The approach described herein is based on the fact that certain objects in the subsurface backscatter seismic waves, for example in the form of spherical waves. If the seismic waves are generated and recorded in a controlled manner, locations where these objects may be found may be determined in an exploratory method by means of travel time corrections and statistical evaluations.

FIGS. 1a and 1b show flowcharts of embodiments of a method for detecting one or more objects in the sea floor. The method comprises obtaining 110 a receiver signal. The receiver signal is based on a scattering of multiple acoustic signals at the one or more objects in the sea floor. The receiver signal is generated by a plurality of receivers. The method further comprises grouping 120 portions of the receiver signal to points of a detection grid. The detection grid represents a grid at the points of which the one or more objects are being localized. The method further comprises performing 130 a travel time correction of the portions of the receiver signal with respect to the points of the detection grid. The method further comprises combining 140 the travel time corrected portions of the receiver signal at the points of the detection grid. The method further comprises detecting 190 the one or more objects at the points of the detection grid based on the combination of the travel time corrected portions of the receiver signal. The detection of the one or more objects is based on the scattering of the multiple acoustic signals at the one or more objects.

FIG. 1c shows a block diagram of an embodiment of a corresponding apparatus 10 for detecting one or more objects in the sea floor. The apparatus 10 includes an interface 12 for obtaining a receiver signal. The apparatus 10 further comprises a processing module 14 which is coupled to the interface 12. The processing module is configured to perform the method of FIGS. 1a and/or 1b. For example, the processing module is configured to group the portions of the receiver signal to points of the detection grid. The processing module is further configured to perform the travel time correction of the portions of the receiver signal with respect to the points of the detection grid. The processing module is further configured to combine the travel time corrected portions of the receiver signal at the points of the detection grid. The processing module is further configured to detect the one or more objects at the points of the detection grid based on the combination of the travel time corrected portions of the receiver signal.

The following description refers to both the method of FIGS. 1a and/or 1b and to the corresponding apparatus 10 of FIG. 1c.

Embodiments of the present disclosure deal with a method, an apparatus as well as a computer program for detecting one or more objects in the sea floor. The term “sea floor” is not to be construed as limiting in the context of the present disclosure—embodiments are equally applicable to detecting objects in a riverbed or at the bottom of a lake or other body of water.

Accordingly, in the light of this application, the term ‘sea floor’ also includes the bottom of bodies of water in general, i.e., also ‘riverbed’, ‘bottom of the lake’ or ‘lakebed’.

For example, the system is here intended to detect objects underneath the sea floor or riverbed, such as objects located, for example, in the sediment of the sea floor. For example, the system may be designed to detect the one or more objects within 10 m (or within 15 m, within 20 m) beneath the sea floor. For example, a detection depth of the system may protrude at least 10 m (or at least 15 m, at least 20 m) into the sea floor. The one or more objects may be, for example, (large, isolated) pieces of rock, such as glacial erratics, which are located in the sea floor. In other embodiments, the one or more objects may be, for example, ammunition that has not exploded, so-called UneXploded Ordnance (UXO). These objects may pose a risk, for example, if structures such as wind turbines or drilling platforms are to be built in the sea floor on foundations that protrude into the sea floor. To detect these objects, embodiments, unlike other approaches, do not use reflections caused by a signal source on the objects, but use the scattering of an acoustic signal at the respective objects. To take advantage of these scattering effects, the wavelength of the acoustic signal used for detection is matched to the size of the objects to be detected. To obtain wide-angle scattering on which the detection may be based, the wavelength of the acoustic signal may be of the same order of magnitude as the size of the objects themselves.

The method comprises obtaining 110 the receiver signal. In embodiments, the receiver signal is generated by a plurality of receivers. The receivers of the plurality of receivers may be, for example, hydrophones, which are microphones that may be used underwater to record or hear underwater sound. The plurality of receivers may be configured, for example, to capture a wavefront formed by scattering (and optionally reflection) of the multiple acoustic signals at the one or more objects (and optionally at the sea floor as well) and to generate the receiver signal based on the captured wavefront. Thus, the plurality of receivers is configured to generate the receiver signal based at least on a scattering of the multiple acoustic signals at the one or more objects.

Receivers often capture both the scattering and the reflection of the acoustic signal. In these cases, the portions based on the scattering and the portions based on the reflection may be separated by the processing module. In other words, the receiver signal may comprise a first signal portion based on the scattering of the multiple acoustic signals at the one or more objects. The receiver signal may comprise a second signal portion based on a reflection of the multiple acoustic signals. The method may comprise suppressing 115 the second signal portion relative to the first signal portion. The subsequent processing steps, such as grouping the portions of the receiver signal, performing the travel time correction, combining the travel time corrected portions, and/or detecting the one or more objects may be based (exclusively or predominantly) on the first signal portion, for example. In other words, the method may comprise suppressing 115, such as attenuating the received power by at least 50%, in the receiver signal, the reflection of the multiple acoustic signals (such as at the sea floor). Good results are obtained with a methodology based on singular value decomposition (SVD; see Bansal and Imhof, 2016: “Diffraction enhancement in prestack seismic data”) after a travel time correction for reflections. After the eigenvalue decomposition, the travel time correction is calculated back so that the original tracks are restored without reflections. Other approaches to reflection suppression are conceivable and do not change the way forward. In other words, suppressing the second signal portion, or the reflections, may be based on an eigenvalue decomposition of a travel time corrected version of the receiver signal.

In some embodiments, it is advantageous for the exemplary solution path shown herein that the backscattered wave field is sampled sufficiently densely. Therefore, it is assumed in the following by way of example that the locations at which the seismic waves are artificially generated (shot locations) and recorded (receiver locations) are known very precisely in space. In addition, the time points of the generation of the waves may be synchronized with the start of each recording. For example, a wave field is sampled sufficiently densely if no aliasing occurs in space and time, i.e., if each wavelength is sampled at least twice in space and time.

In at least some embodiments, the wavelength of the acoustic signal is matched to an expected size of the one or more objects. For example, the expected size of the one or more objects may be a value specified by the target of the detection. If, for example, glacial erratics are to be found, a different wavelength may be used than in a case where ammunition is to be found or, for example, a sunken ship. Thus, the wavelength of the multiple acoustic (or seismic) signals may be in the same range as the expected size of the one or more objects themselves. The wavelength of the multiple acoustic signals may correspond to at least 10% (or at least 20%, at least 30%, at least 50%) of the expected size of the one or more objects. The wavelength of the acoustic signals may, for example, correspond to at most 1000% (or at most 800%, at most 500%) of the expected size of the one or more objects. The present system and method may be used in many cases to detect rather larger objects, such as glacial erratics or unexploded ordnance. For example, the wavelength of the multiple acoustic signals may be approximately at least 50 cm (or at least 80 cm, at least 100 cm, at least 150 cm). For example, the wavelengths of the multiple acoustic signals may be substantially the same, i.e., differ by less than 5% of the wavelength. Also the distance between adjacent receivers of the plurality of receivers may be made dependent on the wavelength, for example to avoid aliasing. For example, a distance between adjacent receivers of the plurality of receivers may be at most half the wavelength of the multiple acoustic signals.

In an exploratory method, it may then be checked at all positions reached by the artificially generated seismic wave whether an object with the properties described above is located at this location. Limiting factors are the attenuation of the seismic signal with increasing travel time and the beam pattern of the objects. In order to test all possible locations of point diffractors in the survey area, a grid configuration (such as the detection grid) may be created in which the calculations are performed on the acoustic data. To follow the common conventions, the individual grid points (the points of the detection grid) are referred to as CFPs.

FIG. 2 shows a schematic diagram of a common-mid-point (CMP) and common-fault-point (CFP) sorting principle. While CMPs 210 combine time series of receivers 212 whose midpoints 216 coincide between source 214 and receiver 212, CFPs 220, in contrast, combine all time series that include a secondary spherical wave originating from a point diffraction source 222 (reference number 224 shows the acoustic source, reference number 226 shows the receivers).

In the next step, the CFPs are systematically checked. In this regard, the CFPs correspond to the points of the detection grid. The subsequent steps are performed for each point of the detection grid. For example, at each of the CFPs, at least a portion of the same sequence of steps is repeated as shown in FIGS. 3a-3h and described in detail below. In other words, steps 120-190 may be performed for each point of the detection grid. For example, grouping 120 the portions of the receiver signal, performing the travel time correction, combining the travel time corrected portions, and detecting the one or more objects (as well as the steps described below as optional) may be performed (separately) for each point of the detection grid. This approach makes the solution path an exploratory method. If there is an object at the tested location, a coherent signal is generated by a travel time correction of the acoustic recordings. With the help of this corrected and thus coherent signal, the one or more objects are detected. If there is no object at the location, no coherent signal is generated by the travel time correction and therefore no detection takes place.

The individual processing steps are explained below by means of schematic illustrations. FIGS. 3a to 3h show an illustration of the processing steps by means of synthetic data. The source-receiver offset describes the distance from the source to the receiver, while the distance to CFP describes the distance from a receiver to a Common Fault Point; Common Fault Points (i.e., the points of the detection grid) are potential positions of a source of scattering.

In the first illustration of the series, FIG. 3a, the receiver signal is illustrated, which includes, besides the diffractions, also reflections. FIG. 3a shows a synthetic example of raw shot gather in which reflections may be seen as hyperbolic branches. FIG. 3a shows model artifacts 302, reflections caused by the sea floor 304, and another group of reflections and diffractions 306. Normally, reflections, such as the reflection of the sea floor marked as “sea floor” are most clearly visible in marine seismic datasets. In the raw data illustration of all recorded signals for a shot, reflections appear as hyperbolic branches. Diffractions are included in the data, but are difficult or impossible to see in this illustration because of their weak amplitude and the travel paths which are different from reflections. The x-axis of FIG. 3a and FIG. 3b shows the source-receiver offset (in m), while the TWT (Two-Way-Traveltime) is plotted in ms on the y-axis of FIGS. 3a to 3h.

FIG. 3b illustrates a relative signal amplification of the diffractions in the capturing of the acoustic signal (“shot gather”). This signal amplification may be achieved, for example, by suppressing 115 the reflections in the receiver signal. The aim of this step is to amplify the diffractions that are weak compared to the reflections. For this purpose, the reflections are attenuated in a targeted manner, for example. This is possible because the travel time curves of diffractions and reflections are different. In FIG. 3b, due to an attenuation of reflections, the noise and especially the diffractions 312, which give a chaotic impression when plotted against the source-receiver offset, are better differentiable. As shown, even after suppression of the reflections, diffractions are still not clearly recognizable in the illustration of all recordings for a shot. In the case of strong diffractions, this step may also be skipped. Again, model artifacts 314 may be seen in FIG. 3b.

Subsequently, the portions of the receiver signal are assigned to the points of the detection grid. In the following, the points of the detection grid are also referred to as CFPs (Common-Fault-Points). FIG. 3c illustrates a sorting of the existing seismic data (the portions of the receiver signal) into Common Fault Points (the grouping of the portions of the receiver signal to the points of the detection grid), as performed for example by Kanasewich and Phadke, 1988 in “Imaging discontinuities on seismic sections”. In the illustration of the signals against the distance of the receivers from a CFP (x-axis, also in FIG. 3d) in FIG. 3c, diffractions 322 may now be seen as hyperbolic branches. Model artifacts 324 may also be seen in FIG. 3c. FIG. 3c shows the diffractions in the CFP capturing (“CFP gather”).

For example, to enable a statistical evaluation and to allow a better illustration, all shots and receivers that would contain the diffraction of an object at that location are combined to one CFP. In other words, the receiver signal may include portions of multiple receivers and portions of multiple shots (acoustic signals). These portions are now assigned to the points of the grid.

There are at least three different possibilities of assigning the acoustic recordings to the CFPs. The sorting and/or illustration of the recordings in the CFP combination takes place based on the distance of the CFP from the relevant receiver of the recording. In other words, the portions of the receiver signal may be grouped to the points of the detection grid based on a distance of the points of the detection grid from the receivers of the plurality of receivers and from the at least one signal source of the multiple acoustic signals.

A first possibility is the so-called Real Aperture Processing. In this approach, each shot is processed individually, i.e., the portions of the receiver signal are grouped to the points of the detection grid separately for each acoustic signal. For this purpose, starting from the positions of the receivers and the source, the detection grid (which may be configured in a grid configuration) is spanned by CFPs up to a maximum distance from the source which depends on the properties of the source used (e.g., aperture angle). The tracks, i.e., the recordings of all receivers for one shot of the source, of the shot to be processed are assigned to each CFP/point of the detection grid.

Alternatively, Synthetic Aperture Processing may be used, for example, i.e., processing by means of a synthetic aperture. When using a synthetic aperture, a number of receivers in the direction of movement may be synthetically increased by emitting multiple acoustic signals. Consequently, in this approach, a fixed number of successive shots are combined. In other words, the portions of the receiver signal are grouped to the points of the detection grid for a predefined number of time points combined in the predefined time sequence. Starting from the positions of receivers and shooting points, a CFP grid (detection grid) is spanned. The tracks of the combined shots are assigned to each individual CFP.

Alternatively, the portions of the receiver signal may be assigned purely on the basis of their distance from the points of the detection grid, i.e., all tracks within a certain distance from a CFP are assigned to that CFP. In other words, the portions of the receiver signal may be grouped to the points of the detection grid for a predefined distance of the points of the detection grid from the receivers of the plurality of receivers and from the at least one signal source. The distance should be large enough to include most of the diffraction hyperboloid to be expected. If this distance is chosen too large, the signal quality will decrease. A usable distance may be estimated from the raw data and/or after signal amplification of the diffractions by measuring the size of the contained diffraction hyperboloids.

Subsequently, the travel time correction is performed for the assigned portions of the receiver signal with respect to the points of the detection grid, i.e., a positioning of the receivers and the signal source to the point of the detection grid is included in order to achieve that different distances of the receivers and of the receiver to the respective point are compensated in the combination of the portions of the receiver signal. FIG. 3d illustrates the travel time correction of the receiver signal. By means of an appropriate travel time correction, the diffractions become a coherent seismic event, as shown in FIG. 3d. The aim of the travel time correction is to correct the diffraction signal of an object at the location of the CFP such that it occurs at the same time in all tracks.

Here, reference number 332 shows the travel time corrected diffraction, reference number 334 shows a multiple of the diffraction, and reference number 336 shows model artifacts. FIG. 3d shows the travel time corrected diffractions in the CFP capturing (“CFP gather, Diffraction Move Out (DiffMO)”).

Different travel time correction approaches may be followed. For example, a travel time correction may be used at constant velocities, i.e., a constant seismic velocity may be assumed for the travel time correction. This allows for an improved second order travel time correction equation when a constant velocity may be assumed:

The travel time correction at constant velocities described herein is based on a second order travel time calculation for a point diffraction (e.g., Yilmaz 1991: “Seismic data processing”; Sheriff and Geldart 1995; Clearbout 2010). The correction depends on the spatial position of the source of diffraction D=(x_d, y_d, z_d), of the receiver R=(x_r, y_r, z_r), of the source S=(x_s, y_s, z_s) and the root mean square velocity v_rms:

TWT
d

=

(

x
d

-

x
s

)

2

+

(

y
d

-

y
s

)

2

+

(

z
d

-

z
s

)

2

v
rms

+

(

x
d

-

x
r

)

2

+

(

y
d

-

y
r

)

2

+

(

z
d

-

z
r

)

2

v
rms

Equation
⁢

5

Equation 5 may be simplified based on the assumption that the source and receiver are located at the sea surface. This allows both z_rand z_sto be set equal to zero. In addition, the positions of the source and receiver are reduced to their distance from a CFP

d

s
,
r

=

(

x
d

-

x

s
,
r

)

2

+

(

y
d

-

y

s
,
r

)

2

Equation
⁢

6

This results in the following simplified travel time equation:

TWT
d

=

d
s
2

v
rms
2

+

z
d
2

v
rms
2

+

d
r
2

v
rms
2

+

z
d
2

v
rms
2

Equation
⁢

7

In equation 7, the vertical travel time is contained in the following form:

TWT

d
,
0

=

2
⁢

z
d

v
rms

Equation
⁢

8

This vertical travel time is constant for a point diffractor in a CFP and thus a DiffMO correction may be derived from Equation 7:

TWT

d
,
0

(

TWT
d

,

d
s

,

d
r

)

=

(

d
s
2

+

d
r
2

)

2

v
rms
4

-

2
⁢

TWT
d
2

(

d
s
2

+

d
r
2

)

v
rms
2

+

TWT
d
4

TWT
d

Equation
⁢

9

As may be seen in FIG. 3d, this equation corrects a diffraction in a CFP to a horizontal and thus coherent event.

Alternatively, a Radon transform may be used for unknown velocities, for example. One of the most important variables for the previously described travel time correction is the assumed average seismic velocity. If it is not known, the travel time correction may alternatively be performed for a range of possible velocities. In other words, the travel time correction may be performed for a range of possible seismic velocities.

If a diffraction is included in the data of the processed CFP, the stacked signal will show a maximum at the location and with the most appropriate average velocity. For example, a seismic velocity from the range of possible seismic velocities may be selected for the travel time correction based on a magnitude of a local maximum in the corresponding combination of the travel time corrected portions of the receiver signal. By this approach, positions of scattering bodies and average velocities may be determined. This methodology represents a Radon transform as transformation takes place from the signal data space to a velocity data space.

Alternatively, a travel time correction may be used at known velocities. In most cases, it cannot be assumed that the seismic velocity v_rmsremains constant with increasing travel time, e.g., because of compacted sediments that have an increased seismic velocity. Rather, it may be assumed that this velocity is a function of the travel time: v_rms(TWT). Following the nomenclature described above, according to Guigné et al. (2014), a travel time curve for diffractions may also be described using the following equation:

TWT
d

=

0.5

√

TWT
2

(

d
s
2

)

+

0.5

√

TWT
2

(

d
r
2

)

Equation
⁢

10

The terms

T
⁢
W
⁢

T

2

(

d
s

2

)

⁢

and
/
or
⁢

T
⁢
W
⁢

T

2

(

d
r

2

)

separately describe the travel times from the source and from the receivers to the respective CFP to be processed. These travel time equations may be formulated as an equation depending on v_rms(TWT) (see for example Yilmaz, 1991). Consequently, seismic velocities matched to different material layers between the plurality of receivers and the points of the detection grid may be used for the travel time correction with respect to the points of the detection grid.

After travel time correction, the tracks within the CFPs (i.e., at each point of the detection grid separately) are summed (combined) and, for example, the number of tracks is divided. A single sum track is created which may correspond to the combination of the portions of the receiver signal, in which the amplitude is plotted against the travel time. If the travel time corrected tracks are now stacked, as shown in FIG. 3e, a clearly recognizable signal amplification results at the time of the coherent event (as stacked diffraction, reference number 342). Consequently, the detection of the one or more objects may already be performed on the combination of the portions of the receiver signal (i.e., without including further method steps), i.e., the one or more objects may be detected based on an amplitude of the combination of the travel time corrected portions of the receiver signal. The travel time (TWT) shown in FIG. 3eh corresponds to the vertical travel time, therefore a conversion to depth may be made if a velocity model exists. This step represents a first statistical evaluation and has, for example, the purpose of suppressing noise and amplifying the signal. Here, all signals that are not diffractions are called noise. The individual tracks may be additionally scaled to compensate for the energy loss due to the spatial propagation of the signal (Spherical Divergence Correction).

In other words, the method, as shown in FIG. 1b, may comprise adjusting 145 the combination of the travel time corrected portions of the receiver signal to achieve a signal amplification of signal portions based on the scattering of the multiple acoustic signals at more distant objects of the one or more objects (for example by means of using the spherical divergence correction). Reference number 344 shows a multiple of the diffraction, which may be caused, for example, by a reflection of the diffraction. In FIG. 3e, the amplitude of the stacked signal is plotted on the x-axis, from −4 to +4×10⁻⁹.

In order to improve the interpretability of the combination of the travel time corrected portions of the receiver signal, a so-called envelope of the combination of the travel time corrected portions of the receiver signal may be calculated. In other words, the method may, optionally, comprise calculating 150 an envelope of the amplitude of the combination of the travel time corrected portions of the receiver signal. The one or more objects may be detected based on the envelope. FIG. 3f illustrates the optional calculation of the envelope of the stacked signal. The optional calculation of the stack envelope, as shown in FIG. 3f, simplifies the illustration of the amplitude of the stack to a single maximum 352. An evaluation of the signal with the phase information is possible and may also include further information about the illustrated point diffractor. The multiple 354 of the diffraction is also visible in FIG. 3f. In FIG. 3f, the envelope is plotted on the x-axis, on a scale from 0 to approx. 4.5×10⁻⁹.

In addition, a coherence function may optionally be calculated 160 via the combination of the travel time corrected portions of the receiver signal, such as a so-called semblance. In FIG. 3g, the semblance/coherence (a similarity or coherence measure) is optionally calculated as a measure of coherence from FIG. 3d (the travel time corrected signals). The semblance, i.e., the similarity of the tracks within a CFP over a defined time window, is a measure of the coherence of the travel time corrected signals. In other words, the coherence function may be based on a similarity between temporally successive portions of the receiver signal. For example, the coherence function may be based on a semblance analysis. The use of other coefficients, e.g., correlation coefficients, is conceivable and does not change the general solution path. The one or more objects may be detected based on the coherence function. Reference number