Data processing unit for predicting a position of a sea mine

Abstract

The invention relates to a data processing unit (20) for predicting a position (22) of a sea mine in a search area. The data processing unit (20) is designed to check, by means of a model (24) based on known positions (26) of sea mines in the search area, whether there are one or more further positions (22) in the search area at which there is a high probability that a sea mine is located.

Description

[0001] Data processing unit for predicting the position of a sea mine

[0002] Description

[0003] The invention relates to the detection of sea mines.

[0004] Mine detection is carried out by minehunting vessels equipped with one or more sensors optimized for minehunting. Depending on the detection range of the sensors and the size of the search area, the minehunting vessel will scan the search area multiple times, particularly in serpentine patterns. The sensors typically used include at least a sonar with multiple underwater sound transducers. However, the sensors are not infallible, so mines can be overlooked. In particular, it is difficult to detect mines embedded in the seabed. However, an undetected mine poses a significant threat to following ships attempting to pass through the area.

[0005] The object of the present invention is therefore to create an improved concept for detecting sea mines.

[0006] This problem is solved by the subject matter of the independent patent claims. Further advantageous embodiments are the subject matter of the dependent patent claims.

[0007] Embodiments show a data processing unit for predicting, i.e., forecasting, the position of a sea mine in a search area. The data processing unit is configured to use a model trained with machine learning, based on known positions of sea mines in the search area, to check whether there are further positions in the search area where a sea mine is highly likely to be located. The model is a computer-trained model, preferably an artificial intelligence model. In principle, any algorithm that can be trained using machine learning can be used to create the model. Examples of such algorithms are an artificial neural network, a hidden Markov model, a linear regression, a Bayes classifier, a support vector machine, etc.

[0008] Preferably, the position of the sea mine(s) can be predicted using artificial intelligence. The artificial intelligence comprises the model. A model describes a trained algorithm. Most common algorithms are suitable as algorithms, in particular artificial neural networks, preferably deep learning approaches such as artificial neural networks based on transformer technology and / or convolutional artificial neural networks. This means that the model is preferably implemented as an artificial neural network. The artificial neural network can, for example, be based on transformer technology and / or be or comprise a convolutional artificial neural network.

[0009] In particular, it has been shown that no special algorithms need to be used to create the model. It is sufficient to train algorithms offered for commercial use with appropriate data. Such algorithms are also referred to as COTS (Components-Off-The-Shelf). It has been shown that the structure, for example the number of neurons or the number of layers in artificial neural networks, is irrelevant for training the model as long as a minimum level of complexity is achieved. Models that also perform well in public benchmarks are particularly suitable. Examples of benchmarks are the "COCO dev-test" or the "COCO val2017". Examples include models based on the DINO (DETR with Improved deNoising anchOr box) algorithm or the YOLO (You Only Look Once) algorithm.

[0010] In “Tianhe Ren, Jianwei Yang, Shilong Liu, Ailing Zeng, Feng Li, Hao Zhang, Hongyang Li, Zhaoyang Zeng, Lei Zhang. A Strong and Reproducible Object Detector with Only Public Datasets. arXiv:2304.13027 [cs.CV], (https: / / doi.org / 10.48550 / arXiv.2304.13027)“ and „Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M. Ni, Heung-Yeung Shum. DINO: DETR with Improved DeNoising for Anchor Box End Detection. arXiv:2203.03605 [cs.CV], (https: / / doi.org / 10.48550 / arXiv.2203.03605)“ are Examples for the Models available.

[0011] These algorithms are particularly suitable for analyzing image data. This means that images, such as nautical charts, can be used for training, which show the positions of known, especially previously cleared, sea mines. Advantageously, the images are scaled to the same scale. This means that the same actual distance between two sea mines corresponds to the same distance between the two sea mines in the image.

[0012] Trained with a large number of such images, the model is then able to recognize patterns in the images. This is usually done using unsupervised learning. The trained model can then predict the locations of further mines once the first sea mine(s) have been found in a new minesweeping area.

[0013] However, it is also possible to train a model, preferably with other algorithms, for example on the raw data of the positions, such as the GPS data.

[0014] The applicant, as a provider of minehunting vessels and minehunting sensors, in particular minehunting sonars, has access to a large amount of data from sea mine clearance operations in various sea areas. Based on the mine detections per sea area, the model can be trained using machine learning, typically using three datasets: a training dataset, a test dataset, and a validation dataset. The sea mine detections in a sea area can be stored in one dataset. This means that the model can be trained using datasets of known positions (locations) of sea mines in search areas already cleared. As already described, the datasets can consist of or include images with the positions of the sea mines, in particular one image per cleared sea area.

[0015] Since the mines are not deployed randomly, but based on a deployment algorithm designed to block as large a sea area as possible with the least possible use of resources (mine deployment), the deployment is not random. Thus, the model can be trained on the resulting deployment patterns.

[0016] The idea is to improve mine hunting with the help of ever-advancing technology of computer-implemented modeling of problems, particularly through machine learning, for example artificial intelligence. Typically, the model will not take over the entire search. This is not possible on its own because the model needs some locations, i.e. known positions, of sea mines in order to be able to predict further locations. But the model can support current mine hunting, for example by checking whether all mines have been found or whether there are any gaps during or after the conventional search of the search area by a minehunter. Furthermore, the model can verify an ambiguous, i.e. inconclusive, detection of a mine by the sensors in order to determine whether it is actually a mine or another object or artifact.

[0017] Furthermore, a method for locating sea mines in a search area is disclosed. The method comprises scanning the search area using a mine detection sensor, in particular a sonar. Minehunters can be used for this purpose, as described above. Furthermore, the method comprises verifying a possible position of a sea mine, detected by the mine detection sensor, using the data processing unit. This means that the verification of the possible position of the sea mine is performed by entering the possible position into the data processing unit, which automatically verifies the possible position of the sea mine. In other words, the model receives the possible position and verifies it internally. This reduces the probability of false detections.

[0018] Alternatively, the possible location of the sea mine can be verified by comparing it with one or more locations in the search area provided by the data processing unit where a sea mine is highly likely to be located. This means that the model can output its potential detection locations, and the possible location is compared manually or using another computer program with all potential detection locations to verify the possible location. This reduces the probability of false detections.

[0019] Analogous to the data processing unit, a computer program is disclosed, comprising instructions which, when executed on a computer, cause the computer to check, by means of a model based on known positions of sea mines in the search area, whether there are further positions in the search area at which a sea mine is highly likely to be located.

[0020] Preferred embodiments of the present invention are explained below with reference to the accompanying drawings. In the drawings:

[0021] Fig. 1 : a schematic block diagram of a data processing unit for predicting a position of a sea mine in a search area.

[0022] Before exemplary embodiments of the present invention are explained in more detail below with reference to the drawings, it is pointed out that identical, functionally equivalent or equivalent elements, objects and / or structures in the different figures are provided with the same reference numerals, so that the description of these elements shown in different exemplary embodiments is interchangeable or can be applied to one another.

[0023] Fig. 1 shows a schematic block diagram of a data processing unit 20 for predicting a position 22 of a sea mine in a search area. Using a model 24 based on known positions 26 of sea mines in the search area, the data processing unit 20 can check whether there are further positions 22 in the search area where a sea mine is highly likely to be located.

[0024] Optionally, the data processing unit 20 can have an input interface for inputting a possible position 28 detected by the sensors, in particular the sonar, into the data processing unit. The possible position can be verified by the model 24, so that the verification result 30 is output as the output of the data processing unit instead of the position 22 of the sea mine.

[0025] Although some aspects have been described in connection with a device, it is understood that these aspects also represent a description of the corresponding method, so that a block or component of a device can also be understood as a corresponding method step or as a feature of a method step. Similarly, aspects described in connection with or as a method step also represent a description of a corresponding block, detail, or feature of a corresponding device.

[0026] The above-described embodiments are merely illustrative of the principles of the present invention. It is understood that modifications and variations of the arrangements and details described herein will be apparent to others skilled in the art. Therefore, it is intended that the invention be limited only by the scope of the following claims and not by the specific details presented in the description and explanation of the embodiments herein.

[0027] List of reference symbols:

[0028] 20 Data processing unit

[0029] 22 Position of a sea mine, determined by the model

[0030] 24 Model

[0031] 26 known positions of sea mines in the search area (previous locations found)

[0032] 28 possible, unclear position of a sea mine, determined by a sensor

[0033] 30 Verification result of the unclear position

Claims

Patent claims 1 . Data processing unit (20) for predicting a position (22) of a sea mine in a search area, wherein the data processing unit (20) is designed to check, by means of a model (24) generated using machine learning and based on known positions (26) of sea mines in the search area, whether there are one or more further positions (22) in the search area at which a sea mine is located with a high probability.

2. Data processing unit (20) according to claim 1, wherein the model (24) is trained by means of data sets of known positions (22) of sea mines in search areas that have already been cleared.

3. Data processing unit (20) according to claim 2, wherein the data sets comprise positions, in particular GPS data, of the sea mines already found.

4. Data processing unit (20) according to claim 2 or claim 3, wherein the data sets comprise images with the positions of the sea mines.

5. Data processing unit (20) according to one of the preceding claims, wherein the model is an artificial intelligence.

6. Procedure for detecting sea mines in a search area comprising the following steps: - scanning the search area with a mine detection sensor, in particular a sonar; - Verifying a possible position (28) of a sea mine detected by the mine detection sensor by means of the data processing unit (20) according to one of the preceding claims.

7. The method according to claim 6, wherein the verification of the possible position of the sea mine is carried out by entering the possible position (28) into the Data processing unit (20) and wherein the data processing unit (20) automatically verifies the possible position (28) of the sea mine.

8. The method according to claim 6 or 7, wherein the verification of the possible position (28) of the sea mine is carried out by comparing the possible position with one or more positions (22) in the search area provided by the data processing unit, at which a sea mine is highly likely to be located.

9. A computer program comprising instructions which, when executed on a computer, cause the computer to check, using a model (24) based on known positions (26) of sea mines in the search area, whether there are further positions (22) in the search area at which a sea mine is highly likely to be located.

Images