CN110728228A

CN110728228A - Seismic wave image fault identification method based on ant colony tracking algorithm

Info

Publication number: CN110728228A
Application number: CN201910955253.7A
Authority: CN
Inventors: 任仲远; 陈凯; 吕涛; 张楠楠; 许鹏飞; 刘宝英; 汤战勇
Original assignee: Northwest University
Current assignee: Northwest University
Priority date: 2019-10-09
Filing date: 2019-10-09
Publication date: 2020-01-24
Anticipated expiration: 2039-10-09
Also published as: CN110728228B

Abstract

The invention belongs to the field of image recognition, and discloses a seismic wave image fault recognition method based on an ant colony tracking algorithm. And performing morphological data processing on the seismic wave image according to the characteristics, wherein the morphological data processing mainly comprises image enhancement, image binarization, formation line thinning and thickening, breakpoint filtering, pseudo breakpoint removal and the like, and the positions of the breakpoints are obtained from the seismic wave image through the operations. And fitting a line segment by carrying out Hough transform on the obtained breakpoints to determine the position of the break layer in the clear picture. For an image with unclear image or distorted fault lines, the result of Hough transform is used as a fault range, and a watershed image of the image is cut according to the range. And obtaining the accurate position of the fault by using an ant colony algorithm on the cut image.

Description

Seismic wave image fault identification method based on ant colony tracking algorithm

Technical Field

The invention belongs to the field of image identification, and particularly relates to a seismic wave image fault identification method based on an ant colony tracking algorithm.

Background

In seismic data, faults are mainly represented by displacement or interruption of a reflection horizon, the amplitude of the fault is discontinuous in the transverse direction, and the reflection amplitude has obvious difference on two sides of the fault, and the characteristics can be recognized by naked eyes on a seismic wave image. Therefore, for the analysis and interpretation of fault systems, the conventional approach is to pick up faults line by line along the main stratigraphic line direction, or to select the inline direction perpendicular to the fault strike. The interpreter typically uses time slices or along-the-horizon slices to control the spatial extent and contrast of faults, and amplitude slices at the same time may contain information of different horizons due to fluctuations in the formation's attitude, and using along-the-horizon slices is an improved approach. However, the use of the slice along the horizon to control the spatial extent of the fault is affected primarily by the artifacts of horizon interpretation and is less reliable in the case of poor seismic data quality. Therefore, the conventional fault interpretation and analysis method has the defects of long period, strong subjectivity and the like, and when the fault system is relatively complex, the conventional interpretation method has low reliability and limited interpretation capability.

Disclosure of Invention

The invention aims to provide a seismic wave image fault identification method based on an ant colony tracking algorithm, which is used for solving the problem of low reliability in the prior art when a complex fault system is analyzed and identified.

In order to realize the task, the invention adopts the following technical scheme that:

step 1: acquiring a seismic wave image to be identified, and preprocessing the seismic wave image to be identified to obtain an effect enhancement image;

step 2: performing breakpoint extraction on the effect enhanced image obtained in the step 1 to obtain a plurality of breakpoints, judging the relative positions of the upper and lower walls of the fault of the seismic wave image to be identified in the step 1, and executing a step 3 if the relative positions of the upper and lower walls of the fault are unchanged; if the relative positions of the upper and lower plates of the fault are changed, executing a step 4;

and step 3: carrying out Hough transform on the plurality of break points obtained in the step 2 to obtain fault lines of the seismic wave image to be identified;

and 4, step 4: firstly, carrying out Hough transform on a plurality of breakpoints obtained in the step 2 to obtain a fault line range of the seismic wave image to be recognized, then cutting off watershed of the image according to the fault line range, and finally fitting the fault line by utilizing an ant colony tracking algorithm according to the cut-off image to obtain the fault line of the seismic wave image to be recognized.

Further, the preprocessing of step 1 is to enhance the input seismic wave image by anisotropic diffusion filtering, where the expression of the anisotropic diffusion filtering is:

wherein u (x, y, t) represents the diffusion-filtered image at time t, u₀(x, y) is the initial image, c is the dominant factor for the degree of diffusion,

which represents the gradient of the image or images,

is the diffusion coefficient.

Further, the breakpoint extraction includes the following sub-steps:

step a: carrying out binarization processing on the effect enhanced image obtained in the step 1, and then extracting formation line information to obtain a binary image containing the formation line information;

step b: thinning, thickening and filtering the binary image containing the formation line information to obtain a binary image containing a pseudo breakpoint and a breakpoint;

step c: and c, removing the false breakpoints of the image obtained in the step b to obtain a plurality of breakpoints.

Further, the threshold of the short line of the hough transform is 40, and the threshold of the merging interval of the hough transform is 70.

Further, step 4 comprises the following substeps:

step 4.1: carrying out Hough transform on the plurality of break points obtained in the step 2 to obtain a fault line range of the seismic wave image to be identified, wherein the fault line range represents a set of points of which the distances from the points to two end points of the initial fault line are less than 15 after the Hough transform is carried out, and the fault line range is equal to the identification range of the ant colony tracking algorithm;

step 4.2: cutting off watershed of the image according to the range of the fault line;

step 4.3: and fitting the fault line by utilizing an ant colony tracking algorithm according to the cut image to obtain the fault line of the seismic wave image to be identified.

Compared with the prior art, the invention has the following technical characteristics:

(1) under the condition of only using the seismic wave image, the method does not need to carry out complex operations of large data quantity such as coherence analysis and the like on seismic wave data, and can realize fault line detection which is more accurate and has higher reference value.

(2) The method can perform morphological operation on the seismic wave image according to the morphological characteristics of the fault to obtain the breakpoint.

(3) The method can fit the fault position in the seismic wave image through an improved ant colony algorithm.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a clear seismic image to be identified in the embodiment;

FIG. 3 is a graph showing the effect of the embodiment after anisotropic diffusion filtering;

FIG. 4 is a detailed view of the formation in the example;

FIG. 5 is a diagram of the positions of breakpoints obtained by XOR and filtering in the embodiment;

FIG. 6 is a diagram of the clear image for identifying the position of the fault line after Hough transformation is performed on the breakpoint in the embodiment;

FIG. 7(a) is an unclear seismic wave image to be identified in the embodiment;

FIG. 7(b) is a schematic view of the fault line range in the embodiment;

fig. 7(c) is a schematic diagram of the ant forward path of the ant colony tracking algorithm in the embodiment;

FIG. 8 is a diagram of the location of the identified fault lines obtained by the ant colony tracking algorithm in the example.

Detailed Description

Example 1

The embodiment discloses a seismic wave image fault identification method based on an ant colony tracking algorithm, which comprises the following steps:

Due to the existence of clutter, the definition of the seismic wave image obtained by us often cannot meet the requirement of easy identification, so that the image enhancement is particularly important. Therefore, an inputted seismic wave image is enhanced by using anisotropic diffusion filtering through an MATLAB software platform, and an anisotropic diffusion filtering algorithm is an image enhancement algorithm in a space domain based on a heat conduction method in physics and mainly smoothes the image through selective smoothing; the smoothing is not limited to uniform regions, but is suppressed only across edges, thereby selectively reducing noise and interference with minor details.

Specifically, the preprocessing in step 1 is to enhance the input seismic wave image by anisotropic diffusion filtering, where the expression of the anisotropic diffusion filtering is:

wherein u (x, y, t) represents the diffusion-filtered image at time t, u₀(x, y) is an initial image, which is a more classical initial value problem in the field of partial differential equations, an image subjected to diffusion filtering at the moment can be obtained at different moments t, c is a main control factor of the diffusion degree,

representing the gradient of an image-it is a vector of gradients in image space, represented as

It is generally considered that, the gradient is a physical quantity reflecting the characteristics of an image,

is the diffusion coefficient. Is a decreasing function of the gradient amplitude, i.e. where the gradient is larger, the diffusion coefficient is smaller; where the gradient is small, the diffusion coefficient is larger, thus forming a directional adaptive diffusion, i.e. a filter function related to the local gradient of the image. Based on the principle, the method is applied to an internal coordinate system of the image characteristic direction, so that the diffusion mechanism becomes very intuitive and has the following characteristics:

(1) the smooth quantity is controlled in the region with more image features, and the smooth quantity is as small as possible or even not smooth; the smoothing amount is large in a region with few image features or a region without image features;

(2) the control of the smooth direction has a large diffusion amount along the image feature direction, and the diffusion amount passing through the image feature is small or even unsmooth. The image is smoothed by using anisotropic diffusion filtering, which preserves the image edge when smoothing the image, and bilateral filtering is very similar, thereby overcoming the defect of Gaussian blur. By using the filtering mode, the original image is processed to be more beneficial to recognition, and the fault layer information is highlighted.

Further, the breakpoint extraction includes the following sub-steps:

step a: carrying out binarization processing on the effect enhanced image obtained in the step 1, and then extracting formation line information to obtain a binary image containing the formation line information; in order to extract a breakpoint in a seismic wave image, a certain judgment needs to be made on the general trend of the stratum of the seismic wave image, in order to extract main information of the stratum in the image, redundant information except for forms in the image needs to be filtered continuously, for a gray image, namely, a threshold value is set to change the gray value of the gray image into a binary value, namely, black and white images are used for representing information in the gray image.

Step b: thinning, thickening and filtering the binary image containing the formation line information to obtain a binary image containing a pseudo breakpoint and a breakpoint; as can be known from morphological analysis, the position of the breakpoint is not directly connected with the thickness of the stratum, and non-position features in the stratum need to be deleted in order to obtain the position of the breakpoint and keep the main information of the fault. Therefore, the formation line thinning operation is carried out on the binarized image, so that the main existing information of the formation can be retained, and unimportant information can be filtered out, so that the trend of the formation line is more obvious; meanwhile, the length of the fault is artificially shortened after the thinning operation, so that the thinned formation line is subjected to thickening operation after main information is retained in the thinning operation, and fault information is amplified and recovered to be continuous in length.

Specifically, the stratigraphic line refining operation is to perform vertical upper bound and lower bound on the pixels of the stratigraphic lines in the image, then to take a median, and to delete the information except the median, so as to obtain the refined stratigraphic lines. By such processing, a refined stratigraphic image can be obtained. The thinned stratum line is subjected to thickening operation, and the thinned image is extended up and down by a certain length, so that the length of the fault is extended, main information of the position where the fault line does not continue is highlighted, and breakpoint judgment is facilitated.

Specifically, morphological analysis on the fault shows that the fault generally has a large fall in the stratigraphic direction, and points with overlarge vertical difference in the stratigraphic line can be extracted by performing exclusive or operation on the image after the stratigraphic thickening in the horizontal direction according to the principle. And then filtering out (filtering process) the dots whose vertical direction continuous number is equal to the number of thickened dots, so that the breakpoint with the highest probability can be obtained.

Specifically, by extracting stratum points with excessively large vertical differences as breakpoints, points with originally large vertical differences may be retained, and therefore filtering needs to be performed on the pseudo breakpoints. It can be known from morphological characteristics that (a fault has little or small drop in the horizontal direction but a stratum has strong vertical difference and can have certain continuity in the horizontal direction), the fault generally has continuity, so that an individual breakpoint cannot exist, and therefore, the individually existing breakpoint can be considered to be a false breakpoint generally, so that an image of a suspicious breakpoint obtained after filtering is filtered, and the individually existing breakpoint is deleted, that is, the false breakpoint can be deleted. The specific implementation method is that each point in the image is traversed, and the isolated points in the image, namely the points without break points around, are deleted. In the algorithm implementation we take the point values in a rectangular area of 7 × 7, consisting of 7 units per point in the horizontal direction and 7 units in the vertical direction, and add them, if the resulting value is 12240, that means that 48 points around the point are 255 (empty points). This point is deleted from the original image.

The positions of break points in the seismic wave image are obtained in the previous steps, but the break points are discrete, so that the fault lines cannot be directly obtained, and the fault lines are accurately fitted through Hough transform or Hough transform and ant colony algorithm according to different image types.

The collected seismic wave images are generally divided into two types, the first type of seismic wave images can obtain clear images in better stratums, the stratums are clear in trend, the fracture is obvious between a set of stratums, and the relative positions of the upper and lower plates of the fault are stable (as shown in figure 2). The second type of unclear image has more than three dislocation phenomena in a set of stratum due to uneven substances in the stratum or excessively complex fault positions, so that the positions of an upper plate and a lower plate of the fault alternately appear, and a large part of vacancy or distortion appears in the fault in a picture (as shown in fig. 7 (a)). The method respectively identifies the morphological operation and the ant colony algorithm of the two images. The first type of images have good stratum shapes, accurate breakpoints can be obtained after morphological operation, clear images of the fault can be obtained by fitting line segments of the breakpoints through Hough transform, and ant colony algorithm fitting is not needed for the images. And for another type of fault with a poor shape, after the fault existence range is identified through Hough transformation, an ant colony algorithm is carried out to fit a complete fault line.

Specifically, for a seismic wave image with a clear break layer in an original image, the position of a break point in the seismic wave image is obtained. And (3) identifying the breakpoint obtained by filtering through Hough transform, so as to obtain the accurate position of the breakpoint in the seismic wave image. The Hough transform can map straight lines in a planar coordinate system into points in a polar coordinate system by converting the planar coordinate system into the polar coordinate system; the more points passing through the straight line, the more times the same point in the polar coordinate system is mapped, the most satisfactory straight line in the plane coordinate system can be obtained by searching the point with the maximum mapping times,

preferably, the identification rule is: 1. identifying a single straight line as short as possible (the algorithm sets the threshold value of the short line to be 40, namely all line segments with the length being more than 40 are identified), so that each possible fault can be reserved; 2. the continuity of the fault is enhanced by merging line segments whose directions are consistent and whose intervals are smaller than a certain threshold (the selection threshold is 70). After Hough transformation is carried out on all breakpoints and an optimal result is selected, an accurate line segment of the position of the fault line is obtained.

Specifically, step 4 includes the following substeps:

watershed image segmentation is a mathematical morphology segmentation method based on a topological theory. The basic idea is to look at the image as a geodetic topological landscape, with the gray value of each point pixel in the image representing the height of that point. The watershed algorithm has good response to weak edges, and noise in the image and fine gray level changes on the surface of an object can be identified and reflected on the image. Therefore, the image obtained by the watershed algorithm can finely reflect the change of the stratum, so that the internal information of the stratum and the information between the stratum can be retained to the highest degree. As is clear from the analysis of the watershed image, the fault line is located substantially on the boundary of the watershed image, and therefore the fault line needs to be extracted from the complex watershed boundary.

The watershed boundary map can be obtained by operating on the grayscale map using a watershed function in MATLAB. However, as can be seen from the watershed image, the boundary line occupied by the fault line has a very small specific gravity, and if the ant algorithm is directly used, the whole watershed image is identified, so that a large number of iterations and the number of ants are required, and the efficiency of extracting the accurate fault line is very low. Therefore, the range of the fault line obtained by Hough transform is used, and the recognition range of the ant colony algorithm is cut according to the range. The concrete implementation is as follows: and storing the coordinates of two end points of the rough fault line segment obtained by Hough transformation into the matrix to obtain an analytical equation of the line segment. By traversing each point in the watershed image, finding a point on the analytic equation, and selecting a point whose distance from the point is less than a certain threshold (herein, the threshold is set to be 15), a fault existence range is formed. And the graph is used as a path graph for identifying the fault by an ant colony algorithm.

Specifically, the ant colony tracking algorithm comprises the following specific steps:

a. algorithm initialization

Firstly, parameters of the ant colony algorithm are assigned with initial values, which mainly comprise the following steps: the pheromone evaporation coefficient ρ is 2, the pheromone increase intensity coefficient Q is 9, the pheromone matrix τ is initially the same as the clip graph path and has a value of 1, the maximum number of iterations is NC _ MAX, the total number of ants is set to M, the ant M is M (M is 1,2,3 … M), the ants are randomly placed on the clip graph nodes on the bottom side during initialization, Tau is the clipped image, and imH is the Y value on the bottom side of the image. The find function takes the x-coordinates of the points on all bottom edges from the clip map and stores them in XINDEXes. When selecting the location where the initialization ant is placed, the coordinates in XINDEXes are randomly selected by the rand function.

b. Determination of ant advancing direction

The ants advance in four directions in the cutting drawing, which are respectively up, down, left and right, and are sequentially indicated by numbers 1,2,3 and 4 (as shown in fig. 5.6). Because the ant colony is completely arranged at the bottom of the image during initialization, the fault is always above the ant for the ant, so the probability of the ant advancing upwards is artificially increased, the probability of the ant advancing downwards is artificially decreased, and the probability of the ant advancing towards each direction next step is obtained, the calculation method is shown in a formula 5.1, and k is the direction of the ant.

c. Update of tabu tables

The tabu table is used as a path for finding a problem in the ant forward path, and points on the paths are written into the tabu table in order to avoid other ants and continue to select the paths in subsequent iterations. The taboo table mainly stores points which cannot be selected in the forward process of the ant colony. There are two main cases, headless and loop, respectively. The following operations are performed for updating the tabu table for these two cases, respectively.

The ants delete the points on the path point by point if encountering the broken path in the advancing process. The specific implementation is that when an ant encounters a path which cannot be taken, the point is placed into a taboo table, so that the ant cannot take the point again in the following ants and the following iterations, and the points in all broken paths can be placed into the taboo table in sequence.

When the ants have loops in the process of advancing, loop coincidence points are placed into the tabu table, and then the loop coincidence points are placed into the tabu table according to the way of processing broken ends, so that the loops are removed.

Where NewECsheet is the updated tabu table and AntRTY, which is the coordinates of the point that has currently been judged to be a loop or broken head.

d. Pheromone update

When the ant finishes fitting the path once, all ants finish walking, and an optimal path is born. At this time, the pheromone concentration is decreased by performing pheromone volatilization once on the whole pheromone matrix. And then, performing pheromone enhancement on the points on the fitted optimal path so as to complete the updating of the pheromone matrix. . Through the operation, ants can select the optimal path in each iteration more in multiple iterations, so that the accurate position of the fault line can be obtained.

And when each ant finishes one path, the ant dies, the total number of the ants is reduced by one, and AntNum is AntNum-1. When all ants die, namely all ants select a path on the graph to finish walking, one iteration is finished. But the iteration number is equal to the initialized maximum iteration number, namely the algorithm is ended when NC is equal to NC _ max. And outputting the optimal ant path. In this way, the position of the fault line in the unclear or fault distorted seismic wave image can be obtained.

Example 2

The embodiment discloses a seismic wave image fault identification method based on an ant colony tracking algorithm, and discloses the following technical characteristics on the basis of the embodiment 1:

to automate fault identification MATLAB2016 was used as an experimental environment. And writing corresponding codes and algorithms according to the previous image processing source, and storing the codes and algorithms into a working space for processing the images. The input image is then processed according to the processing flow and the position of the slice therein is displayed in the input image.

(a) Fault identification process of seismic wave image with obvious fault

1. Reading the original seismic wave image into an experiment platform, and displaying the result as shown in figure 2. From this, it can be seen that the image definition is low, the precise position of the fault line cannot be visually confirmed, and the noise around the formation line is too large, so that it is necessary to use image enhancement in order to obtain the fault form thereof.

2. The image enhancement is performed on the input original image, and the specific implementation is to smooth the original image by using anisotropic diffusion filtering to achieve the position of a smooth stratigraphic salient fault, and the result is shown in fig. 3.

3. The fault existence position can be confirmed in a large value by each different diffusion filtering result graph, but information irrelevant to the form, such as excessive colors, is still contained in the different diffusion filtering result graph, and main form information in the formation line is obtained by carrying out binarization and formation thinning on the image, as shown in fig. 4. In the position where the image visible faults are located at the stratigraphic line breaks, there are still many stratigraphic line faults due to non-fault reasons for which filtering is required.

4. The method comprises the steps of carrying out thickening operation on thinned stratum lines to strengthen main stratum information, restoring continuity of fault lines, carrying out horizontal XOR to obtain the jump points of the stratum in an image, and filtering line segments with the length smaller than the thickening width to obtain breakpoints caused by fault reasons. And then filtering the pseudo break points due to overlarge stratum drop, thereby obtaining the break points with the highest probability in the image. The results are shown in fig. 5.

5. And carrying out Hough transform straight line recognition on the obtained discrete breakpoints, fitting the most breakpoints into a line segment to obtain the position of the fault line in the image, and displaying the position in an original image, wherein the effect is as shown in figure 6.

(b) Fault identification process of seismic wave image with unobvious fault

1. The early identification of the seismic wave image with an unobvious fault position is approximately the same, and the specific process is not described again. The range of the obtained hough transform and the fault line obtained by clipping the watershed image using the hough transform are shown in fig. 7(a), 7(b), and 7 (c).

2. The existence position of the fault is then fitted using the ant colony algorithm by clipping the watershed image to the range of the obtained fault line and using it as a heuristic, as shown in fig. 8.

Claims

1. A seismic wave image fault identification method based on an ant colony tracking algorithm is characterized by comprising the following steps:

2. The method for identifying seismic wave image faults based on the ant colony tracking algorithm as claimed in claim 1, wherein the preprocessing of the step 1 is to enhance the input seismic wave image through anisotropic diffusion filtering, and the expression of the anisotropic diffusion filtering is as follows:

which represents the gradient of the image or images,is the diffusion coefficient.

3. The method for identifying seismic wave image faults based on the ant colony tracking algorithm according to claim 1, wherein the breakpoint extraction comprises the following sub-steps:

4. The method for identifying seismic wave image faults based on the ant colony tracking algorithm as claimed in claim 1, wherein a threshold value of a short line of the Hough transform is 40, and a threshold value of a merging interval of the Hough transform is 70.

5. The method for identifying seismic wave image faults based on the ant colony tracking algorithm as claimed in claim 1, wherein the step 4 comprises the following sub-steps: