EP2989610A1

EP2989610A1 - Method for identifying or detecting an underwater structure, computer and watercraft

Info

Publication number: EP2989610A1
Application number: EP14725352.0A
Authority: EP
Inventors: Benjamin Lehmann; Anton LORENSON
Original assignee: Atlas Elektronik GmbH
Current assignee: Atlas Elektronik GmbH
Priority date: 2013-04-26
Filing date: 2014-04-01
Publication date: 2016-03-02
Also published as: WO2014173393A1; DE102013104306A1

Abstract

In order to identify or detect an underwater pipeline, for example, template-based methods are used according to the prior art, in which the total solution space with the individual synthetically obtained images is calculated in advance and an image obtained using sonar is then compared with each individual synthetic image of the total solution space. This is extremely intensive in terms of computation and/or main memory, with the result that these solutions cannot be used in AUVs (Autonomous Underwater Vehicles). The invention therefore relates to a method for identifying or detecting an underwater structure, wherein the method comprises the following steps of: determining an original image using a (Synthetic Aperture) Side Scan Sonar mounted on a watercraft, in particular an underwater vehicle, determining a watercraft position and/or an underwater structure position, with the result that there is additional information, determining a reference image space with associated reference images using the additional information and known geometry of the underwater structure, with the result that there is a reduced reference image space, and comparing the original image with the determined reference images of the reduced reference image space, with the result that the underwater structure is detected and/or identified.

Description

Method for identifying or detecting an underwater structure, computer and watercraft

[01] The invention relates to a method for identifying or detecting an underwater structure, to a computer which is set up such that the method can be carried out, and to a watercraft which has the computer and an imaging system, in particular a sonar.

[02] Object detection and object identification (ie ect classification) in sonar-generated images are still a challenge for automated image processing. The measurement-specific properties such as turbulent inhomogeneous environment, no apriori knowledge about the position of the object to be recorded and low signal -to-noise ratios make it difficult to detect and identify the objects and increase the false alarm rate.

To identify or detect, for example, an underwater pipeline template-based methods are used in the prior art, in which previously calculated the total solution space with the individual synthetic images and then compared a sonar image with each individual synthetic image of the total solution space. This is extremely compute and / or memory intensive, so these solutions in AUVs

(Autonomous Underwater Vehicles) can not be used. Real-time processing is absolutely impossible because a time interval between two sonar images is orders of magnitude lower than a calculation sweep time.

[04] The object of the invention is to improve the state of the art.

[05] The object is achieved by a method for identifying or detecting an underwater structure, the method comprising the following steps

- Determining an original image by means of a watercraft, in particular an underwater vehicle, in particular a equipped with sonar or other suitable imaging system underwater vehicle

Determining a watercraft position and / or an underwater structure so that additional information is available

Determining a reference image space with associated reference images using the additional information and known geometry of the underwater structure, so that there is a reduced reference image space, and

Comparing the original image with the determined reference images of the reference image space so that the underwater structure is identified or detected.

[06] By including additional information, a total solution space can be dramatically reduced, so that there is only one reference picture space. Thus, a real-time detection and identification method can be provided. In particular, the above method can be used in AUVs (Autonomous Operated Vehicles) or ROVs (Remotely Operated Vehicles), since in particular the calculation duration of the reference solution space with corresponding reference images of the time interval lies between the determination of two sonar recordings in the AUVs or ROVs.

[07] The following terminology is explained:

[08] "Identify" or "detect" includes recognizing and characterizing the underwater structure in an original image. Information about the distance to the underwater structure and / or the geographic position of the underwater structure can then be obtained. In addition, special objects such as underwater mines, whose geometry is known, can be found.

[09] The "underwater structure" may include all natural or technical structures underwater, in particular pipelines, foundations of off-shore wind turbines or underwater mines are of particular interest.

[10] The "determination of an original image" is carried out in particular by means of sonar, whereby it is possible to use bow, stern, down-side, side-scan, synthetic aperture and other suitable sonars Basically, passive and active sonars are included. [11] For example, in the case of a ship, the vessel's position may include the coordinates determined by means of satellite navigation, while in the case of an underwater vehicle this may be the height above ground, which is obtained, for example, by echosounding all other data relating to any coordinate system can be used in the determination The number and quality of the data obtained are of great importance, since they have a considerable influence in reducing the total reference image space, so that a reference image space can be determined which compared to the overall reference image space (FIG. clearly) is reduced.

Alternatively or additionally, a "determination of an underwater structure layer" can take place, For example, the underwater structure water sound can have strongly reflecting regions, so that in the case of a linear structure, such as a pipeline, the reference image space is reduced such that essentially the "thickness" of the Pipeline in the original image is relevant for a distance determination. Also can be attached to the underwater structure transmitter, which emit signals so that information regarding a position of the underwater structure can be determined and evaluated.

[13] The "Determining a reference image space with associated reference images" is done using the

Extra information. This means the more extensive the additional information regarding the position of the vessel and / or the underwater structure, the more the overall reference image space can be reduced to the reference image space. The fewer reference images in the reference image space, the fewer comparisons with the original image must be made, so that the computing time is significantly reduced.

[14] "Additional information" includes all information that makes the overall reference image space reducible, in particular metrological data such as the position / coordinates of the vessel, altitude above seabed and / or submarine location, etc. Parts of the additional information can also be estimated can be metrologically determined over ground and a distance to the underwater structure can be estimated.

[15] In the present case, the "geometry of the

For example, the modeled synthetic structure may be a computer-aided image of a pipeline or a mine.

Highlight) a shadow of the actual

Underwater structure additionally modeled, as a highlight-shade pair of underwater structure can also be found in the original image. For example, modeling can be done with modeling software such as Matlab / Simulink® or POV-RAY or OCTAVE or other suitable software.

By exploiting the additional information, a "reduced reference image space" can be provided, for example, a reference image space has fewer reference images than a total reference image space, so that fewer reference images are to be compared with the original image.

[17] In the case of "comparing" the reference images with the original image, suitable image processing methods are used, in particular correlating, or other suitable method for calculating the match, by which at the highest

Degree of agreement of the reference image with the original image the location of the largest matches is determined. On the basis of the reference image, further data such as, for example, distance to the underwater structure, location of the structure, geographical coordinates of the structure can be obtained, or a navigation to or a repair / maintenance / removal of the searched object can take place.

[18] In a further embodiment, the reference image space is determined during use of the watercraft by means of calculation by a computer, wherein in particular the computer is a component of the watercraft.

Thus, in particular, an AUV can be provided that can autonomously adapt adaptively to underwater conditions. Also, the default to a Underwater structure can be changed without the extensive data must be stored in a memory. Especially in the case of ROVs, the calculation can also take place on a ship or a platform, so that a corresponding adaptation by an operator on board the ship / platform can take place.

[20] A "mission" in AUVs or ROVs includes, in particular, underwater activities in which one

Underwater structure should be found.

[21] In "Calculation by a Calculator", the reference image space is determined by means of a suitable computer on the basis of the model parameters of the underwater structure and the additional information.

[22] In order to further accelerate the identification or detection, a total reference image space can be structured, in particular stored in a database, on a storage medium and, when the reference image space is determined, the reference image space extracted from the total reference image space and loaded into a main memory of the computer.

[23] Thus, a more effective mechanism than that used in the prior art template method is available. In the present case, due to the structure of the stored data, only the data which is necessary for the comparison of the reference images with the original image due to the additional parameters determined are loaded into a working memory. By the structured deposit in a storage medium can indeed Basically more data volumes (unit [byte]) are available, but the speed advantage in the selection occurs. The "disadvantage" of the larger data volumes is rather secondary, since storage media are much cheaper and have the capacity to offer more capacity than corresponding memory.

[24] The "total reference image space" corresponds to the complete solution space for all

Mission parameters to provide all reference images. For example, all elevations above ground and all associated distances of an AUV to the underwater structure are determined. In addition, other data such as image resolutions and / or model parameters, etc. may be included in addition to the reference images.

[25] Structured data storage can be understood to be either hierarchical and / or relational storage of the data.The data can be normalized or even redundantly redundant.In the present case, it is important that a selection of reference images and data is used based on the structure Comparison of the reference images with the original image are required.

[26] A "database" includes all databases with all database models, especially relational databases.

In particular, external storage media such as CD-ROMs, hard disks, flash memories, DVDs, floppy disks, USB sticks or the like are encompassed as "storage medium." As a rule, all storage facilities are included which are not included the working memory of the comparison of reference image and original image performing computer are. It should be noted, however, that certain data may also be stored in the main memory. This offers little changing data.

[28] The data stored accordingly on the storage medium can be "extracted" in accordance with additional information determined and loaded into the main memory so that the respectively relevant data is stored in the main memory so that a time-effective identification or detection can take place.

[29] In another embodiment, the determination of the original image is carried out by means of a sonar. This procedure is particularly effective for underwater structure, underwater original images are also visually detectable.

[30] "Sonar" is a method for locating objects under water by means of emitted sound pulses ". The word is an English acronym of sound navigation and ranging, which translates to sound navigation and distance determination and describes the function well.

[31] In order to reduce the overall reference image space effectively, the additional information includes a water altitude above seabed, a measurement distance to the underwater structure and / or an expected underwater structure of the underwater structure, which are to be determined in particular by measurement. [32] The "water elevation above the seabed" includes altitude determinable by sonar.

[33] The "measurement distance to the underwater structure" corresponds essentially to the length of a line of sight or its projection onto the seabed.

[34] "An underwater structure of the underwater structure" corresponds to the spatial arrangement of the structure, for example, the underwater structure includes the course of a pipeline on the seabed.

[35] In a further embodiment, the additional information is discretized. Thus, the amount of data can also be significantly reduced.

[36] "Discretize" divides continuous areas linearly or nonlinearly into subunits, so that the elevation above seabed can be discretized, whereby the structure with associated shadow and the resolution of the original image have an influence on the width of the elevations above ground.

[37] In order to speed up the computing speed or the comparison, the original image can be subdivided into partial areas, in particular overlapping partial areas.

[38] The subareas are also called Ranges. Dividing into ranges thus generally results in further reducing the overall reference image space. The overlap are chosen in particular so that in the Overlap area a highlight-shadow image of the respective underwater structure fits inside.

[39] In a further embodiment, the additional information is discretized or the original image is subdivided on the basis of the image resolution of the measurement method and / or an underwater structure with associated length of the highlight-shadow pair of the underwater structure. Thus, the method can be adapted very effectively to the additional information.

[40] In order to determine the additional information particularly efficiently, the additional information (s) can be determined metrologically.

[41] According to a further aspect of the invention, the object is achieved by a computer which is set up in such a way that a previously described method can be carried out. In particular, the computer can be located both in the AUV and on a ship, separated from the ROV.

[42] In a further aspect of the invention, the object is achieved by a watercraft, in particular an unmanned underwater vehicle (AUV, ROV), which has a previously described computer or which is set up such that a previously described method can be carried out.

[43] In the following, the invention will be explained in more detail by means of exemplary embodiments. Show it Figure 1 is a schematic representation of an under

Water-working ROV, which detects a cylindrical underwater e ect,

Figure 2 is a schematic representation analogous to

1, wherein the detectable or detectable for detecting sizes are given,

FIG. 3 a schematic representation of a modeled cylinder with additionally modeled shadow,

Figure 4 is a schematic representation of a split Sonarranges with Pictured overlap and

FIG. 5 shows a structured representation of a

File structure for a template-based method.

[44] An AUV 100 dives below the water surface 111 at a deployment height 121 above the seabed 113. The AUV has an echosounder system which emits echosound signals 131 in order to determine the deployment height 121 by measurement. In addition, the AUV 100 has a sonar system that detects sonar images. In addition, in a first alternative, a high-performance computer 101 is arranged in the AUV, which uses the measured data of the AUV and the known geometry of an object to be searched 104 to model a comparison of determined sonar images Performs reference images, as well as the required reference images modeled in real-time operation.

[45] In the second alternative, only one computer 101 and one data memory (e.g., flash memory) are arranged in the ROV 100. In this second alternative, a database with reference images is stored on the flash memory, the database having the memory structure as shown in FIG. 5.

Rebuilt is the AUV 100 or an ROV 100, which in contrast to AUV is connected via a supply cable 108 to a ship 102. The ship 102 has a high-performance computer 103, which performs a comparison of determined sonar images with modeled reference images based both on the measurement data of the ROV and the known geometry of the pipeline 104 to be searched.

[47] An object (e.g., a pipeline) 104 is attached to the seabed 113. In an alternative version, the object 104 may include additional signal amplification sound reflectors 135 which may be included

Underwater sound signal emitted by the Sonar of AUVs / ROVs 100, strongly reflect. By means of a plurality of these sound reflectors 135, the AUV 100 or ROV 100 can additionally determine an underwater course / a position of the object.

In another version, instead of the underwater sound reflectors underwater sounder 135 arranged on the object (eg pipeline), which emit a coded underwater sound signal 133. About this Underwater sound signal 133 may also be a

Underwater course or a position of the object can be determined. A direct "line of sight" 123 between AUV / ROV and Pipeline is referred to herein as Line of Sight (LOS).

[49] In AUV operation, identification is carried out in two variants. In variant a), the height 121 is determined by means of depth sounder. The geometry of the object is known.

[50] Based on the measured altitude 121 and the known geometry of the object 104, the object is modeled for different distances 123 and different object attitude angles. If additional information about the position of the object can be determined from the measurement, only the determined object position can be used for the modeling, which additionally significantly reduces the solution space. After modeling, there is a large number of modeled reference images. These are stored in the main memory of the computer 101.

[51] In advance, in parallel or subsequently, the determination of an original image takes place by means of an imaging system, for example the sonar. This original image is compared to all reference images by correlation or other suitable method for determining the degree of matching. The reference image with the highest degree of agreement is selected. The information underlying this image is stored, for example, for further navigation to the object or otherwise used. In the variants b), all simulation data were determined before the mission on a high-performance computer and stored in a flash memory as a database, the database having the structure according to FIG. 5. In turn, the height 121 is determined metrologically by means of echosounder and depending on the height 121, the data relevant for this height (reference images) are loaded into the main memory of the computer 101. A discrete image (pixel image) of the region to be searched is also created and a comparison is made as described.

[53] If the underwater vehicle is designed as an ROV, the computer-intensive work steps take place by means of the computer 103 on board the ship 102. The data exchange takes place e.g. via supply line 108 or other suitable supply channel.

[54] Mathematically, Fig. 1 can be described as follows to a good approximation (see Fig. 2), wherein the height h

(Reference numeral 121 in Fig. 1) is discriminated so as to have a height discretization Ah which is designed such that within Ah a shadow change of the acoustic or optical shadow cast by the object is below an image resolution of the sonar, e.g. can be determined by the following example calculation:

[55] d / (h-a) = l / a

[56] follows from it

[57] h = (a * d) / l + a [58] it applies

[59] Ah = hi - h ₂ = a * d [Al / (lil ₂ )]

[60] it follows

[61] Δ1 <= r from which it follows that Ah <= a * d (r / l _d ² ),

[62] where "l _d " is the shadow length of the pipeline 104 at the smallest pipeline distance (given by the measurement method), "d" is the smallest possible object distance

(given by the measurement method), "r" is the image resolution of the scene (in a sidescan sonar in the AUVs / ROVs orthogonal direction) and "a" is the height of the smallest searched object.

[63] To reduce the number of reference images to be compared, a range discretization (FIG. 4) is performed. The measuring range (sonar range) is subdivided into overlapping partial ranges. For each of these subareas, templates b) are generated as if the object (e.g., pipeline) 104 in question was in the middle of the area. The following steps take place:

[64] A: An offset of the 1st rank 451 from the AUV is set to the blanked range 465. (In the case of a sonar measurement, the blanked range 465 is a metrologically hidden area).

[65] B: An extension of the 1st rank 451 in the AUV / ROV direction of travel results from the - for a given AUV (flight) altitude - offset of the searched object, which would be necessary to extend the shadow 344 significantly (eg 10%).

[66] C: The offset of the beginning of the following rank 453 to the end of the preceding rank 451 is chosen such that it is greater than the longest highlight shadow structure 344 that generates an object at the beginning of this rank 453, at least by the image resolution would. In particular, a maximum object height from the searched object set is taken into account.

[67] D: The steps A to C are repeated until the maximum possible extent of the scene to be searched is covered.

[68] This can be done analogously for variant a) as well as for ROVs.

[69] A template file (see structogram Fig. 5) is structured as follows:

[70] 571 Number of discrete altitudes

[71] 573 For each altitude

[72] 575 altitude number

[73] 577 altitude value

[74] 579 Number of discrete rank

[75] 581 For every range

[76] 583 range number [77] 585 Range value (start, stop)

[78] 587 number of templates in the range

[79] 589 For each template (ect model picture)

[80] 591 template number

[81] 593 Template type (type of object)

[82] 595 Position of the object center point in the model image

[83] 597 template data: model image, resolution, assumed object position, etc.

[84] The procedure of the detection method is performed as follows. If there is a generated discrete image of the environment to be searched (e.g., video camera or sonar image), the model images (templates) are loaded into the main memory of the computer. In an alternative, it is assumed that all of the altitude levels 121 given at the time of the image generation AUV / ROV fit into the range memory of the computer 101. So will this range template data

(The 579 times 587) Number of templates loaded in the same memory and a comparison as in the paragraph

[51] described. The original image is divided into the ranges and each of these ranges is compared with corresponding model images / templates 597. In other alternatives, it is assumed that at the fly height 121 given at the time of the image generation AUV / ROV, the resulting range templates do not fit into the main memory of the computer 101. That's the picture the scanned scene is divided into rank. Subsequently, for each of these areas will be included model images / templates 597 in the

Memory of the computer 101 loaded, subjected to the comparison and unloaded again.

Reference sign list

100 AUV, ROV

101 computers

102 ship

103 high-performance computer

104 pipeline

108 supply cable

111 water surface

113 seabed

121 application height

123 line of sight

133 Underwater sound signal

135 sound reflector

304 modeled pipeline

344 modeled shadow structure

346 shadow mixing zone

451 1st rank offset from AUV

453 subsequent range

465 Blanked Range 571 Number of discrete altitudes

573 For each altitude

575 altitude number

577 altitude value

579 number of discrete rank

581 For every range

583 range number

585 Range value (start, stop)

587 number of templates in the range

589 For each template (object model image)

591 template number

593 Template type (type of object)

595 Location of the object center point in the model image

597 Template Data: Model image, resolution, assumed object position etc.

Claims

Claims:

A method of identifying or detecting an underwater structure, the method comprising the steps of:

Determining an original image by means of a watercraft, in particular an underwater vehicle,

- Determine a vessel location

- Determining a reference image space with associated

Reference pictures using the additional information and known geometry of the searched

Underwater structure, so a reduced

Reference picture space is present and

Comparing the original image with the determined reference images of the reference image space so that the underwater structure is detected and / or identified.

A method according to claim 1, characterized in that the determination of the reference image space during use of the watercraft by means of a calculation by a computer, wherein in particular the computer is a component of the watercraft.

Method according to one of the preceding claims, characterized in that a total reference image space is structured, in particular in a database, stored on a storage medium and extracted when determining the reference image space of the reference image space from the total reference image space and loaded into a main memory of the computer. Method according to one of the preceding claims, characterized in that the determination of the original image by means of a sonar.

Method according to one of the preceding claims, characterized in that the additional information is a Wasserfahrzeughöhe seabed, a measuring distance to the underwater structure and / or a

Underwater structure of the underwater structure includes.

Method according to one of the preceding claims, characterized in that the additional information is discretized.

Method according to one of the preceding claims, characterized in that a subdivision of the original image into subregions, in particular overlapping subregions, takes place.

A method according to claim 5, characterized in that the discretization of the additional information or the subdivision of the original image is based on an image resolution and / or a shadow length of the underwater structure.

Method according to one of the preceding claims, characterized in that the additional information is determined by measurement.

Method according to one of the preceding claims, characterized in that the metrological extraction of the additional information is facilitated by the attachment of additional reflectors / signal generator to the object to be searched.

Computer, which is set up such that a method according to one of the preceding claims is feasible. Watercraft, in particular unmanned

Underwater vehicle having a computer according to claim 10 or which is arranged such that a method according to any one of claims 1 to 9 is feasible.