JPH0438491A - Method of analyzing two-dimensional image - Google Patents

Abstract

PURPOSE: To determine a linear structure having approximate characteristic values by dividing each linear structure contained in an image into curve sections and displaying all sections having at least one characteristic value belonging to a predetermined same range. CONSTITUTION: Measurements are determined based on parameters along two axes and composed along a linear structure to form an image where the linear structure is distributed preferentially along one axis and then the linear structure contained in the image is mapped. Each linear structure is divided into a plurality of curve sections touching each other at the ends thereof and a curve section is selected such that the maximum distance between a point in the curve section and a line connecting the opposite ends thereof does not exceed a given limit value. Each section is then allotted with one characteristic value characteristic of all measurements of parameters associated with the section and all sections having at least one characteristic value belonging to a predetermined same range are displayed.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a method of construction particularly suitable for images having one preferred direction, such as images of seismic sections.
This relates to point 6). More specifically, the present invention includes:
The present invention relates to a method for automatically selecting a seismic ephemeris portion of an earthquake cross section having a certain number of characteristic values, which is used for seismic exploration, and performing structural analysis and stratigraphic analysis of the cross section.

In general, underground geology is considered to consist of a stack of multiple strata having different characteristic values, and these multiple strata are three-dimensionally organized according to a certain geometric form.

In order to elucidate the geometrical form of underground geology, underground geology experts, especially geophysicists involved in oil exploration,
It uses a special technique called "earthquake reflection method." In the seismic reflection method, a sound wave signal is transmitted from the earth's surface, and the sound wave signal is recorded as it propagates into the underground geology and is reflected at specific acoustic interfaces. Acoustic interface is the interface between different geological formations,
It is called a reflective surface.

The recorded signals are processed to obtain a detailed picture of the underground geology called a "seismic section." The seismic cross-section simulates a vertical cross-section of the underground geology, and in this cross-section the reflecting surfaces appear as a nearly horizontal or deformed linear structure (l 1neation) that overlaps each other, is more or less conspicuous, and is more or less continuous. Experts call this ``seismic layer quasi-J''.

A seismic section is called a "seismic trace" and consists of a series of vertical records that are continuous from left to right or right to left. A seismic trace consists of a set of signals of more or less large amplitude consisting of positive and negative arcs, each of which corresponds to one of the reflective surfaces.
Corresponds to an image of a point. The image signals of the same reflective surface correlate the traces to form a visible line structure.

Some of these line structures are extremely short but extremely important. For example, there is a ``double B'' between two distinct principal line structures consisting of a combination of arcs of amplitude of the same sign and separated over a wide range by an arc of signal of opposite sign. When signal arcs having a higher frequency and the same sign as the arcs constituting the structure are present over several traces, it is considered that a doublet exists in the seismic section.

Seismic sections prepared by geophysical researchers are entrusted to geophysical or geological experts who interpret said sections, for example to detect the presence or absence of oil fields. The latter task is
The method is to create a seismic stratigraphy map by vointing a number of predetermined seismic horizons in a plurality of matching seismic sections drawn in the same geographical area. Pointing work has traditionally been done manually. In this work, signals considered to belong to the same seismic horizon are combined using a color pen by optical correlation with each other.

Different calendar wheels are usually pointed with different colors to make them easier to identify. Image signals of the same reflective surface have almost the same characteristics such as form, frequency components, number of arcs, and amplitude.

Such drawing work is a laborious work that requires a long time, and is therefore expensive. In fact, when it is necessary to draw a large number of calendar wheels with a large number of cross sections, it is not uncommon for several experts to spend several weeks on one elucidation task.

For example, when conducting exploration in a very wide geographical zone,
Or, in the case of ``seismic exploration for ore deposits,'' which observes multiple cross-sections in a narrow zone but very close together, if the individual calendar wheels to be plotted are not separated enough or the geology of the exploration zone is not stable, it may be necessary to manually It is easy to make drawing errors during work. In particular, when a fault exists, the reflection plane is interrupted by the fault, so it is necessary to correlate two signals that do not belong to the same reflection plane with two line structures on both sides of the fault to solve the cross section. This is an extremely difficult task even for experts.

Therefore, in order to interpret earthquake documents, experts
It is necessary to read multiple pieces of information by comparing and examining a large amount of data displayed on one screen. This is generally a tedious task that takes a long time, and is likely to be an extremely subjective judgment depending on the ability of the person in charge and the quality of the display screen.

Computer imaging has made tremendous advances in recent years, allowing computers to be used not only as display and storage means, but also as image analysis means.

The seismic cross section can be thought of as an image 2D of two dimensions I(k l). Characters indicate time on the ordinate character! indicates the seismic trace number on the abscissa. ■ indicates the amplitude as a function of the value of the reflection coefficient at the reflective surface.

Assuming that the amplitude I is represented by gray steps, the detection of seismic horizons theoretically reduces to edge detection or contour factors. However, this method has a fundamental problem. That is, since seismic images have extremely large noise, they are not suitable for analysis using conventional image processing methods. The results obtained after using the conventional contour detection method, ie "extrema filling", are disappointing. Traditionally used in the processing of geophysical data, for example to determine the slope of a formation, this extremum marking method first marks the maximum amplitude value in the characteristic trace at time TO, and then marks the peak value at To in the adjacent trace. Find the maximum value at the closest time and process subsequent traces in the same way. This method is extremely sensitive to noise and requires a large amount of threshold data.

The present invention considers a number of characteristics peculiar to an image of an earthquake cross section, such as the ambient noise of the earthquake cross section or the unidirectionality of the information contained in the seismic cross section, and calculates the measured value using parameters along two axes. The measured values are organized along a line structure, an image is formed in which the line structure is preferentially distributed along one axis, and a zone having a line structure with similar characteristic values is determined based on this image. A method for analyzing two-dimensional images is provided. The characteristics of the method of the present invention are as follows: First, a line structure included in an image is drawn, and then each line structure is decomposed into a plurality of curved sections that touch each other at the ends, and each of the curved sections has a point between the curved section and the curved line. selected such that the maximum distance between the straight sections connecting the ends of the section does not exceed a given limit value, and then for each of said sections at least one step characterizing all the different measurements of the parameters relating to said section. The method includes the steps of performing the assignment of one characteristic quantity and finally displaying all sections having at least one characteristic quantity belonging to the same predetermined range.

According to one variant, the alignment
In order to display all alignments in which at least one of the parameters with respect to The parameter is assigned to each alignment consisting of a junction of sections corresponding to one characteristic quantity and touching each other at the ends.

The method of the invention is particularly suitable for the structural and stratigraphic analysis of seismic images or seismic sections where information (in this case images of seismic horizons) is preferentially distributed along the horizon; In particular, the entire seismic horizon of one cross-section can be automatically plotted and the boundaries of geological assemblages with approximate characteristic values can be defined. The characteristics of the method of the present invention used for seismic exploration are, on the one hand, that in order to construct the calendar wheel, each trace of the seismic section image is replaced by a binary function that can take only two values, and that the amplitude of the trace is correct; When the trace amplitude has a negative value, the first value is given, and when the amplitude of the trace has a negative value, the second value is given. Contouring) program, i.e. a program capable of defining the contours of zones corresponding to one or the other value on a binary plane, and then tracing each of the contours in a given direction of rotation, , a series of chains (c
on the other hand, if the contour has folds, before decomposing the contour into chains,
The characteristics of each of the folds are determined by two adjacent cusps and two cusps existing on the contour, and a simple contour with two cusps and a multiple of two isolated classes are determined. , and through this conversion, anomalies can be detected as seismic doublets existing in the contour.

Finally, for applications in seismic exploration, the characteristic parameters assigned to intervals and alignments include at least the slope, which corresponds to the angle formed by the straight line connecting the ends of the interval or alignment and the reference vertical axis; or the amplitude corresponding to the average of the optimal amplitudes along the reflective surface section with respect to the alignment, the amplitude variance corresponding to the variance of the set of optimal amplitudes along the reflective surface section, the length corresponding to the number of dots in the interval or alignment, and the approximate period is used.

According to a variant, the image obtained after "contouring" may be filtered, in order to remove from the image very small surface elements that do not extend into more than one trace, for example.

Although the method of the present invention gives good results regarding seismic images,
It may also be advantageously used for flaw detection in other types of images, such as VLSI (large scale integrated circuit) images, natural scenes such as satellite, 5POT, LΔNDSAT images, and industrial components.

The advantages of the method according to the invention, especially when used in seismic exploration, are:
Time and cost savings as well as reliable and sustainable seismic information can be extracted from seismic images. For example, when it comes to time reduction, two
The method of the present invention allows elucidation work that took months to be completed in a few days or hours.

Other advantages and features of the invention will become apparent from the following description based on an exemplary embodiment illustrated in the accompanying drawings.

The seismic cross section shown in Figure 1 was obtained by an exploration method using so-called "seismic reflection." The principle of seismic reflection was explained above. Since there are many documents regarding this method, it is recommended to refer to these documents regarding the formation of seismic sections. As shown in FIG. 1 and in detail in FIG. 6, a seismic section is generally constructed by a parallel arrangement of isolated records with variable amplitudes, called traces 1.

The most common analog representation of these traces is the 6th
As shown in the rightmost trace of the figure, the arcs of positive amplitude of the signal are shown in black and the arcs of negative amplitude are shown in white. The reference axis 2 with zero amplitude does not need to be illustrated. 1
A plurality of traces belonging to one seismic section are recorded in a plane with distance [X on the abscissa axis and time T or depth P on the ordinate axis.

The seismic cross-sectional image of FIG. 1 is composed of rows of pixels, each representing one trace of the cross-section. The amplitude of the trace is coated with multiple gray levels, with the highest amplitude corresponding to black and the lowest amplitude corresponding to white.

In seismic cross-sections, the set of coherent arcs that make up the seismic almanac can be discerned with the naked eye. The purpose of observing seismic cross-sections is to find arc coherence. Each of the arcs can be represented by extrema that are considered to be approximations of the arc. Extrema are localized representations of arc time and amplitude. To construct a chain that may represent a portion of an earthquake almanac wheel,
Combine extreme values based on a number of promises selected by experts.

FIG. 6 shows three chains 4, 5.6 of local maxima.

An extremum may also be defined as a point at which the sign of the slope of the signal's amplitude reverses. The first step in using the method of the invention for seismic exploration is to replace the trace with a step function, such as function 7 in FIG. The value of the trace is set to 1 in a time range in which the amplitude has a positive slope, and the value is set to 0 in a time range in which the amplitude has a negative slope. If the cross section is "binarized" for each trace in this way, a diagram similar to FIG. 2 will be obtained. Second
According to the figure, a white "solid part" consisting of a set of values 1 surrounded by values O and a black f-void J consisting of a set of values 0 surrounded by values 1 can be distinguished. In the particular example shown,
A solid part corresponds to a positive gradient, and a void part corresponds to a negative gradient. Tracking in the direction of increasing ordinate, each white-to-black transition point corresponds to a local maximum, and each black-to-white transition point corresponds to a local minimum.

It may be advantageous to apply a morphological filter to this image in order to remove homogeneous surfaces of very small dimensions that extend over, for example, one or two traces and only indicate the presence of noise.

A conventional image processing program that performs "edge tracking", also referred to as a "contour plotting" program, is then used on the binary image. Examples of such programs include Pa
vlidis or Rosenfeld's "contour construction" algorithm or the inventor of the present invention N.

“Coup” developed by KESKES
There is a method. This method is based on the theory of delimiting a range or "contour" of zones corresponding to the same value. A contour 8 drawn in this manner is shown in FIG.

For each closed contour drawn as described above, a specific number of points that do not intersect the edges of the cross section may be determined.

In the case of simple contours, at least two points are determined. These points are the turning points when tracing the contour in a given direction of rotation, for example the left-right or right-left I\ turning points 9, 10 in the illustrated embodiment (FIG. 78). matches. These are called cusps.

A specific rotation direction (clockwise or counterclockwise) from the same cusp
If the contour is traced in one rotation direction, a solid part can be drawn, and if it is traced in the opposite direction, a void part can be drawn. Whether the tracked boundary line corresponds to a maximum value or a minimum value of the amplitude can be clearly identified by the direction of rotation and the direction of movement (right→left or left→right).

Some contours, such as the contour shown schematically in FIG. 78, have more than two cusps. When the method of the invention is used for seismic exploration, such additional cusps are treated in a special procedure, since they indicate the presence of folds 13, called "duplexes", which are important in seismic reflection.

These double classes 14 are related to important phenomena from an elucidation point of view, such as the presence of carbonaceous rock, oil or gas formations, and require careful treatment.

In the method of the present invention, the characteristics of each fold existing in a binary cross section are determined by several parameters. These parameters include the distances dxl and dx2 between each of the two cusps belonging to the fold and the point closest to each cusp in the contour to which the fold belongs, that is, the point that intersects the vertical line passing through the cusp; the horizontal distance d between two vertical lines passing through two cusps belonging to the same fold (this value defines the size of the fold) and, in some cases, i.e. in a particular direction in one direction of rotation; When tracing the contour in , if there is another fold behind one fold, the distance D between these two folds (
(not shown).

At the stage of decomposing each contour into "chains", it is determined whether the fold is an image of the double class. The term "chain" refers to a set of local maxima or minima that exist in the portion between two cusps of a contour. If the contour under observation is that of an unfolded solid containing only two cusps (e.g. when using conventional signs and values), then this contour consists of two strands and an upper The chain below corresponds to the local maximum value, and the lower chain corresponds to the local minimum value.

The contour containing the fold before processing the double class consists of four chains such as chains 15, 16, 17, and 18 in FIG. 78.

By processing the double class, it is determined whether the cusp of the fold is the end of the chain. If it is determined that the fold is not an image of the double class, the chain is cut at the fold and in subsequent processing the two chains, such as chain 18.16 in Figure 78, are observed separately. When the fold is determined to be a double-class image,
Correct the chain's trajectory without cutting it at the fold. That is,
As shown in FIG. 7B, for each fold, three contour sections are determined that are exponentiated between the vertical lines passing through the two cusps defining the fold. An upper contour portion 19 belonging to the upper chain 18 and having the fastest time, a lower contour portion 21 belonging to the lower chain 16 and having the slowest time, and an intermediate contour portion existing between the two contour portions and overlapping as one complete chain. 20 is determined.

On the one hand, when a fold is present, the upper contour part 1
The main chain 22 is formed by the combination of the upper chain 18 to which 9 belongs and the part of the lower chain not including the lower contour part, and on the other hand two single chains, namely the intermediate contour part 20 and the lower contour part 2
1 exists.

Whether it is a dual class or not depends on the parameters dxl,
dx2. Judgment is made based on d and D. In the described embodiment, the characteristics of the fold are determined by four parameters, and at least one of the distances Ndxl or dx2 is determined by a threshold value set as a function of the frequency characteristics of the seismic section (e.g. if the earthquake is between 10 and 80 t (z) is less than approximately 16 ms), and the length d is a function of the adaptation distance, e.g. the distance between the dual class being processed and the next dual class.
When it is less than D, it is determined that the fold is a double-class image.

All upper and lower folds belonging to the same contour are sequentially processed starting from the smallest size. Recalculate the distance after each processing of the dual class.

Decompose all contours into chains and then decompose each class into intervals. The section is a portion of a chain such that no point on the section is separated by more than 6 lengths of a straight line connecting the ends of the section. h is a threshold set by experts and is on the order of a few milliseconds.

The reason for performing the above decomposition is that it is impossible to be certain that one chain belongs to one and only one seismic ephemeris wheel. This is because it is important to define the boundaries of the homogeneous part of .

polygon approximation method
In order to perform this process in a chain tracking direction, a processing algorithm such as Pavlidis' algorithm may be used, which gives results independent of the chain tracking direction.

In the method of the invention, in particular when the method is used for seismic surveys, selection and Advantageously, the display takes place automatically. Each section is assigned a reference numeral and stored in a storage device, and a number of parameters that determine the characteristics of the arc of the earthquake calendar wheel portion corresponding to the section are calculated and stored. These parameters include, for example, the length corresponding to the number of traces divided by the section, the local slope corresponding to the angle formed by the vertical axis of the section and the straight section connecting the ends of the section, and the area that is most easily detected in seismic sections. One of the parameters is the amplitude, which is calculated in the form of the average value of the optimal amplitude of the arc along the interval and its standard deviation, and the average approximate period along the interval and its standard deviation.

In two dimensions, just the presence of sections allows experts to display the seismic elements present in the image. According to the method of the invention, an expert can observe the influence and distribution of parameters such as amplitude, continuity, slope, frequency, etc. within minutes by calculation after reading the magnetic tape.

The entire chain may be displayed to obtain an image similar to that of FIG. 3, or a double impression seismic image may be formed to obtain an image similar to FIG. 4.

By decomposing into intervals and calculating the characteristic parameters of each interval, an expert can select a set of intervals with close characteristic values in the seismic image to display a homogeneous seismic set. The selection is made on the basis of distances based on calculation parameters, requiring threshold and limit data. FIG. 5 shows a section of FIG. 1 selected based on the positive slope of the FIG. 1 cross section. Some of the alignments shown in FIG. 5 have been obtained by juxtaposing various adjacent sections that satisfy the same criteria. These alignments constitute a homogeneous set of extreme values belonging to the same calendar wheel.

It is possible to recalculate characteristic parameters such as length, average amplitude, amplitude variance, apparent frequency or average approximate period for the alignment obtained as a result of the first selection.

The selection criteria may be made more complex and multiple parameters may be used simultaneously. In this way, an expert can determine a set of intervals (
In other words, a set of earthquake calendar wheel parts) can be selected.

Using color coding, the government offices of the examined geological media and their characteristic values can be more clearly identified. For example, an integrated display of amplitudes of multiple seismic chains and different classes defined by experts is possible. Each set of alignments belonging to a given grade is displayed in one color, and a chain may be composed of multiple alignments belonging to different grades.

Color also helps identify additional parameters that were used as selection criteria. For example, various alignments can be coded and displayed in different colors based on length. In this case, one particular color corresponds to each alignment.

In fact, in the first step of the method the contours are quite long. The contour of the cross section spans the entire width of the cross section, and at least
In the case of very clear calendar wheels with large amplitudes, the contour consists of a combination of two or three alignments selected by one or more criteria.

The continuity of the calendar wheel drawing is further improved by joining alignments that lie in extension of each other in the continuous calendar wheel configuration and whose ends are separated by, for example, only two or three traces.

Although the above detailed description uses the method of the invention for the analysis of seismic sections, the invention is not limited to this application. Those skilled in the art will be able to make numerous modifications within the scope of the invention depending on the desired application.

For example, the method of the invention may be used in fields other than geophysics, such as counting cells, simply detecting contours, studying skin tissue, etc.

If the information is distributed closer to the horizontal line than the vertical line, the information is processed column by column, and if the information is distributed closer to the vertical line than the horizontal line, the information is processed row by column. Process.

[Brief explanation of drawings]

Figure 1 is an explanatory diagram of the seismic cross-sectional image, Figure 2 is an explanatory diagram of the binary image in Figure 1, Figure 3 is an explanatory diagram of the image in Figure 2 after using the "contour drawing" program, and Figure 4 is an explanatory diagram of the image in Figure 2. is an explanatory diagram showing the superposition of the result of automatic extremum entry and the initial image in Figure 1, Figure 5 is a schematic diagram showing the selection of alignment regarding negative slope, and Figure 6 is a diagram for understanding the terminology used. 78 and 7B are illustrations of a hypothetical example of the doublet procedure. 1...Trace, 2...Reference axis, 4, 5
．． 6...Extreme value, 7...Step function, 13.
...Fold part, 14...Doublet, 15,
16, 17, 18... chain. 1;〈Engraving on the surface (change to circular stand) Shin IG-1 IGx2 FIG-7A X Year 1990↓8th

Claims (3)

[Claims] (1) An image is formed in which measured values are determined by parameters along two axes, the measured values are organized along a line structure, and the line structure is preferentially distributed along one axis; A two-dimensional image analysis method capable of determining a zone including a line structure with approximate characteristic values based on the image, the method comprising: drawing the line structure included in the image; decomposed into curved sections, each of said curved sections being selected such that the maximum distance between a point of said curved section and a straight section connecting both ends of said curved section does not exceed a given limit value; and performing an assignment of at least one characteristic quantity characterizing all the different measured values of the parameter with respect to the interval, and displaying all the intervals having at least one characteristic quantity belonging to the same predetermined range. Features a two-dimensional image analysis method. (2) In order to display all alignments in which at least one of the parameters related to the alignment belongs to the same region defined by the user, a parameter determining at least one characteristic quantity is calculated, and at least one parameter belonging to the same range is displayed. 2. A method as claimed in claim 1, characterized in that the parameter is assigned to each alignment consisting of a junction of sections corresponding to characteristic quantities and touching each other at their ends. (3) A seismic cross-sectional image is composed of multiple columns representing each trace of the cross-section, and the entire quasi-seismic layer of one cross-section is automatically drawn, and the boundaries of geological assemblages with approximate characteristic values are defined. 3. Use of the method according to claim 1 or 2 for carrying out a structural and stratigraphic analysis of a seismic cross-section, comprising: on the one hand, each trace of the seismic cross-section image being divided into two values in order to construct stratigraphic strata; When the amplitude of the trace has a positive slope, the first value is given, and when the amplitude of the trace has a negative value, the second value is given. , apply an edge tracking (contour drawing) program to the binary seismic cross-sectional image, that is, a program capable of defining the contour of a zone corresponding to one or the other value on the binary cross-section; When tracing a contour in the rotational direction, decomposing each of the contours into a series of chains formed by cusps, which are points on the contour corresponding to left-to-right or right-to-left turning points;
Consider the chain to be an image of a seismic horizon; on the other hand, if the contour has folds, before decomposing the contour into chains,
The characteristics of each of the folds are determined by two adjacent cusps and two cusps existing on the contour, and a simple contour with only two cusps and a multiple of two isolated chains are determined. and the characteristic parameters assigned to the interval and alignment are at least slope, amplitude, amplitude dispersion, length,
Use characterized by being approximately periodic.