Underwater sonar image target detection method and device

Abstract

The invention provides an underwater sonar image target detection method and device, and is applied to the technical field of underwater sonar image processing. The method comprises the following steps: acquiring underwater sonar image data, and preprocessing the underwater sonar image data; inputting the preprocessed underwater sonar image data into a target detection model to determine detection target information; wherein the target detection model is a YOLOv10 optimization algorithm and comprises a backbone network, a neck network and a head network, the backbone network comprises a multi-scale adaptive attention fusion module, and the multi-scale adaptive attention fusion module comprises a plurality of parallel feature extraction branches; the convolution kernels of the plurality of parallel feature extraction branches are different in scale.

Description

Technical Field

[0001] This invention relates to the field of underwater sonar image processing technology, and in particular to a method and apparatus for underwater sonar image target detection. Background Technology

[0002] In scenarios such as underwater rescue, resource exploration, hazard target screening, and seabed modeling, underwater sonar image target detection technology is required to accurately identify underwater targets and provide key information for operational decisions. Therefore, underwater sonar image target detection has significant application value.

[0003] In existing technologies, the YOLOv10 algorithm has become the dominant network model for real-time end-to-end object detection due to its effective balance between computational cost and detection performance, and is often used for image object detection tasks.

[0004] However, there are obvious drawbacks when applying the YOLOv10 algorithm to underwater sonar image detection: First, the underwater environment is subject to seabed reverberation noise interference, and the sonar imaging resolution is poor with a low pixel ratio in the foreground target area. At the same time, YOLOv10 has a weak ability to extract fine features, making it difficult to identify small targets and targets with indistinct features. Second, complex backgrounds and targets have similar features in sonar images, and YOLOv10 is prone to misidentifying noise as targets, leading to false detection problems and failing to meet the accurate detection requirements in complex underwater environments. Summary of the Invention

[0005] This invention provides an underwater sonar image target detection method and apparatus to solve the problems of difficulty in detecting small targets in underwater sonar images, false positives and false negatives caused by complex background interference, and unclear target features. It can improve the target detection and recognition capability of sonar images in complex underwater environments and achieve accurate output of target location, category and confidence level.

[0006] This invention provides a method for underwater sonar image target detection, comprising: acquiring underwater sonar image data and preprocessing the underwater sonar image data; inputting the preprocessed underwater sonar image data into a target detection model to determine the target information; wherein the target detection model is a YOLOv10 optimized algorithm, including a backbone network, a neck network, and a head network, the backbone network including a multi-scale adaptive attention fusion module, the multi-scale adaptive attention fusion module including multiple parallel feature extraction branches, and the convolution kernels of the multiple parallel feature extraction branches having different scales.

[0007] This invention also provides an underwater sonar image target detection device, comprising the following modules: an acquisition module and a processing module; the acquisition module is used to acquire underwater sonar image data and preprocess the underwater sonar image data; the processing module is used to input the preprocessed underwater sonar image data into a target detection model to determine the target information; wherein, the target detection model is a YOLOv10 optimized algorithm, including a backbone network, a neck network, and a head network, the backbone network including a multi-scale adaptive attention fusion module, the multi-scale adaptive attention fusion module including multiple parallel feature extraction branches, the convolution kernels of the multiple parallel feature extraction branches having different scales.

[0008] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the underwater sonar image target detection method as described above.

[0009] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the underwater sonar image target detection method as described above.

[0010] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the underwater sonar image target detection method as described above.

[0011] The underwater sonar image target detection method and apparatus provided by this invention, because the backbone network of the target detection model includes a multi-scale adaptive attention fusion module with parallel feature extraction branches of convolutional kernels of multiple different scales, can extract sonar image features from multiple scale dimensions, expand the feature receptive field and enhance the ability to capture fine features. Therefore, it can effectively solve the defect of the traditional YOLOv10 algorithm in extracting fine features, thereby accurately identifying small underwater targets and targets with inconspicuous features. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating the underwater sonar image target detection method provided by the present invention; Figure 2 This is a schematic diagram of the underwater sonar image target detection device provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0016] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0017] To facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish the same or similar items with essentially the same function and effect. Those skilled in the art can understand that the terms "first" and "second" are not intended to limit the quantity or execution order.

[0018] This application describes some exemplary embodiments for illustrative purposes. It should be understood that this application may be implemented in other ways not specifically shown in the accompanying drawings.

[0019] like Figure 1 As shown, this application provides an underwater sonar image target detection method, which can be applied to an underwater sonar image target detection device. The underwater sonar image target detection method may include steps S101-S102: S101. The underwater sonar image target detection device acquires underwater sonar image data and preprocesses the underwater sonar image data.

[0020] Specifically, underwater sonar images are susceptible to seabed reverberation noise, water flow interference, and other factors, resulting in problems such as low contrast between the target and the background and blurred details. Underwater sonar image target detection devices can first collect sonar image data in different complex underwater environments using sonar equipment, and then perform preprocessing operations.

[0021] Optionally, the preprocessing of the underwater sonar image data includes: performing a sample enhancement operation on the underwater sonar image data, wherein the sample enhancement operation includes at least one of the following: target and background contrast enhancement, random rotation, random scaling, random flipping, and adding noise.

[0022] For example, for images where the grayscale difference between the target and the background is small due to the sonar imaging principle, the sharpness of the target edge can be improved by contrast enhancement. For situations where changes in the angle of the sonar equipment in actual operations may lead to different target imaging directions, random rotation and random flipping operations can be used to simulate the target shape under different shooting angles, making the model adaptable to more diverse target postures. For the problem of small targets having a low proportion in the image, the proportion of small target features can be increased by random scaling. At the same time, noise that conforms to the characteristics of the underwater environment can be added to enhance the model's ability to resist noise interference and reduce false detections in subsequent detections.

[0023] S102. The underwater sonar image target detection device inputs the preprocessed underwater sonar image data into the target detection model to determine the target information.

[0024] The target detection model mentioned above is an optimized YOLOv10 algorithm, which includes a backbone network, a neck network, and a head network. The backbone network includes a multi-scale adaptive attention fusion module, which includes multiple parallel feature extraction branches with different scales of convolution kernels.

[0025] Specifically, after the preprocessed underwater sonar image is input into the target detection model, the model can gradually extract features and output detection results through the collaborative work of the backbone, neck, and head networks. The target detection model is an optimized YOLOv10 algorithm, comprising a backbone network, a neck network, and a head network. The backbone network includes a multi-scale adaptive attention fusion module, which includes multiple parallel feature extraction branches with convolutional kernels of different scales.

[0026] For example, the multi-scale adaptive attention fusion module can be configured with three parallel branches, employing convolutional kernels of 3×3, 5×5, and 7×7 scales respectively. The 3×3 convolutional kernel is used to extract detailed texture features of the target, such as the edge contours of small obstacles, while the 5×5 and 7×7 convolutional kernels are used to capture broader contextual features, such as the spatial relationship between the target and surrounding seaweed and reefs. By fusing multi-scale features, the model can simultaneously take into account the details of small targets and the features of large scenes, effectively solving the problem of traditional YOLOv10's difficulty in recognizing small underwater targets and targets with indistinct features.

[0027] Optionally, the backbone network adopts an enhanced CSPNet structure; the neck network adopts an improved PANet structure, which is used to achieve multi-scale feature fusion; and the head network adopts a one-to-many and one-to-one dual-head design.

[0028] Specifically, the backbone network adopts an enhanced CSPNet structure, which reduces computational redundancy in the feature extraction process by optimizing the gradient flow path. For example, it splits the feature map during feature transfer and retains key information through shortcut connections. This reduces the computational load of the model while improving the integrity of feature transfer, ensuring that weak features of underwater targets are not lost.

[0029] The neck network adopts an improved PANet structure, which is used to achieve multi-scale feature fusion. Compared with the traditional PANet, it dynamically adjusts the fusion weights of features at different scales. For example, it assigns higher weights to low-level features corresponding to small targets, strengthens the proportion of small target features in the fusion process, and avoids small target features being covered by high-level large-scale features.

[0030] The head network adopts a dual-head design of one-to-many and one-to-one. During training, the one-to-many head generates multiple candidate prediction boxes to provide rich supervision signals for the model. During inference, only the one-to-one head is enabled to output a single optimal prediction box, without the need to perform non-maximum suppression operation. This improves the detection speed and avoids the target missed detection that may be caused by non-maximum suppression operation.

[0031] Optionally, the loss function of the target detection model is a dynamic scale loss function, which balances the penalty intensity of scale loss and position loss in the process of detecting targets of different scales by dynamically adjusting the influence coefficients of scale loss and position loss.

[0032] It should be noted that since the positional deviation of small targets has a greater impact on the detection results, the model's learning accuracy in locating small targets can be enhanced by increasing the positional loss weight. For large targets, the scale loss weight should be appropriately increased to ensure that the model accurately identifies the overall size of the target and avoids misclassifying large targets as multiple small targets. This reduces the loss deviation in the detection of targets of different scales and improves the overall detection accuracy of the model.

[0033] Optionally, the detection target information includes the location, category, and confidence level of the detection target.

[0034] Optionally, before preprocessing the underwater sonar image data, the underwater sonar image target detection device can annotate the targets in the enhanced underwater sonar image data to generate an annotation file; divide the annotated underwater sonar image data into a training set, a validation set, and a test set according to a preset ratio; initialize the target detection model and load pre-trained weights, and configure training parameters, which at least include learning rate, batch size, and number of training epochs; train the target detection model using the training set, monitor the model's performance indicators through the validation set during training, and adjust the model's hyperparameters according to the performance indicators; evaluate the trained target detection model using the test set. If the target detection model is overfitted, optimize the model by using regularization methods or increasing data samples; if the model is underfitted, adjust the model structure or increase the number of training epochs, repeating the training and evaluation steps until the model performance meets the requirements. Finally, export the trained model to a suitable deployment format, such as PT, ONNX, TensorRT, etc. In practical applications, underwater sonar image data is input into the deployed model for target detection.

[0035] In this embodiment, the backbone network of the target detection model includes a multi-scale adaptive attention fusion module with parallel feature extraction branches of convolutional kernels of different scales. This module can extract sonar image features from multiple scale dimensions, expand the feature receptive field, and enhance the ability to capture fine features. Therefore, it can effectively solve the defect of the traditional YOLOv10 algorithm in extracting fine features, thereby accurately identifying small underwater targets and targets with inconspicuous features.

[0036] The foregoing mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0037] It should be noted that the device in the embodiments of this application includes a virtual device and a physical device. The virtual device can be an underwater sonar image target detection device, and the physical device can include electronic devices, computer storage media, and computer program products.

[0038] The underwater sonar image target detection method provided in this application can be executed by an underwater sonar image target detection device or a control module for underwater sonar image target detection within that device. This application uses an underwater sonar image target detection device executing the underwater sonar image target detection method as an example to illustrate the underwater sonar image target detection device provided in this application.

[0039] It should be noted that the underwater sonar image target detection device can be divided into functional modules according to the above method examples. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. Optionally, the module division in the embodiments of this application is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0040] like Figure 2 As shown in the figure, this application embodiment provides an underwater sonar image target detection device 200. The underwater sonar image target detection device 200 includes: an acquisition module 201 and a processing module 202.

[0041] The acquisition module 201 can be used to acquire underwater sonar image data and preprocess the underwater sonar image data; the processing module 202 can be used to input the preprocessed underwater sonar image data into a target detection model to determine the target information; wherein, the target detection model is a YOLOv10 optimized algorithm, including a backbone network, a neck network and a head network, the backbone network includes a multi-scale adaptive attention fusion module, the multi-scale adaptive attention fusion module includes multiple parallel feature extraction branches, and the convolution kernels of the multiple parallel feature extraction branches have different scales.

[0042] Optionally, the loss function of the target detection model is a dynamic scale loss function, which balances the penalty intensity of scale loss and position loss in the process of detecting targets of different scales by dynamically adjusting the influence coefficients of scale loss and position loss.

[0043] Optionally, the backbone network adopts an enhanced CSPNet structure; the neck network adopts an improved PANet structure, which is used to achieve multi-scale feature fusion; and the head network adopts a dual-head design with one-to-many and one-to-one communication.

[0044] Optionally, the preprocessing of the underwater sonar image data includes: performing a sample enhancement operation on the underwater sonar image data, wherein the sample enhancement operation includes at least one of the following: target and background contrast enhancement, random rotation, random scaling, random flipping, and adding noise.

[0045] Optionally, before preprocessing the underwater sonar image data, the processing module 202 can be used to annotate the targets in the sample-enhanced underwater sonar image data to generate an annotation file; divide the annotated underwater sonar image data into a training set, a validation set, and a test set according to a preset ratio; initialize the target detection model and load pre-trained weights, configure training parameters, the training parameters including at least the learning rate, batch size, and number of training epochs; train the target detection model using the training set, monitor the model's performance indicators through the validation set during training, and adjust the model's hyperparameters according to the performance indicators; evaluate the trained target detection model using the test set, if the target detection model is overfitted, optimize the model by using regularization methods or increasing data samples, if the model is underfitted, adjust the model structure or increase the number of training epochs, and repeat the training and evaluation steps until the model performance meets the standards.

[0046] Optionally, the detection target information includes the location, category, and confidence level of the detection target.

[0047] In this embodiment, the backbone network of the target detection model includes a multi-scale adaptive attention fusion module with parallel feature extraction branches of convolutional kernels of different scales. This module can extract sonar image features from multiple scale dimensions, expand the feature receptive field, and enhance the ability to capture fine features. Therefore, it can effectively solve the defect of the traditional YOLOv10 algorithm in extracting fine features, thereby accurately identifying small underwater targets and targets with inconspicuous features.

[0048] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an underwater sonar image target detection method. The method includes: acquiring underwater sonar image data and preprocessing the underwater sonar image data; inputting the preprocessed underwater sonar image data into a target detection model to determine the target information; wherein the target detection model is a YOLOv10 optimized algorithm, including a backbone network, a neck network, and a head network. The backbone network includes a multi-scale adaptive attention fusion module, and the multi-scale adaptive attention fusion module includes multiple parallel feature extraction branches, the convolution kernels of the multiple parallel feature extraction branches having different scales.

[0049] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0050] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the underwater sonar image target detection method provided by the above methods. The method includes: acquiring underwater sonar image data and preprocessing the underwater sonar image data; inputting the preprocessed underwater sonar image data into a target detection model to determine the target information; wherein the target detection model is a YOLOv10 optimized algorithm, including a backbone network, a neck network, and a head network. The backbone network includes a multi-scale adaptive attention fusion module, and the multi-scale adaptive attention fusion module includes multiple parallel feature extraction branches, the convolution kernels of the multiple parallel feature extraction branches having different scales.

[0051] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the underwater sonar image target detection method provided by the above methods. The method includes: acquiring underwater sonar image data and preprocessing the underwater sonar image data; inputting the preprocessed underwater sonar image data into a target detection model to determine the target information; wherein the target detection model is a YOLOv10 optimized algorithm, including a backbone network, a neck network, and a head network. The backbone network includes a multi-scale adaptive attention fusion module, and the multi-scale adaptive attention fusion module includes multiple parallel feature extraction branches, the convolution kernels of the multiple parallel feature extraction branches having different scales.

[0052] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0053] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for underwater sonar image target detection, characterized in that, include: Acquire underwater sonar image data and preprocess the underwater sonar image data; The preprocessed underwater sonar image data is input into the target detection model to determine the target information. The target detection model is an optimized YOLOv10 algorithm, which includes a backbone network, a neck network, and a head network. The backbone network includes a multi-scale adaptive attention fusion module, which includes multiple parallel feature extraction branches with different scales of convolutional kernels.

2. The underwater sonar image target detection method according to claim 1, characterized in that, The loss function of the target detection model is a dynamic scale loss function. The dynamic scale loss function balances the penalty intensity of scale loss and position loss in the process of detecting targets of different scales by dynamically adjusting the influence coefficients of scale loss and position loss.

3. The underwater sonar image target detection method according to claim 1, characterized in that, The backbone network adopts an enhanced CSPNet structure; the neck network adopts an improved PANet structure, which is used to achieve multi-scale feature fusion; the head network adopts a dual-head design with one-to-many and one-to-one features.

4. The underwater sonar image target detection method according to claim 1, characterized in that, The preprocessing of the underwater sonar image data includes: The underwater sonar image data is subjected to sample enhancement operations, which include at least one of the following: target and background contrast enhancement, random rotation, random scaling, random flipping, and adding noise.

5. The underwater sonar image target detection method according to claim 4, characterized in that, Before preprocessing the underwater sonar image data, the method further includes: Targets in the enhanced underwater sonar image data are labeled to generate a labeling file; The labeled underwater sonar image data is divided into training set, validation set and test set according to a preset ratio; Initialize the object detection model and load pre-trained weights, and configure training parameters, which include at least the learning rate, batch size, and number of training rounds; The target detection model is trained using the training set. During the training process, the model's performance metrics are monitored using the validation set, and the model's hyperparameters are adjusted based on the performance metrics. The trained target detection model is evaluated using the test set. If the target detection model is overfitted, regularization or additional data samples are used to optimize the model. If the model is underfitted, the model structure is adjusted or the number of training rounds is increased. The training and evaluation steps are repeated until the model performance meets the target.

6. The underwater sonar image target detection method according to claim 1, characterized in that, The target information includes the target's location, category, and confidence level.

7. An underwater sonar image target detection device, characterized in that, include: Acquisition module and processing module; The acquisition module is used to acquire underwater sonar image data and preprocess the underwater sonar image data. The processing module is used to input the preprocessed underwater sonar image data into the target detection model to determine the target information. The target detection model is an optimized YOLOv10 algorithm, which includes a backbone network, a neck network, and a head network. The backbone network includes a multi-scale adaptive attention fusion module, which includes multiple parallel feature extraction branches with different scales of convolutional kernels.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the underwater sonar image target detection method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the underwater sonar image target detection method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the underwater sonar image target detection method as described in any one of claims 1 to 6.

Images