to the method, the non-correlated noise of the environment is anisotropically restrained while the contour of the same phase axis of the seismic data is enhanced, so that the remarkable extraction of the target contour of the same phase axis is realized, and preparation is made for subsequent seismic interpretation, reservoir prediction and the like. The invention adopts an anisotropic environment inhibition method, and can detect the edge variation more accurately than an isotropic method.
Description
Technical Field
The invention belongs to the technical field of seismic exploration, relates to non-classical receptive field suppression of visual cognition, and particularly relates to a seismic data on-phase axis detection method based on an environmental suppression and contour combination model of visual cognition.
Technical Field
The on-phase axis pickup is the basis of seismic interpretation and can be used for researching stratum of an stratum sequence, predicting a reservoir and extracting storage characteristics. The in-phase axis is defined as the line of phase peaks (peaks or troughs), and is marked according to the vibration of similar shape and the rule of regular occurrence when the seismic record is interpreted. The same phase axis can represent seismic waves of different types or strata, and the picked result can be used for tomography and interpretation, can represent stratum interfaces with different lithology, namely a sedimentary interface, and can also represent isochronous stratum interfaces with different ground history periods. The on-axis pick-up plays an important role in the determination of the subsequent reflection interface. Most of the information in the acquired signals in the seismic exploration is contained in the phase axis, for example, the self-excitation time of the phase axis can reflect the depth information of an interface to a certain extent, and the shape of the phase axis reflects the propagation speed information of the seismic waves in the medium. Since the changes of the seismic wavelets of the in-phase axis are derived from conditions such as the seismic wave propagation process, the detection and the pickup of the in-phase axis are very important for the processing and interpretation of the seismic data. The signal to noise ratio of the seismic section can be well improved by using the same-phase-axis edge pickup, so that the originally acquired data can be well used as a pad for seismic exploration interpretation, and the method is very effective for the next oil gas exploration and stratum research. Therefore, in-phase axis pickup of seismic records plays a very important role, and the accuracy of the pickup results directly affects the results of subsequent seismic data processing.
Currently, there are a number of in-phase axis pick-up methods, which are largely divided into four categories: firstly, picking up according to the instantaneous characteristics of the seismic records, such as a maximum amplitude method, an energy ratio method and the like; picking up according to global features of the seismic records, such as a correlation method, a constraint direct wave method and the like; thirdly, picking up the same phase axis by using artificial intelligence, such as an artificial neural network method, a fractal method and the like; and fourthly, an edge detection method based on a digital image processing principle. The most widely used edge detection methods are now algorithms such as Sobel operator, canny operator, etc. which are also quite sophisticated in the study of seismic exploration. The Canny algorithm (Canny, 1986) is applied in many scientific communities due to its high efficiency and intelligent analysis performance in extracting image features. Although the Canny algorithm has advantages in image detection, the calculation data is much larger than that of the traditional derivative-based detection algorithm when the detection is performed, so that the practicability in actual seismic data processing is not strong. The edge detection method converts each sampling point amplitude value of a two-dimensional seismic signal into a different gray value and regards a seismic wave gather as a gray image, so that an edge detection method is used in image processing to detect in-phase axes in the image. The information of the same phase axis picked up by the method is fuzzy, the resolution is greatly interfered by space, frequency and noise, the detection result is the envelope of the gray mutation area, the resolution is very low, and the information of the same phase axis cannot be directly used as the detection result of the same phase axis. The neural network method simulates a neural network using a known in-phase axis as a standard sample, and uses error iteration to modify the connection weights between neurons. The process of the neural network method requires enough sample support, however, the selection of the sample is very difficult, the new sample adopted can also affect the neural network, and a large number of iterative operations cause serious time consumption. The basic principle of the cross-correlation method is that waveform similarity characteristics of in-phase axes among seismic traces are utilized to extract the in-phase axes which are susceptible to noise, and the spatial resolution is reduced along with the increase of calculated amount. The principle of the chain matching method is that each waveform is represented by a peak and valley link with multiple features, and then the pick-up of the in-phase axis is converted into the best matching problem between chains, i.e. the smallest link path is found. This approach is also affected by noise, spatial and frequency interference, so it is also difficult to solve the complex on-phase processing problem. Tu et al propose an automatic pick-up event system, the method comprising three steps: two-dimensional matched filters, kalman filters, and flexible templates.
In the visual cognition process, in the seismic data processing, the position of effective oil gas in stratum layers is obtained through the in-phase axis pickup of the seismic profile according to the influence of the in-phase axis pickup of the seismic wave reflection profile on structural interpretation, and then the processing and interpretation of the seismic data are realized through the effective stratum layer information, so that the efficiency of seismic exploration is improved. It is significant to discuss how to implement in-phase axis extraction based on visual features. Geophysical prospecting techniques introduce computer vision techniques in a launch phase. Scientists explore the application of computer vision technology in the field of seismic exploration, find that a new method for improving the seismic data processing capacity by utilizing vision performance and quickly realizing oil and gas exploration, and research and application of the technology can become a wind vane for the future industrial development of seismic exploration.
Aiming at the extraction of the contour features of the same phase axis of the seismic section, the invention provides a novel method for realizing the extraction of the contour of the same phase axis of the seismic data based on a contour and boundary detection model of a primary visual cortex sensing mechanism. The method uses Gabor function to describe the simple receptive field profile of the mammalian visual cortex, introduces non-classical receptive field excitation and suppression mechanisms to reduce background noise and highlight the boundaries of the region. Finally, the contour is thinned by non-maximum suppression, and the binarized contour is extracted by a hysteresis threshold method. Compared with other traditional image edge detection, the calculation model with the visual environment inhibition and space enhancement functions provided by the invention can enable the detection of the phase axis of the seismic section to be more advantageous, and is more in line with a biological visual mechanism.
Disclosure of Invention
In order to solve the problems in the prior art, the invention aims to provide a visual cognition based method for detecting the same phase axis of seismic data based on an environmental suppression and contour combination model, which is used for the same phase axis pickup technology of the seismic data, and the picked same phase axis can be used for researching stratum sequence, predicting a reservoir and extracting storage characteristics.
In order to achieve the above purpose, the technical scheme adopted by the invention is that the method for detecting the same phase axis of the seismic data based on the environmental suppression and contour combination model of visual cognition comprises the following steps:
step 1): processing the seismic data by using a self-adaptive direction Gabor energy filter;
step 2): under the processing result of the step 1), constructing a Gaussian difference function (DoG) of a central environment to realize an environment suppression effect;
step 3): providing a local energy aggregation function combined by the contours according to edge symbiotic and co-circular constraint, realizing a Gestant continuity criterion, enhancing a consistent space structure, and enabling the contours to protrude from a disordered background;
step 4): iterating according to the result of the step 2) restraining the background noise and the step 3) enhancing the signal until the termination condition is met, and then exiting the iteration; otherwise, returning to the step 1);
step 5): and (3) performing binarization processing on the detection result in the step (4) by using a non-maximum value inhibition and hysteresis threshold method.
Further, in the step 4), there are two termination conditions for the iterative computation, and the iteration is terminated when any one of the termination conditions is satisfied: terminating the iteration once the hard threshold condition is met, namely the iteration number reaches a given threshold value; secondly, under the condition of a soft threshold, a group trunk obtained by manual drawing is used for defining an ideal output which can be realized, and a contour detection evaluation operator is used for reflecting the edge detection performance of an algorithm; e (E) DO 、B DO Respectively representing a contour image set and a background image set in the group trunk, E D 、B D Respectively representing a contour image set and a background image set after the operation result of the event detection operator;
E＝E DO ∩E D representing the actual set of contours detected, E FN ＝E DO ∩B D Represents a missed contour present in the group contour, E FP ＝E D ∩B DO Representing the detected false contours; the edge detection performance operator is defined as:
card (X) represents the number of elements of set X; p=1, then the true in-phase is correctly detected, the closer P is to 0, meaning that more in-phase is missed or false detected.
Further, in the step 1), the seismic data is preprocessed, and then is processed by a self-adaptive direction Gabor energy filter, wherein the preprocessing is a wavelet fusion of the envelope peak instantaneous frequency and the inclined superposition peak instantaneous amplitude of the pre-stack seismic data.
Furthermore, in the step 1), the direction selection method of the Gabor filter adopts an adaptive direction selection method based on Radon transformation.
Furthermore, the degree of inhibition of the responses of the non-classical receptive fields to cells in the receptive fields in different directions and at different positions is different, and the step 2) adopts an anisotropic environment inhibition method, namely, the anisotropic environment inhibition based on the primary visual cortex of biological vision, and the method can detect the position where the azimuthal homogeneity decays relative to the isotropic inhibition.
Further, in the step 3), the in-phase axis curvature of the seismic data has the characteristics of continuity and low curvature, and under the guidance of a gelalt edge perception aggregation criterion, the characteristics of vision when perceiving a space structure with consistency are described by adopting co-circle constraint and combining the preference of vision for the low curvature and the continuous curvature.
Further, in the step 5), a non-maximum suppression and hysteresis threshold method of a canny edge detection operator is adopted to process the result obtained in the step 4), and then edge connection is carried out to obtain a binarized image.
Further, the seismic data is in sgy format.
Further, in-phase axis picking is performed based on pre-stack seismic data.
The invention provides a seismic data on-phase axis pickup method based on biological vision primary visual cortex environment inhibition and space enhancement function, which adopts a human vision primary visual cortex environment perception mechanism, and a neuroscientist research shows that: the biological vision primary visual cortex has the functions of inhibiting environmental information and enhancing the circulation excitation of an interest target, so that the visual cortex can rapidly and efficiently extract the outline or boundary of an observed object from a complex environment, and important characteristic information is provided for subsequent visual cognition. The invention constructs a seismic data on-phase axis pickup method based on analyzing a biological vision primary visual cortex information processing mechanism. The calculation model suppresses uncorrelated environmental noise while enhancing the target contour of the same phase axis of the seismic data, thereby realizing the significance extraction of the target contour of the same phase axis.
Compared with the prior art, the invention has the advantages that the invention provides the method for picking up the same phase axis of the seismic data based on the biological vision primary visual cortex environment inhibition and the space enhancement function, adopts the human vision primary visual cortex environment perception mechanism, reduces the response of texture edges in the background, highlights the boundaries of areas, enables smooth contours to be highlighted from the disordered background, carries out anisotropic inhibition on the non-relevant noise of the environment while enhancing the contour of the same phase axis of the seismic data, realizes the remarkable extraction of the target contour of the same phase axis, and prepares for subsequent seismic interpretation, reservoir prediction and the like.
Furthermore, the invention firstly carries out preprocessing on the seismic data, and obtains the fusion of the instantaneous frequency of the envelope peak value and the instantaneous amplitude of the inclined superposition peak value, thereby providing the optimal quality data to the maximum extent; compared with the traditional method, the method can provide a reliable angle range and improve the recognition rate of the filter during decomposition; specifically, the conventional Gabor filter adopts fixed directions with certain intervals, such as 0 °, 45 °, 90 °, 135 °, generally 4 or 8 directions, but the conventional direction selection method ignores characteristics of image content and differences among classes, so in order to better embody the direction selection characteristics of the Gabor filter, the invention provides a Radon transform-based adaptive direction selection method. After wavelet transformation is carried out on the original seismic data to obtain the instantaneous amplitude, local Radon transformation is carried out on the basis of the instantaneous amplitude profile. For each sampling point, the average value of the instantaneous amplitudes in different inclination angles is obtained, the maximum value in the average value is picked up, and the value is placed at the corresponding position to form the inclination superposition peak amplitude profile used in the step 1), and the inclination angle corresponding to the maximum value forms the inclination angle profile. Based on the inclination angle section, the whole section is traversed by adopting an edge connection method, the continuously-changing inclination angle (namely corresponding to the related phase axis) is found, the angle range and the interval of the inclination angle are determined, a reliable angle range is provided for the decomposition of the Gabor filter, and the recognition rate of the filter during the decomposition can be improved.
Furthermore, the invention adopts an anisotropic environment inhibition method, and can detect the edge variation more accurately than an isotropic method. When the contour aggregation is enhanced, the invention establishes a consistent space structure perception model by adopting the co-circle constraint and combining the characteristics of low curvature, continuously-changing curvature and the like of the common phase axis of the earthquake.
Furthermore, in the field of image processing, gray images are commonly used, elements of the gray images are pixels, a part of elements of the images also comprise colors and the like, and seismic data is sgy format and contains a large amount of information.
Furthermore, the invention performs on-phase axis pickup based on pre-stack seismic data. The pre-stack seismic data contains more abundant information than post-stack seismic data, including common shot gathers acquired by the earthquake, VSP data and the like; and the amplitude and frequency information is not damaged without NMO stretching and superposition, and some fine stratum characteristics can be reflected, but the disadvantage is low signal-to-noise ratio and high noise. Therefore, the invention provides a high-precision in-phase axis picking method aiming at the characteristic of low signal-to-noise ratio of pre-stack seismic data.
Drawings
FIG. 1 is a schematic diagram of a Radon transform.
Fig. 2 is a wavelet fusion flow chart.
FIG. 3 is a schematic illustration of environmental suppression, equidirectional suppression, highlighting significant objectives.
FIG. 4 (a) original prestack CSP data; fig. 4 (b) instantaneous amplitude profile.
Fig. 5 is an envelope peak instantaneous frequency.
Fig. 6 is a wavelet fusion profile of the envelope peak instantaneous frequency and the oblique superimposed peak amplitude.
Fig. 7 is a dip profile obtained by Radon transform.
Fig. 8 is the real part of the Gabor filter.
FIG. 9 (a) is a desired "group trunk" output; FIG. 9 (b) shows the result of inhibition and enhancement; fig. 9 (c) is a result of the on-phase axis pickup of the binarization map.
Fig. 10 is a flow chart of a non-classical receptive field suppression based seismic data event detection algorithm.
Detailed Description
The invention is further described below with reference to the drawings and the detailed description.
The phase axis is defined as the connecting line of phase peaks (peaks or troughs), and phase axis pickup is the basis of seismic interpretation and can be used for researching stratum of an stratum sequence, predicting a reservoir and extracting storage characteristics. The invention carries out in-phase axis pickup based on pre-stack seismic data, the pre-stack seismic data contains more abundant information than post-stack seismic data, the pre-stack seismic data is not stretched and overlapped by NMO, the amplitude and frequency information is not damaged, and some fine stratum characteristics can be reflected, but the invention has the defects of low signal to noise ratio and large noise. Therefore, the invention provides a high-precision in-phase axis picking method aiming at the characteristic of low signal-to-noise ratio of pre-stack seismic data, as shown in fig. 10;
firstly, preprocessing the seismic data, wherein the preprocessing is wavelet fusion of the envelope peak instantaneous frequency and the inclined superposition peak instantaneous amplitude of the pre-stack seismic data; then processing the adaptive direction Gabor energy filter, wherein the direction selection method of the Gabor energy filter adopts an adaptive direction selection method based on Radon transformation;
constructing a Gaussian difference function (DoG) of a central environment to realize an environment suppression effect; providing a local energy aggregation function combined by the contours according to edge symbiotic and co-circular constraint, realizing a Gestant continuity criterion, enhancing a consistent space structure, and enabling the contours to protrude from a disordered background; after iteration, the result is binarized by using a non-maximum value inhibition and hysteresis threshold method.
In an embodiment of the present invention, it is assumed that the source wavelet is approximated with a constant phase wavelet having 4 parameters as follows:
where σ is the modulation angular frequency, δ is the energy attenuation factor, and A' and φ are the amplitude and phase, respectively. Fourier transform is performed on both sides of the (1) to obtain:
according to the definition of Barnes, the Envelope Peak Instantaneous Frequency (EPIF) of a constant-phase wavelet is the average frequency of the wavelet amplitude spectrum weighted against Fourier frequencies, i.e
Substituting the mode of the formula (2) into the formula (3) to obtain EPIF of zero-moment source wavelet:
the tilt superimposed peak amplitude is calculated based on the instantaneous amplitude, which is calculated as follows:
A(t)＝|s(t)+i·H[s(t)]| (5)
wherein, the liquid crystal display device comprises a liquid crystal display device,for the resolved part of the signal calculated with wavelet transform, A (t) is the modulus of the instantaneous amplitude, S (t) is a trace of the seismic data, S (b, a) is the wavelet transform of the seismic data,the wavelet transform for g (t) for which the wavelet function g (t) is the real part of the Fourier transform, a for which the scale factor b for which the shift factor is, s (t) for g (t) is defined as:
wherein t, b.epsilon.R, a>0；g(t)∈L 1 (R,dt)I L 2 (R,dt)，Is the complex conjugate of g (t).
The local Radon transformation is the inclined superposition, and the local linear Radon transformation is adopted in the invention. The integral path of a linear Radon transform is linear, also known as a tau-p transform, whose discrete form in the frequency domain is:
wherein M (f, p) = ≡m (τ, p) e -j2πfτ dτ，For instantaneous amplitude A (t, x m ) Fourier transform, x m For reference, m (τ, p) is the Radon transform of the time domain. The calculation process of the local linear Radon transform is shown in fig. 1: selecting a reference track on the instantaneous amplitude profile, and for a certain time intercept τj on the reference track, taking several tracks near the reference track along n p Each having a different slope p j (j＝1,2,L,n p ) Is superimposed (slope sampled at intervals of Δp), the sum of the instantaneous amplitudes superimposed in different directions at the time intercept is calculated, and the superimposed value is recorded at the position (τ) corresponding to the τ -p coordinate axis j ,p j ) When the selected superposition slope is close to or equal to the slope of the in-phase axis, the superposition value of the record in the t-x domain along the straight line is maximum. The average value of the maximum overlap value is placed at the corresponding position (τ) in the t-x domain (time-distance domain) j ,x m ) In this regard, a superstrate can be constructed to increase the signal-to-noise ratio, referred to as a sloped superimposed peak amplitude profile, with the slope corresponding to the maximum superimposed value also placed at the corresponding location in the t-x domain, constituting a gradient profile. The calculation method of the linear Radon transformation is various, and the invention adopts high-precision frequency-space domain matrix phase multiplication calculation. Before Radon transformation, the instantaneous amplitude data in the time domain is zero-padded to increase the frequency resolution, and the zero-padded length is three times of the data length in the invention.
The envelope peak instantaneous frequency and the oblique superimposed peak amplitude have been found so far, and then the two are wavelet fused. Let f (t) 1 ,t 2 )∈L 2 (R 2 ) Represents one at R 2 Two-dimensional signal on, f (t 1 ,t 2 ) T in (b) 1 And t 2 The abscissa and the ordinate, respectively. Psi (t) 1 ,t 2 ) The mother function representing the two-dimensional wavelet is subjected to resolution stretching and displacement to obtain a definition formula of the two-dimensional continuous wavelet mother function as follows:
the two-dimensional continuous wavelet transform is as follows:
in the method, in the process of the invention,and as a normalization factor, the conservation of energy before and after wavelet stretching can be ensured. In order to be able to implement a two-dimensional wavelet transform in a computer, it is necessary to implement discretization of the two-dimensional wavelet (DPWT). The two-dimensional discrete wavelet transform is shown as follows:
in the formula, in two-dimensional continuous waveletj、k 1 、k 2 ∈Z，a 0 、τ 10 、τ 20 Is constant. And carrying out progression after the integration discretization, so as to obtain wavelet transformation in a two-dimensional discrete space. The two-dimensional Discrete Spatial Wavelet Transform (DSWT) is shown as follows:
and finally, performing specialization on the two-dimensional discrete space wavelet transformation to obtain the two-dimensional discrete wavelet transformation. The two-dimensional discrete wavelet transform is shown as follows:
it can be seen that a is 0 ＝2，τ 10 ＝τ 20 ＝1。
The two-dimensional image fusion based on wavelet transformation mainly comprises three steps:
(1) And respectively carrying out wavelet decomposition on each original image to be fused to obtain decomposition coefficients in different directions.
(2) And carrying out fusion processing on the coefficients in different directions obtained by decomposition.
(3) And carrying out wavelet inverse transformation on the fused coefficient to obtain a final result image.
The two-dimensional wavelet transform-based image fusion process is shown in fig. 2.
The two-dimensional Gabor function can approximate the characteristics of a simple cell receptive field in the visual cortex of a mammal, is a Gaussian function modulated by a complex sine function, and has the following expression:
g(x,y)＝h(x',y')exp(j2πfx')(13)
in the method, in the process of the invention,θ is the direction of the filter, and filtering in any direction can be realized through rotation of a coordinate system, which is showing the characteristic of sensitivity of the cell receptive field to the direction selectivity, f represents the center frequency, determines the effective area of the filter, and h (x, y) is a gaussian function, and has the expression:
σ x ，σ y is the covariance of the gaussian function in the horizontal and vertical directions. The traditional Gabor filter adopts fixed directions, such as 0 degree, 45 degree, 90 degree, 135 degree and the like, has low recognition rate, does not consider the characteristics of object contents and the differences among classes, and the fixed selected directions are not as high as possible. In consideration of the specificity of the on-phase axial gradient of the seismic data, the inclination angle section obtained by Radon transformation is used for setting the direction of the filter, so that the recognition rate can be improved, and the method is better than the traditional method. The tilt angle profile is obtained when the tilt superimposed peak amplitude profile is obtained. The inclination of the same phase axis is continuous, so that from the first row of the inclination profile, the surrounding points of one point are detected, and by comparing the surrounding points (for CSP data, only the surrounding points on the right side can be comparedThe calculated amount is saved), if the difference value between the amplitude and the reference track is within a certain range, the difference value is considered as the inclination angle on the same phase axis, and the value of the point is reserved; otherwise, if the noise exceeds a certain range, the noise is considered as noise, and the noise is set to zero. After traversing all elements, picking up the maximum value and the minimum value as the upper limit and the lower limit of the Gabor filter, and carrying out Gabor integral transformation at intervals of 5 degrees. The 5 ° spacing is used because angular resolution below 5 ° creates false directions.
The non-classical receptive fields allow for a difference in the extent of inhibition of cellular responses in the receptive fields in different directions and at different locations. As shown in fig. 3, 2 edges of the triangle are highlighted in a direction different from the background texture, while edges in a direction identical to the background texture are completely suppressed, so that non-classical receptive field suppression is an essential feature of biological boundary detection, by which the boundary and isolated outline of the region are mainly detected. Unlike on-phase axis picking, the object to be picked has consistent or continuous texture, which needs to be highlighted, and background noise, which is disordered, needs to be completely suppressed. Therefore, in the invention, the relation between the degree of inhibition and the azimuth is expressed by the following formula in consideration of the anisotropic inhibition characteristic of the non-classical receptive field:
Δ(x,y,x-x 0 ,y-y 0 )＝|sin(Θ θ (x,y)-Θ θ (x-x 0 ,y-y 0 ))| (15)
wherein Θ is θ (x, y) represents the gradient direction angle of the point (x, y), Δ (x, y, x-x) 0 ,y-y 0 ) Representing points (x, y) and points (x, y, x-x) 0 ,y-y 0 ) And the suppression weights due to the direction difference. In addition to the direction influencing the degree of inhibition, the distance also influences the degree of inhibition between cells, and the invention uses a gaussian difference function to simulate the distance inhibition relationship between cells:
wherein, the liquid crystal display device comprises a liquid crystal display device,the value of g is the first order norm of the vector, w σ (x, y) represents a distance weight. The total inhibition of the environment for a given stimulation point is:
i represents the environmental point around the point of interest o, M i The excitation response of the environmental stimulus point, i.e. the gray value of the pixel or the response value of the primary cell, is represented.
Gestalt perception criteria tell us that vision can aggregate features that locally have a consistency rule into a whole, enabling objects to stand out from a complex background. The co-circular constraint determines a local aggregation function of the greatest likelihood for contour aggregation in natural scenes and further considers that the degree to which neurons are enhanced by external environmental stimuli depends on the curvature of the two neuron preferential directions. The curvature of the same phase axis of the earthquake has consistency or continuity, and if the curvature of the co-circular outline is smaller, the strengthening effect is stronger. To reflect the effect of low curvature, the excitement is weighted according to curvature, and the weighting function expression is:
wherein, the liquid crystal display device comprises a liquid crystal display device,for curvature (S)>Representing the distance between two neurons, σ=0.3 is a constant, α, β are the preferred orientations of the two neurons, and γ is the orientation of the two neuron lines. The exponential function is decaying, so the smaller the curvature the greater the weight. The receptive field center point (x ', y'), the neurons whose gradient direction angle α are subjected to the enhancement input of the external environmental neurons are:
wherein M (x ', y', beta) is the response value of the neuron located at (x ', y'), and beta is the preferred direction angle of the neuron at (x ', y'), namely the gradient direction angle of the pixel point.
In order to obtain a definite binarization result, the invention adopts a method of non-maximum suppression and hysteresis threshold of a canny edge detection operator for processing. The principle of non-maximum value inhibition refinement is as follows: dividing the periphery of the point (x, y) into four adjacent areas, and dividing the gradient direction theta of the point (x, y) σ (x, y) for the respective region, comparing the gradient value of the point (x, y) with the gradient values on (x ', y') and (x ", y") of the region, and if the gradient value of the point (x, y) is the largest, retaining; otherwise, the pixel for that point is set to 0.
The principle of the hysteresis threshold method is as follows: the image is scanned by taking a double threshold. Let two threshold values be gamma A And gamma B And meet gamma A ＝β·γ B (0 < beta < 1). Greater than threshold gamma A The pixels of (1) are called quasi-contour points, given a reservation, set as image B; less than gamma A Is a non-contour point, giving a removal; the pixel value is at gamma A And gamma B The candidate contour points are in between, and the filtered image is A. Performing contour line tracking from a point of the image B, and if the pixel value in the image A is contained in the tracking path, considering the corresponding point in the image A as a contour line and reserving the contour line; otherwise, removing points outside the tracking line to realize edge connection.
Calculation case analysis
We will propose a method for the actual common shot gather (CSP) with 595 shots, a minimum offset of 90m, a distance between detectors of 10m,1250 samples, and 2ms samples. Fig. 4 (a) is original pre-stack seismic data, and it can be seen from the figure that the original data has obvious in-phase axis, but the background noise is also very large, especially in-phase axis at the near offset distance and shallow earth surface can not be distinguished, and the wave crest and the wave trough of the in-phase axis are mutually influenced, which is not beneficial to picking up other attribute information such as dip angle. Fig. 4 (b) is an instantaneous amplitude profile. Fig. 5 is a profile of the envelope peak instantaneous frequency (epaf), from which it can be seen that the envelope peak instantaneous frequency reduces the mutual interference of peaks, troughs, the information is further concentrated, but the noise in the deep layer is still very large. Fig. 6 shows the wavelet fusion result of the epaf and the oblique superimposed peak amplitude profile, and it can be seen that the fused image highlights the same phase axis, and the influence of noise is weakened to some extent. Fig. 7 is a dip profile of Radon transform pickup. Fig. 8 is the real part of a Gabor filter having 8 directions and 5 dimensions. Fig. 9 (a), 9 (b) and 9 (c) are results of picking up the same phase axis by the method of the present invention, it can be seen that the picking up of the same phase axis in the shallow layer is complete, the deep layer is submerged in the noise due to the larger noise, the picking up of the same phase axis is less, but the picking up of the same phase axis in the thick stratum is complete, which illustrates the effectiveness of the present invention.
Claims (8)
1. The seismic data event detection method based on the visual cognition environment suppression and contour combination model is characterized by comprising the following steps of:
step 1): processing the seismic data by using a self-adaptive direction Gabor energy filter;
step 2): under the processing result of the step 1), constructing a Gaussian difference function of a central environment to realize an environment suppression effect;
step 3): providing a local energy aggregation function combined by the contours according to edge symbiotic and co-circular constraint, realizing a Gestant continuity criterion, enhancing a consistent space structure, and enabling the contours to protrude from a disordered background;
step 4): iterating according to the result of the step 2) restraining the background noise and the step 3) enhancing the signal until the termination condition is met, and then exiting the iteration; otherwise, returning to the step 1);
step 5): performing binarization processing on the detection result in the step 4) by using a non-maximum value inhibition and hysteresis threshold method;
in the step 1), the seismic data is preprocessed and then is processed by a self-adaptive direction Gabor energy filter, wherein the preprocessing is wavelet fusion of the instantaneous frequency of the envelope peak value and the instantaneous amplitude of the inclined superposition peak value of the pre-stack seismic data.
2. The method for on-phase detection of seismic data based on a visual-cognitive environmental suppression and contour combination model according to claim 1, wherein the termination condition of the step 4) is that the iteration number satisfies a hard threshold condition or that the detection result satisfies a visual-perceived ground score.
3. The method for detecting the same phase axis of the seismic data based on the visual-recognition environment suppression and contour combination model according to claim 1, wherein in the step 1), the direction selection method of the Gabor filter adopts an adaptive direction selection method based on Radon transformation.
4. The method for on-phase detection of seismic data based on a visual-recognition-based environmental suppression and contour combination model according to claim 1, wherein the step 2) adopts an anisotropic environmental suppression method.
5. The method for on-phase detection of seismic data based on a visual-recognition environmental suppression and contour combination model according to claim 1, wherein in the step 3), the symbiotic co-circle constraint is adopted in combination with the preference of vision for low curvature and continuous curvature to describe the characteristics of vision when perceiving a space structure with consistency.
6. The method for on-phase detection of seismic data based on a visual-recognition environment suppression and contour combination model according to claim 1, wherein in the step 5), the result obtained in the step 4) is processed by adopting a non-maximum suppression and hysteresis threshold method of a canny edge detection operator, and then edge connection is performed to obtain a binarized image.
7. The visual-cognition-based environmental suppression and contour-combining-model-based seismic data event detection method of claim 1, wherein the seismic data is in sgy format.
8. The visual-cognition-based environmental suppression and contour combination model-based seismic data event detection method according to claim 1, wherein event picking is performed based on prestack seismic data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110249015.1A CN113031059B (en) | 2021-03-08 | 2021-03-08 | Visual cognition-based seismic data event detection method based on environment suppression and contour combination model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110249015.1A CN113031059B (en) | 2021-03-08 | 2021-03-08 | Visual cognition-based seismic data event detection method based on environment suppression and contour combination model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113031059A CN113031059A (en) | 2021-06-25 |
| CN113031059B true CN113031059B (en) | 2023-10-13 |
Family
ID=76466902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110249015.1A Active CN113031059B (en) | 2021-03-08 | 2021-03-08 | Visual cognition-based seismic data event detection method based on environment suppression and contour combination model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113031059B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114002739B (en) * | 2021-11-11 | 2024-01-26 | 中海石油(中国)有限公司 | Edge detection method, device and medium based on geometric non-parallel statistical attribute |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103903251A (en) * | 2012-12-30 | 2014-07-02 | 南京理工大学 | Night vision image salient contour extracting method based on non-classical receptive field composite modulation |
| CN104714251A (en) * | 2015-03-24 | 2015-06-17 | 西安交通大学 | Oblique stacking peak amplitude portion edge detecting method for event automatic picking |
| CN107515420A (en) * | 2017-04-28 | 2017-12-26 | 西安石油大学 | It is a kind of for local correlation lineups when walking with gradient precision pick method |
| CN107644400A (en) * | 2016-07-21 | 2018-01-30 | 中国石油化工股份有限公司 | The Texture Segmentation Methods and device of seismic section image |
| CN109709603A (en) * | 2018-11-23 | 2019-05-03 | 中国石油天然气股份有限公司 | Seismic horizon identification and method for tracing, system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7203342B2 (en) * | 2001-03-07 | 2007-04-10 | Schlumberger Technology Corporation | Image feature extraction |
| CN111796324B (en) * | 2019-04-09 | 2023-02-10 | 中国石油天然气股份有限公司 | Seismic all-horizon tracking method and device |
-
2021
- 2021-03-08 CN CN202110249015.1A patent/CN113031059B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103903251A (en) * | 2012-12-30 | 2014-07-02 | 南京理工大学 | Night vision image salient contour extracting method based on non-classical receptive field composite modulation |
| CN104714251A (en) * | 2015-03-24 | 2015-06-17 | 西安交通大学 | Oblique stacking peak amplitude portion edge detecting method for event automatic picking |
| CN107644400A (en) * | 2016-07-21 | 2018-01-30 | 中国石油化工股份有限公司 | The Texture Segmentation Methods and device of seismic section image |
| CN107515420A (en) * | 2017-04-28 | 2017-12-26 | 西安石油大学 | It is a kind of for local correlation lineups when walking with gradient precision pick method |
| CN109709603A (en) * | 2018-11-23 | 2019-05-03 | 中国石油天然气股份有限公司 | Seismic horizon identification and method for tracing, system |
Non-Patent Citations (8)
| Title |
|---|
| 《Automatic events extraction in pre-stack seismic data based on edge detection in slant-stacked peak amplitude profiles》;Jing Zhao et al;Journal of Petroleum Science and Engineering;第459-466页 * |
| 《Gabor滤波器的纹理分割》;黄伟;盐城工学院学报(自然科学版);第21卷(第3期);第9-13页 * |
| 《N次根倾斜叠加方法在间断面研究中的应用》;臧绍先等;地球物理学报;第45卷(第3期);第407-415页 * |
| 《利用LOG算子和改进相干算法提高地震资料分辨率》;杨立强,郝天珧,宋海斌;物探与化探;第29卷(第4期);第351-354页 * |
| 《利用改进相干算法提高地震资料分辨率的有效算法》;杨立强;中国科学院研究生院学报;第22卷(第3期);第359-362页 * |
| 《基于初级视皮层感知机制的轮廓与边界检测》;唐奇伶;中国优秀博士学位论文全文数据库 信息科技(第2009年第05期期);第15-73页 * |
| 《应用小波变换和C3相干算法检测地震剖面同相轴》;杨立强等;物探化探计算技术;第26卷(第4期);第302-305页 * |
| 《数学形态学在地震同相轴边缘检测及机器视觉中的研究与应用》;赵静等;IFEDC;第1-9页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113031059A (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hu et al. | First-arrival picking with a U-net convolutional network | |
| US8095319B2 (en) | System and method for fault identification | |
| Schneider | Developments in seismic data processing and analysis (1968–1970) | |
| Di et al. | Improving seismic fault detection by super-attribute-based classification | |
| Admasu et al. | Autotracking of faults on 3D seismic data | |
| CN109001801B (en) | Fault variable-scale identification method based on multiple iteration ant colony algorithm | |
| Wang et al. | Data-driven microseismic event localization: An application to the Oklahoma Arkoma basin hydraulic fracturing data | |
| MX2015000034A (en) | Methods and systems for optimizing generation of seismic images. | |
| Dou et al. | Attention-based 3-D seismic fault segmentation training by a few 2-D slice labels | |
| CN113031059B (en) | Visual cognition-based seismic data event detection method based on environment suppression and contour combination model | |
| CN104730572B (en) | A kind of diffracted wave formation method based on L0 semi-norm and device | |
| CN114859410A (en) | Seismic event extraction method of B-COSFIRE filter based on differential evolution algorithm | |
| CN110261905A (en) | Complex value based on pitch angle control is concerned with microfault recognition methods | |
| Zhao et al. | A Comprehensive Horizon‐Picking Method on Subbottom Profiles by Combining Envelope, Phase Attributes, and Texture Analysis | |
| Xing et al. | Ground roll intelligent suppression based on spatial domain synchrosqueezing wavelet transform convolutional neural network | |
| Picard et al. | Seafloor description in sonar images using the monogenic signal and the intrinsic dimensionality | |
| Zhong et al. | Multiscale residual pyramid network for seismic background noise attenuation | |
| Mora et al. | Fault enhancement using probabilistic neural networks and Laplacian of a Gaussian filter: A case study in the Great South Basin, New Zealand | |
| Wang et al. | High-resolution seismic faults interpretation based on adversarial neural networks with a regularization technique | |
| Zhao et al. | Effective detection of seismic events by non-classical receptive field visual cognitive modelling. | |
| Zhou et al. | Surface diffraction noise attenuation for marine seismic data processing with mathematical morphological filtering | |
| Wu et al. | High-resolution imaging: An approach by compensating absorption and dispersion in prestack time migration with effective Q estimation and Fresnel zone identification based on deep learning | |
| US20230375735A1 (en) | Detection of hydrocarbon presence in subsurface from seismic images using relational learning | |
| Kim | Machine learning applications for seismic processing and interpretation | |
| Li et al. | N-th root slant stack for enhancing weak seismic signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |