EP1869640A1

EP1869640A1 - Method for hierarchical determination of coherent events in a seismic image

Info

Publication number: EP1869640A1
Application number: EP06755429A
Authority: EP
Inventors: Serge Beucher; Etienne Decenciere; Luc Sandjivy; Cedric Magneron; Thimothée FAUCON
Original assignee: Earth Resource Management Services ERMS
Current assignee: Earth Resource Management Services ERMS
Priority date: 2005-04-15
Filing date: 2006-04-18
Publication date: 2007-12-26
Also published as: US20080170756A1; US8385603B2; FR2884636B1; WO2006108971A1; FR2884636A1

Abstract

The invention concerns the determination of coherent events in a seismic image comprising the following operational phases: a phase of selecting a segmentation criterion based on a variable and at least one sliding window on said zone and its characteristics, a phase of hierarchical segmentation, for overlapping positions of the sliding window, including a segmentation of said zone into n regions and for each pixel located at least once by a segmentation boundary; a phase of assigning to the non-located pixels a value of indices (EI and EIC) corresponding to a numerical or alphanumeric characteristic non-value; a phase of determining coherent events of the image by thresholding the indices (EI and EIC), that is by selecting only located pixels of the zone corresponding to values lower or higher than a fixed threshold. The invention is applicable to imaging data obtained by seismic imaging physical methods: amplitude data or seismic attributes and likewise to medical imaging, sonar, non-destructive control of materials and the like.

Description

METHOD FOR THE HIERARCHICAL DETERMINATION OF COHERENT EVENTS IN A SEISMIC IMAGE

The present invention relates to a method for the hierarchical determination of coherent events in a seismic image.

It applies more particularly, but not exclusively, to the image data obtained by physical seismic imaging methods: amplitude data or seismic attributes, in the pre-stack and post-stack domains, the stack being a central operation of the Seismic processing that compresses seismic data (reducing the number of data) and acts as a powerful noise filter. It also applies to medical imaging data, sonar, non-destructive testing of materials, etc.

Mathematical Morphology, developed in the same way as Geostatistics by Professor Georges Matheron, is based on set-theoretic and topological notions. Its principle is to study the morphological characteristics (shape, size, orientation, ...) of the objects in an image. Mathematical Morphology provides the language and nonlinear tools appropriate for the recognition and processing of shapes in an image regardless of its dimension (1D, 2D, 3D ... nD).

More particularly, Mathematical Morphology provides tools for hierarchical segmentation of images. Indeed, it allows the segmentation of images into several regions according to one or more criteria such as, for example, the amplitude, the contrast, the amplitude gradient ... The boundaries between these regions defining segments that are not necessarily rectilinear which generally represent energetic and continuous forms of the image.

By definition, a coherent event will be called a "coherent event" in a 3D image according to criteria of continuity and energy. For a 2D image, a coherent event will correspond to a line that is not necessarily straight.

The usual methods for determining coherent events in seismic images are generally based on so-called propagation algorithms. Germs are placed on the image, that is to say anchor points at the level of coherent events that we want to determine. These germs are usually the result of a human interpretation. The propagation algorithms determine the completeness of each of the selected coherent events by looking in the image for paths of greater spatial correlation from the germs relating to each event. They "propagate" from point to point, pixel by pixel, the coherent events of the image. This type of algorithm can be unstable, especially in a noisy environment, and a "bad" path of greater spatial correlation is sometimes quickly used. In addition, this propagation approach is not optimal for real-time 3D volume interpretation of coherent events in seismic cubes.

A direct application of the invention concerns the quality control of a horizon point (coherent events corresponding to geological interfaces) on a seismic cube.

The pointing of coherent events called horizons on seismic cubes is done semi-automatically. An interpreter creates seeds that serve as anchors for a propagation algorithm to generate all or part of the "pointed" horizon. But, this type of algorithm can be unstable, especially in noisy environment. This instability results in offsets of the point with respect to the local extremum corresponding to the targeted seismic event, or sometimes more radical errors corresponding to phase jumps. These errors can have a very large impact on the value of amplitude or any other attribute extracted for example in the context of a "reservoir-oriented study". However, a reliable and rapid control of the quality of the pointing of these horizons proves essential. Indeed, a "reservoir-oriented study" is an imaging study of a reservoir and an estimation of the reserves it contains, and the economic stakes induced by such studies are considerable, the precision of the results they provide. must be optimal. Small mistakes can have big economic consequences.

On the other hand, visual examination is poorly suited to quickly identify spatial anomalies extending in the three dimensions of space.

Another possible application relates to the determination of coherent events on different seismic cubes of the same geographical area, these cubes being generated with different processing parameters.

In the solutions currently proposed, it is common in seismic processing and interpretation that two (or more) seismic cubes are defined on the same geographical area. Depending on the operational context, the analysis of the differences between the two cubes is then generally instructive. It can highlight related differences:

- the production of hydrocarbons over time (reservoir problem 4D, ie defined according to the three spatial dimensions and a temporal dimension);

- to different types of variable studied as in a 4C problematic where two types of seismic waves are propagated in the subsoil: P compression waves and S shear waves in order to produce two cubes, a cube in P amplitudes and a cube in S amplitudes;

- different treatment parameters. The characterization of these differences (localization, quantification) is never easy, the simple difference between the two seismic cubes being often insufficient or even sterile. Indeed, even small geographical offsets of the coherent events related to each of the two seismic cubes significantly decrease the value of the information of the difference cube.

In order to minimize these drawbacks, the invention proposes to provide a tool for assisting the determination of coherent events in a seismic image by associating with each pixel of the image a pair of EII indexes "Event Importance Index", ECI "Event Confidence Index" calculated by applying, on "overlapping windows", a hierarchical segmentation. By "overlapping window", it is meant that instead of realizing a segmentation over the entire image as is commonly done, successive segmentations are performed on parts of the image defined by a "moving window" that moves on the image in the manner of a scanner.

The computation of the pair of indices EII and ECI solves the problem of the real-time determination of the coherent events of a seismic image.

Advantageously, the determination of the coherent events of a seismic image is done by means of the method according to the invention which comprises the following operating phases:

a phase of choice: of a segmentation criterion according to a variable represented in the form of a seismic image such as amplitude, contrast, gradient, or an area of the image to be qualified, at least one sliding window on said zone and its characteristics, a hierarchical segmentation phase, for positions of the overlapping sliding window, comprising a segmentation of said zone into n regions and for each pixel identified at least once by a segmentation boundary: saving the number of times the pixel is identified by a segmentation boundary for each sliding window to which it belongs and the calculation of the ratio of this number by the maximum theoretical number of times that this same pixel could be located, this ratio corresponding to the index ECI; o the allocation for each segmentation boundary of a hierarchical level based on a calculation according to a specific criterion such as the surface of the boundary, so as to obtain the index EII;

an allocation phase to the non-marked pixels of an index value

El and EIC corresponding to a non-numeric or alphanumeric characteristic value.

a phase of determining the coherent events of the image by thresholding the indices EII and ECI, that is to say by selecting only pixels identified from the zone corresponding to values lower or higher than a fixed threshold.

The variable can be a seismic attribute such as amplitude, reflectivity, impedance or any other attribute calculated on a seismic cube.

This thresholding may depend in particular on the business application pursued.

The characteristics of the sliding window may include its dimensions, the recovery rate of a position i of the window with respect to a position ir, the recovery ratio being defined by 1 / r, where r is the modulus of the displacement vector in a main directions of the multiple seismic image of a grid step (definition an image is defined on a grid with as parameter the grid step in each of the main directions of the image (usually Euclidean axes)) in this direction.

The number n of segmentation regions of the zone may be dependent on the segmented zone.

The boundaries of the segmentation for each position of the sliding window define the pixels identified by the segmentation process and for which the indices ElI and ECI can be computed, the low EII and high ECI pixels defining the most coherent events of the segmentation process. 'picture.

Advantageously, the segmentation phase can be carried out:

- by the hierarchical segmentation algorithm of "Water Sharing Line"; it is a mathematical morphology algorithm that makes it possible to perform hierarchical segmentation of images;

on a previously filtered seismic image, that is to say an image transformed by a filter, for example a filter making it possible to improve the continuity of the image and to eliminate high frequency random noise;

The recovery of the positions of the sliding window can be defined by a recovery rate parameter expressed for example in pixels.

The value of this recovery rate parameter can be decreased to lighten the calculation times.

Of course, in this case, the quality of the results should not be degraded.

This approach, in accordance with the method according to the invention, may be implemented in the context of any operational context involving a seismic cube or seismic cubes of the same geographical area.

The pixels identified by the process of segmentation of the guide zone surrounding the pointed horizon and characterized by their EII and EIC indices may allow the development of horizon qualification attribute maps in order to enable quality control. of said horizon. Embodiments of the invention will be described below, by way of non-limiting examples, with reference to the accompanying drawings in which:

FIG. 1 is a perspective representation of a seismic cube and a horizon to be qualified;

FIGS. 2a and 2b are vertical sections of FIG. 1 representing respectively the horizon to be qualified and the horizon to be qualified surrounded by a guide zone;

Figure 3 is a representation of a section of the guide area having a sliding window;

Figure 4 is a representation of the displacement of the window of Figure 3;

Figures 5a, 5b and 5c are representations of a sliding window segmented respectively into two, three and thirty regions;

Figure 6 is a representation of the displacement of the window of Figure 3 showing the boundaries according to their hierarchy;

Figures 7a, 7b and 7c are representations of qualification attribute cards.

FIGS. 8a and 8b are representations on vertical sections of anomalies, respectively deviations from a local extremum and a phase jump of the horizon, as evidenced by the qualification attribute maps of FIGS. 7a, 7b and 7c.

The first example is illustrated in Figures 1 to 8b. This example concerns the quality control of a point of "horizons" (coherent events corresponding to geological interfaces) and in particular of a pointed horizon 1 on a seismic cube 2 (Figures 1 and 2a).

A horizon is very wide in space and represents a geological interface of sedimentation between geological layers, it generally results from an interpretation by a geophysicist or a geologist.

The determination of the indices EII and ECI, in accordance with the process according to the invention, contributes to defining the parts of horizons deemed to be anomalous which should be examined in order to guarantee a geological coherence of the pointed horizons.

The variable or seismic attribute used is the amplitude of the cube.

The first steps of the process include the choice:

a criterion of hierarchical segmentation of the variable such as the amplitude,

an area of the seismic cube comprising the pointed horizon to be qualified, that is to say on which a quality control must be carried out,

a sliding window 3 on said zone and its characteristics (dimensions, recovery rate, etc.) (FIG. 3),

The criterion of hierarchical segmentation chosen is the amplitude corresponding to the values of the variable itself, that is to say that one takes a dynamic criterion related to the values of the variable. Thus, for example, the larger the average value of the variable associated with a segmentation boundary, the more pixels associated with this boundary will be in the first hierarchy levels.

The zone of the seismic cube to be qualified is limited in space in a guide zone 4 of the seismic cube 2 comprising and surrounding the pointed horizon to be qualified 1 (FIG. 2b). This zone 4 serves as a mask for selecting the pixels of the seismic cube on which the following steps of the method according to the invention will be carried out. The width of the zone 4 is defined according to a vertical section (Figure 2b) of the seismic cube by a vertical parameter of deviation in pixels on either side of the pointed horizon 1.

It should be noted that the dimensions of the guide zone surrounding the horizon can be varied spatially.

The dimensions of the sliding window must be fixed.

For this, one can advantageously perform beforehand a geostatistical analysis to define objective criteria for choosing the dimensions of the sliding window. These criteria may include lengths of spatial correlations for example determined by the identification of the ranges of a variogram. The variogram is a statistical function that can be used to analyze spatial correlations within a spatial dataset, the observed ranges on the variogram providing information about the average dimensions of the structures (events) in an image.

If necessary, the value of the recovery rate parameter of the different positions of the sliding window 3 may be decreased to lighten the calculation times. Of course, this value will be chosen so as not to degrade the quality of the results.

In FIG. 4, said recovery ratio is chosen equal to one pixel and a fixed reference R is placed. The position 6 of the window is shifted one pixel to the right relative to the position 5 of the window and the position 7 of the window is shifted one pixel to the right with respect to the position 6 of the window. The pixels marked are represented by points P.

The segmentation criterion chosen, the dimensions of the sliding window and its fixed recovery ratio value and the guide zone 4 defined, all pixels of the seismic cube of the guide zone are segmented by the sliding window.

In this example, the hierarchical segmentation phase, for positions of the overlapping sliding window, takes the segmentation of the image into thirty regions and for the pixels identified at least once by a segmentation boundary:

saving the number of times the pixel is marked by a segmentation boundary for all the positions of the sliding window to which it belongs and calculating the ratio of this number by the maximum theoretical number of times that this same pixel could be spotted this ratio corresponding to the ECI index (FIG. 4);

the assignment for each segmentation boundary of a hierarchical level according to a scale B based on a calculation according to a given criterion such as the surface of the boundary, so as to obtain the index EII (FIG. different levels that can be materialized by colors.

It should be noted that the number of segmentation regions varies from one sliding window to another. Indeed, for example, if we put a sliding window on France, the segmentation will show 22 regions, it will not be the same for a slippery window falling on Germany.

More particularly, when the image is segmented into two regions 8, 9, a boundary 10 appears (Figure 5a). This border corresponds to the first hierarchy level and its value is set to one.

Then, the image is segmented into three regions 8, 11, 12, a border 13 appears (Figure 5b). This border corresponds to the second level of hierarchy and its value is fixed at two. Segmentation continues to reach thirty regions and the corresponding boundaries (Figures 5c and 6).

Then the process comprises the following steps: an allocation phase to the non-marked pixels of a value of indices E1 and EIC corresponding to a non-digital characteristic value;

determining the coherent events of the zone by thresholding the indices EII and ECI, the index EII having to be less than a threshold value and the index ECI having to be greater than a threshold value,

Since all the pixels identified by the segmentation process of the guide zone surrounding the pointed horizon are characterized by their indices EII and EIC, several qualification maps of the pointed horizon can be calculated in order to make it possible to carry out the control. of the horizon. In this area, a pixel is defined by three dimensions x, y and z.

A qualification attribute map is a representation in two dimensions x and y. An attribute defined by a pair of x and y coordinates may correspond to an average value per vertical (line oriented along the vertical axis z of depth of the seismic cube) for all the pixels located in the guide zone surrounding the horizon ( for example ECI or average EII by vertical) or else the selection of a pixel on the vertical, for example that of the strongest ECI or weaker EII, and a value associated with this pixel.

Figure 7a shows a map of the horizon to qualify. A color code represents the Z depth of the horizon.

Figure 7b shows a position qualification map) of pixels selected because belonging to the most consistent local event.

Figure 7c is a map of the qualification attribute of the horizon pointer to be qualified, this attribute being named LPE coherence attribute.

For each vertical of the guide zone defined by a pair of coordinates x and y was selected the pixel belonging to the most coherent local event defined as being the pixel of weaker EII and stronger ECI. The difference between the vertical position of the pointed horizon and the position vertical pixels retained allows to highlight anomalies of position of the pointed horizon.

These cards make it possible to highlight anomalies and to locate them in the plane XY. These anomalies can then be found, from the interpretation of the anomalies on the 2D maps of qualification, on the data cube, in section for example as illustrated in Figures 8a and 8b).

Moreover, it happens that in the particular case of the LPE coherence attribute, this attribute can also be represented in the cube.

These anomalies may consist, for a given pixel, in a deviation from the local extremum, that is to say a discrepancy between the most coherent event, ie the pixel detected and the horizon. which is greater than the maximum allowed deviation or less than a minimum allowed deviation.

In the example of FIG. 8a, the pointed horizon is represented by a black line 14 and the qualified marked pixels, that is to say the pixels of higher ECI and weaker EII, are represented by circles 15. The frame 16 contains a pointed horizon portion located higher than the pixels thus constituting.

These anomalies can also consist, in phase jumps of the horizon corresponding to pointing errors which represent the unauthorized passage from one horizon to another or from one event to another: On the example of the figure 8b, the pointed horizon is represented by a black line 17 and the qualified marked pixels, that is to say the pixels of higher ECI and lower EII are represented by a white line 18. The frame 19 contains a phase jump of the pointed horizon.

The second example concerns the determination of coherent events on different seismic cubes of the same geographical area, these cubes being generated with different processing parameters. Indeed, in the solutions currently proposed, it is common in seismic processing and interpretation that two or more seismic cubes are defined on the same geographical area. Depending on the operational context, the analysis of the differences between the two cubes is then generally instructive. It can highlight related differences:

- the production of hydrocarbons over time (reservoir problem 4D); - to natures of different studied variable (problematic 4C for example);

- different treatment parameters.

However, the characterization of these differences (localization, quantification) is never easy, the simple difference between the two seismic cubes proving themselves often insufficient, or even sterile. Indeed, even small geographical offsets of the coherent events related to each of the two seismic cubes significantly decrease the value of the information of the difference cube. Thus, the simple difference between the two seismic cubes does not make it possible to distinguish between differences related on the one hand to the geographical offsets of the coherent seismic events and on the other hand to variations in energy (in value) of these events.

However, the quantification of the impact of the processing parameters on the amplitude images of seismic cubes is essential in a reservoir-oriented oil study, this impact ultimately affecting the calculation of hydrocarbon reserves.

The determination of the indices EII and ECI on each of the two cubes of the same geographical area, in accordance with the method according to the invention makes it possible to overcome spatial spatial shifts of coherent events of the same geological origin and thus to allow the implementation of complementary quantitative tools for analyzing the differences between seismic cubes of the same geographical area. The process then comprises the following steps:

the determination, for each seismic cube, of the indices EII and ECI according to the method according to the invention developed in example 1 by choosing: the amplitude of the variable "seismic amplitude" as segmentation criterion, the dimensions of the sliding window so that, the more we want to retain coherent events of small size, the more the dimensions of the sliding window are restricted, o a threshold value important to perform a thresholding on the ECI index (selection of values High ECI) and the EII index

(selection of high EII values) in order to determine a number of significant coherent events on the two seismic cubes,

once the coherent events have been identified on each of the seismic cubes, a pairing of these events consisting in defining that a given geological event is located on each of the two cubes and corresponds to given seismic events on the cubes; thus paired, these events serve as a support for the calculation of characterization attributes of the differences in the seismic cubes.

The criterion of pairing may for example be a criterion of spatial proximity: two events "sufficiently" close, that is to say belonging to a common neighborhood are considered to correspond to the same geological event.

According to one variant, the results of segmentation of one of the seismic cubes can be used to constrain the segmentation of the other cube. In particular, the Water Division Line algorithm makes it possible to construct markers for the segmentation of a seismic cube, these markers being obtained from the segmentation results of the other cube. According to another variant, it is possible to reduce the value of the sliding window overlap ratio parameter to reduce the calculation times so that the quality of the results is not degraded.

The pairing of coherent events on each of the two images can be achieved by searching, with a certain vertical tolerance, pixels defining coherent events of one of the two seismic images closest to the pixels defining coherent events of the other. seismic image.

The selection of matched coherent events makes it possible to dispense with the effects of geographical shift of coherent events and consequently to better characterize the differences between seismic cubes by calculating, for example, the simple difference in values, the geographical shift of coherent events on each. two cubes or any other relevant attribute.

This approach, in accordance with the method according to the invention, may be implemented in the context of any operational context involving different seismic cubes but the same geographical area (4D, 4C, multi-3D, etc.).

Claims

claims

A method for determining coherent events of a seismic image, characterized in that it comprises the following operating phases:

a phase of choice: of a segmentation criterion according to a variable represented in the form of a seismic image such as the amplitude, the contrast, the gradient of an area of the image to be qualified, at least one sliding window on said zone and its characteristics (dimensions, recovery rate, etc.),

a hierarchical segmentation phase, for positions of the overlapping sliding window, comprising a segmentation of said zone into n regions and for each pixel identified at least once by a segmentation boundary: saving the number of times the pixel is identified by a segmentation boundary for each sliding window to which it belongs and the calculation of the ratio of this number by the maximum theoretical number of times that this same pixel could be located, this ratio corresponding to the index ECI; o the allocation for each segmentation boundary of a hierarchical level based on a calculation according to a specific criterion such as the surface of the boundary, so as to obtain the index EII; an allocation phase to the non-marked pixels of an index value

El and EIC corresponding to a non-numeric or alphanumeric characteristic value;

2. Method according to claim 1, characterized in that the variable is a seismic attribute such as amplitude, reflectivity, impedance

3. Method according to claim 1, characterized in that the characteristics of the sliding window include its dimensions, the rate of recovery of a position i of the window relative to a position ir, the recovery ratio being defined by 1 / r, r being the vector module moving in one of the main directions of the multiple seismic image of a grid pitch in that direction.

4. Method according to claim 1, characterized in that the number of n segmentation regions of the zone is dependent on the segmented zone.

5. Method according to claim 1, characterized in that the second phase is performed by the hierarchical segmentation algorithm of "Water Sharing Line".

6. Method according to claim 1, characterized in that the second phase is performed on a previously filtered seismic image that is to say image transformed by a filter.

7. The method of claim 6, characterized in that said filter improves the continuity of the image and eliminates high frequency random noise.

8. Method according to claim 1, characterized in that the recovery of the sliding windows is defined by a recovery rate parameter expressed for example in pixels.

9. Use of the method according to claim 1 for the quality control of a horizon point on a seismic cube.

10. Use of the method according to claim 1 for determining coherent events on different seismic cubes of the same geographical area, these cubes being generated with different processing parameters.