Underwater image data processing method, system and electronic equipment

Abstract

The present invention discloses an underwater image data processing method, system and electronic device, and relates to the field of data processing technology. The method comprises: collecting underwater images and underwater environments in real time through a preset underwater acquisition device, and transmitting them to obtain the underwater images and the underwater environment; performing image processing and target detection on the underwater images in combination with the underwater environment based on a deep learning algorithm of a YOLO model; storing sub-images of different categories in the underwater images after detection in different storage paths, and creating an image index for each sub-image accordingly. The present invention performs image processing and target detection on underwater images through a deep learning algorithm based on a YOLO model, and can accurately classify targets in underwater images, thereby solving the problem in the prior art that the classification accuracy is not high when classifying targets in underwater images, resulting in low subsequent utilization value.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an underwater image data processing method, system and electronic equipment. Background Art

[0002] With the development of underwater detection technology, the collection and analysis of underwater data has become increasingly important. Existing underwater data collection methods mainly rely on manual operations, which are inefficient and costly. In addition, the underwater environment is complex and changeable, and traditional image processing technology is difficult to achieve efficient and accurate target detection in underwater environments.

[0003] Therefore, there is an urgent need for an underwater dataset collection system based on advanced deep learning models to improve the efficiency and accuracy of data collection. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide an underwater image data processing method, system and electronic device. The present invention performs image processing and target detection on underwater images through a deep learning algorithm based on the YOLO model, and can accurately classify targets in underwater images, thereby solving the problem in the prior art that the classification accuracy is not high when classifying targets in underwater images, resulting in low subsequent utilization value.

[0005] A first aspect of the present invention is to provide a method for processing underwater image data, the method comprising:

[0006] Collect underwater images and underwater environment in real time through a preset underwater acquisition device, and transmit them to obtain the underwater images and underwater environment;

[0007] Based on the deep learning algorithm of the YOLO model, the underwater image is processed and the target is detected in combination with the underwater environment;

[0008] The sub-images of the underwater image that are detected and determined to be of different categories are stored in different storage paths respectively, and an image index is created for each sub-image accordingly;

[0009] Among them, in the step of performing image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment, the step of performing target detection on the underwater image includes:

[0010] In combination with the underwater environment, the underwater image is imported into the convolution layer and the pooling layer of the YOLO model to extract a multi-scale feature map;

[0011] Importing the feature map into the fully connected layer of the YOLO model to generate the center coordinates and confidence of the bounding box in the feature map;

[0012] According to the center coordinates and confidence of the bounding box, the category probability of each bounding box in multiple categories is predicted respectively, and normalized;

[0013] Non-maximum suppression is performed on the predicted bounding box to filter out bounding boxes with overlap greater than a preset value and retain the bounding box with the highest confidence.

[0014] According to one aspect of the above technical solution, in combination with the underwater environment, the underwater image is imported into the convolution layer and the pooling layer of the YOLO model, and in the step of extracting the multi-scale feature map, the calculation expression is:

[0015] ;

[0016] In the formula, For the The pixel value of the feature map of the layer, is the vertical position offset of the convolution kernel, is the horizontal position offset of the convolution kernel, is the convolution kernel, is the bias, is the activation function.

[0017] According to one aspect of the above technical solution, in the step of importing the feature map into the fully connected layer of the YOLO model and generating the center coordinates and confidence of the bounding box in the feature map, the tuple of the center coordinates and the confidence is expressed as:

[0018] ;

[0019] In the formula, is the confidence level of the target existence, are the center coordinates of the bounding box, and are the width and height of the bounding box respectively.

[0020] According to one aspect of the above technical solution, in the step of predicting the category probability of each of the bounding boxes in a plurality of categories according to the center coordinates and the confidence of the bounding box, and performing normalization processing, the calculation expression is:

[0021] ;

[0022] In the formula, The target in the given bounding box belongs to The probability of the class, For the The score of the class.

[0023] According to one aspect of the above technical solution, non-maximum suppression is performed on the predicted bounding box to filter out bounding boxes with an overlap greater than a preset value, and in the step of retaining the bounding box with the largest confidence, the calculation expression is:

[0024] ;

[0025] In the formula, is the intersection-and-union ratio, and Two bounding boxes respectively.

[0026] According to one aspect of the above technical solution, in the step of performing image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment, the step of performing image processing on the underwater image includes:

[0027] The underwater image is subjected to denoising, color correction and format conversion.

[0028] According to one aspect of the above technical solution, the step of denoising the underwater image is implemented by using a non-local means algorithm, and the calculation expression is:

[0029] ;

[0030] In the formula, is the pixel value after processing, is the neighboring pixel value, After denoising The pixel value of is the pixel pitch, and are filtering parameters, is the normalization factor.

[0031] According to one aspect of the above technical solution, the step of color correction of the underwater image is implemented by using a white balance algorithm, and the calculation expression is:

[0032] ;

[0033] In the formula, is the corrected pixel value, is the channel gain coefficient, is the original pixel value.

[0034] A second aspect of the present invention is to provide an underwater image data processing system, which is applied to the method described in the above technical solution, and comprises:

[0035] A data acquisition module, used to collect underwater images and underwater environment in real time through a preset underwater acquisition device, and transmit them to obtain the underwater images and underwater environment;

[0036] A data processing module, used for performing image processing and target detection on the underwater image based on a deep learning algorithm of a YOLO model in combination with the underwater environment;

[0037] The classification storage module is used to store the sub-images of different categories detected in the underwater image in different storage paths respectively, and to create an image index for each sub-image accordingly.

[0038] The third aspect of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the method described in the above technical solution when executing the computer program.

[0039] Compared with the prior art, the underwater image data processing method, system and electronic device shown in the present invention have the following beneficial effects:

[0040] The present invention uses underwater acquisition equipment to collect underwater images and underwater environments in real time, and then transmits them. Then, the deep learning algorithm of the YOLO model is used to perform image processing and target detection on the underwater images according to the specific environment of the underwater environment. Specifically, the underwater image is introduced into the convolution layer and the pooling layer of the YOLO model, and a multi-scale feature map is extracted. Then, the feature map is introduced into the fully connected layer, and the center coordinates and confidence of the bounding box in the feature map are generated. Then, based on the center coordinates and the confidence, the category probability of each bounding box in multiple categories is predicted and normalized. Then, non-maximum suppression is performed to obtain the bounding box with the maximum confidence, so as to accurately predict the target category, thereby solving the problem of low classification accuracy when classifying targets in underwater images in the prior art, resulting in low subsequent utilization value. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0042] Figure 1 Schematic diagram of the process of underwater image data processing method in one embodiment of the present invention;

[0043] Figure 2 4 is a structural block diagram of an underwater image data processing system in one embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. Several embodiments of the present invention are shown in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0045] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0047] Embodiment 1

[0048] See also Figure 1 The first embodiment of the present invention provides an underwater image data processing method, the method comprising steps S10 to S30:

[0049] Step S10, collecting underwater images and underwater environment in real time through a preset underwater collection device, and transmitting them to obtain the underwater images and underwater environment.

[0050] In this embodiment, the underwater acquisition equipment includes a camera, etc. By increasing the sealing of the camera, the camera can shoot and record below the water surface, thereby acquiring underwater images and underwater environment, and then transmitting and sharing data through a pre-established data transmission channel.

[0051] Step S20, performing image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment.

[0052] Among them, in the step of performing image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment, the step of performing target detection on the underwater image includes:

[0053] In combination with the underwater environment, the underwater image is imported into the convolution layer and the pooling layer of the YOLO model to extract a multi-scale feature map;

[0054] Importing the feature map into the fully connected layer of the YOLO model to generate the center coordinates and confidence of the bounding box in the feature map;

[0055] According to the center coordinates and confidence of the bounding box, the category probability of each bounding box in multiple categories is predicted respectively, and normalized;

[0056] Non-maximum suppression is performed on the predicted bounding box to filter out bounding boxes with overlap greater than a preset value and retain the bounding box with the highest confidence.

[0057] Among them, in combination with the underwater environment, the underwater image is imported into the convolution layer and pooling layer of the YOLO model, and in the step of extracting the multi-scale feature map, the calculation expression is:

[0058] ;

[0059] In the formula, For the The pixel value of the feature map of the layer, is the vertical position offset of the convolution kernel, is the horizontal position offset of the convolution kernel, is the convolution kernel, is the bias, is the activation function.

[0060] Among them, in the step of importing the feature map into the fully connected layer of the YOLO model to generate the center coordinates and confidence of the bounding box in the feature map, the tuple of the center coordinates and the confidence is expressed as:

[0061] ;

[0062] In the formula, is the confidence level of the target existence, are the center coordinates of the bounding box, and are the width and height of the bounding box respectively.

[0063] Specifically, the above formula represents the output of the bounding box prediction, which can simultaneously predict the position (center point, width, height) of the bounding box and the confidence that the object is contained in the bounding box.

[0064] Among them, in the step of predicting the category probability of each of the bounding boxes in multiple categories according to the center coordinates and confidence of the bounding box, and performing normalization processing, the calculation expression is:

[0065] ;

[0066] In the formula, The target in the given bounding box belongs to The probability of the class, For the The score of the class.

[0067] Among them, in the step of performing non-maximum suppression on the predicted bounding box to filter out bounding boxes with an overlap greater than a preset value and retaining the bounding box with the largest confidence, the calculation expression is:

[0068] ;

[0069] In the formula, is the intersection-and-union ratio, and Two bounding boxes respectively.

[0070] In this embodiment, in the step of performing image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment, the step of performing image processing on the underwater image includes:

[0071] The underwater image is subjected to denoising, color correction and format conversion.

[0072] The step of denoising the underwater image is implemented by using a non-local means algorithm, and the calculation expression is:

[0073] ;

[0074] In the formula, is the pixel value after processing, is the neighboring pixel value, After denoising The pixel value of is the pixel pitch, and are filtering parameters, is the normalization factor.

[0075] The step of color correction of the underwater image is implemented by using a white balance algorithm, and the calculation expression is:

[0076] ;

[0077] In the formula, is the corrected pixel value, is the channel gain coefficient, is the original pixel value.

[0078] Step S30, storing the sub-images of different categories detected in the underwater image in different storage paths respectively, and creating an image index for each sub-image accordingly.

[0079] Next, the method shown in this embodiment is described in detail with reference to a specific example.

[0080] When a group of underwater images are received, the underwater images contain fish and seabed vegetation. First, in this embodiment, the image is processed by a denoising algorithm to eliminate image noise; then, the image is color corrected and the image tone is adjusted; the processed image is input into the YOLOv9-10 model, the feature map is extracted through the model, and the feature map is imported into the convolution layer and the fully connected layer to predict multiple bounding boxes and category probabilities. Finally, the optimal bounding box is screened out by the NMS algorithm, and the location information and category of the fish and seabed vegetation are successfully detected. After the above steps, the background data processing center can efficiently and accurately process underwater images and detect targets, and the output detection results include target location, target category and confidence, which provide important support for subsequent data analysis and application.

[0081] Furthermore, the storage and management module connected to the data processing center in the background is used to store, classify and manage the processed data for subsequent analysis and research. The storage and management module consists of a database management system, a high-speed storage device and data management software. The specific implementation steps and processing flow are as follows:

[0082] First, the processed data includes image data, target detection results, and sensor data. These data are transmitted to the storage and management module through the data transmission module. After the data arrives, the system formats it and converts the data into a unified storage format. Among them, image data uses JPEG or PNG format, detection results are stored in JSON or XML format, and sensor data is stored in CSV or SQL format.

[0083] The system then classifies and indexes the data. The classification process is based on the category labels in the target detection results, and the image data is classified and stored according to the detected target category. For example, detected fish images are stored in the "fish" directory, and detected seabed vegetation images are stored in the "vegetation" directory.

[0084] The indexing process is to create an index for each data item for quick retrieval. The index includes information such as timestamp, device number, data type, and data content. The index is stored in the database in the form of key-value pairs to facilitate quick data query.

[0085] The data storage adopts a distributed storage system, which distributes and stores data on multiple storage nodes to improve storage reliability and access speed. Each storage node adopts RAID technology to ensure data redundancy and security.

[0086] The data management software provides data retrieval, access and backup functions. Users can retrieve data through the query interface, which supports fast query based on indexes.

[0087] Among them, data backup adopts a strategy that combines regular backup and incremental backup to ensure data security. Regular backup is performed once a week, and all data is backed up to an external storage device. Incremental backup is performed once a day, and only the data added or modified on that day is backed up.

[0088] Next, still combining the specific case, assume that the system receives a set of image data, which contains images of fish and seabed vegetation. First, the system formats the image data and converts it into JPEG format. Then, according to the category labels in the detection results, the system stores the fish images in the "fish" directory and the seabed vegetation images in the "vegetation" directory. Next, the system creates an index for each image, including the timestamp, device number, data type, and detection result content, and stores the index in the database.

[0089] Subsequently, the system distributes and stores the image data on multiple storage nodes, and uses RAID technology on each node to ensure data redundancy and security. Users can retrieve data for a specific time period or a specific device through the query interface, and the system quickly locates and returns the corresponding data based on the index. Finally, the system can, for example, perform a regular backup once a week to back up all data to an external storage device, and perform an incremental backup once a day to only back up data that was added or modified that day.

[0090] Through the above steps, the storage and management module can efficiently and reliably store, classify and manage processed data, ensure the security and accessibility of data, and provide a solid data foundation for subsequent analysis and research.

[0091] In addition, on the user side, the user interface is used to provide data visualization and interactive functions, and users can view real-time data, test results and historical records through the interface. The specific implementation steps and processing flow are as follows:

[0092] First, the user interface is developed using a front-end framework (such as React or Vue) and a data visualization library (such as D3.js or Chart.js). The interface design includes multiple modules: real-time data module, detection result module, and history module. The interface communicates with the data processing center in real time through WebSocket, receives real-time data and displays it visually.

[0093] In the real-time data module, after the system receives the real-time data, it transmits it to the front-end interface through WebSocket. The front-end interface parses the data and displays it visually. The image data is drawn using HTML5 Canvas, and the sensor data is displayed using charts.

[0094] In the detection result module, the system displays the results of target detection, including target location, category, and confidence. The target detection results are marked on the image using a bounding box, and the category and confidence are displayed through labels.

[0095] In the history module, users can view historical data by selecting dates and devices. The system retrieves the corresponding historical data from the database and displays it, using date selectors and drop-down menus for interaction.

[0096] Next, we will continue to use specific examples. Assume that the system receives real-time image data and sensor data. The user interface receives data through WebSocket and displays it in real time. Image data is drawn through Canvas, and sensor data is drawn as a line chart through Chart.js. The detection result module displays the target detection results, marks the detected target bounding box on the image, and displays the category and confidence. The user selects a date and device to view the history, and the system retrieves the corresponding data from the database and displays it through charts and images. The interactive functions provided by the user interface enable users to easily view and analyze underwater data, providing support for scientific research and applications.

[0097] Compared with the prior art, the underwater image data processing system shown in this embodiment has the following beneficial effects:

[0098] In this embodiment, underwater images and underwater environments are collected in real time by underwater collection equipment, and then transmitted. Then, the underwater images are processed and target detected according to the specific environment of the underwater environment through the deep learning algorithm of the YOLO model. Specifically, the underwater image is imported into the convolution layer and the pooling layer of the YOLO model, and a multi-scale feature map is extracted. Then, the feature map is imported into the fully connected layer, and the center coordinates and confidence of the bounding box in the feature map are generated. Then, based on the center coordinates and the confidence, the category probability of each bounding box in multiple categories is predicted and normalized. Then, non-maximum suppression is performed to obtain the bounding box with the maximum confidence, so as to accurately predict the target category, thereby solving the problem of low classification accuracy when classifying targets in underwater images in the prior art, resulting in low subsequent utilization value.

[0099] Embodiment 2

[0100] See also Figure 2The second embodiment of the present invention provides an underwater image data processing system, which applies the method described in the first embodiment. The system includes: a data acquisition module 10, a data processing module 20 and a classification storage module 30.

[0101] The data acquisition module 10 is used to collect underwater images and underwater environments in real time through a preset underwater acquisition device, and transmit them to obtain the underwater images and underwater environments.

[0102] In this embodiment, the underwater acquisition equipment includes a camera, etc. By increasing the sealing of the camera, the camera can shoot and record below the water surface, thereby acquiring underwater images and underwater environment, and then transmitting and sharing data through a pre-established data transmission channel.

[0103] The data processing module 20 is used to perform image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment.

[0104] The data processing module 20 is specifically used for:

[0105] In combination with the underwater environment, the underwater image is imported into the convolution layer and the pooling layer of the YOLO model to extract a multi-scale feature map;

[0106] Importing the feature map into the fully connected layer of the YOLO model to generate the center coordinates and confidence of the bounding box in the feature map;

[0107] According to the center coordinates and confidence of the bounding box, the category probability of each bounding box in multiple categories is predicted respectively, and normalized;

[0108] Non-maximum suppression is performed on the predicted bounding box to filter out bounding boxes with overlap greater than a preset value and retain the bounding box with the highest confidence.

[0109] Among them, combined with the underwater environment, the underwater image is imported into the convolution layer and pooling layer of the YOLO model, and the calculation expression for extracting the multi-scale feature map is:

[0110] ;

[0111] In the formula, For the The pixel value of the feature map of the layer, is the vertical position offset of the convolution kernel, is the horizontal position offset of the convolution kernel, is the convolution kernel, is the bias, is the activation function.

[0112] The feature map is imported into the fully connected layer of the YOLO model, and the tuple of the center coordinates and confidence of the bounding box in the feature map is generated as follows:

[0113] ;

[0114] In the formula, is the confidence level of the target existence, are the center coordinates of the bounding box, and are the width and height of the bounding box respectively.

[0115] Specifically, the above formula represents the output of the bounding box prediction, which can simultaneously predict the position (center point, width, height) of the bounding box and the confidence that the object is contained in the bounding box.

[0116] Among them, according to the center coordinates and confidence of the bounding box, the category probability of each bounding box in multiple categories is predicted respectively, and the calculation expression for normalization is:

[0117] ;

[0118] In the formula, The target in the given bounding box belongs to The probability of the class, For the The score of the class.

[0119] Among them, non-maximum suppression is performed on the predicted bounding box to filter out bounding boxes with an overlap greater than a preset value, and the calculation expression for retaining the bounding box with the largest confidence is:

[0120] ;

[0121] In the formula, is the intersection-and-union ratio, and Two bounding boxes respectively.

[0122] In this embodiment, the data processing module 20 is also used for:

[0123] The underwater image is subjected to denoising, color correction and format conversion.

[0124] The step of denoising the underwater image is implemented by using a non-local means algorithm, and the calculation expression is:

[0125] ;

[0126] In the formula, is the pixel value after processing, is the neighboring pixel value, After denoising The pixel value of is the pixel pitch, and are filtering parameters, is the normalization factor.

[0127] The step of color correction of the underwater image is implemented by using a white balance algorithm, and the calculation expression is:

[0128] ;

[0129] In the formula, is the corrected pixel value, is the channel gain coefficient, is the original pixel value.

[0130] The classification storage module 30 is used to store the sub-images of different categories detected in the underwater image in different storage paths respectively, and to create an image index for each sub-image accordingly.

[0131] Compared with the prior art, the underwater image data processing system shown in this embodiment has the following beneficial effects:

[0132] In this embodiment, underwater images and underwater environments are collected in real time by underwater collection equipment, and then transmitted. Then, the underwater images are processed and target detected according to the specific environment of the underwater environment through the deep learning algorithm of the YOLO model. Specifically, the underwater image is imported into the convolution layer and the pooling layer of the YOLO model, and a multi-scale feature map is extracted. Then, the feature map is imported into the fully connected layer, and the center coordinates and confidence of the bounding box in the feature map are generated. Then, based on the center coordinates and the confidence, the category probability of each bounding box in multiple categories is predicted and normalized. Then, non-maximum suppression is performed to obtain the bounding box with the maximum confidence, so as to accurately predict the target category, thereby solving the problem of low classification accuracy when classifying targets in underwater images in the prior art, resulting in low subsequent utilization value.

[0133] Embodiment 2

[0134] A third embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the method described in the above embodiments when executing the computer program.

[0135] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0136] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for processing underwater image data, characterized in that: The method comprises: Collect underwater images and underwater environment in real time through a preset underwater acquisition device, and transmit them to obtain the underwater images and underwater environment; Based on the deep learning algorithm of the YOLO model, the underwater image is processed and the target is detected in combination with the underwater environment; The sub-images of the underwater image that are detected and determined to be of different categories are stored in different storage paths respectively, and an image index is created for each sub-image accordingly; Among them, in the step of performing image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment, the step of performing target detection on the underwater image includes: In combination with the underwater environment, the underwater image is imported into the convolution layer and the pooling layer of the YOLO model to extract a multi-scale feature map; Importing the feature map into the fully connected layer of the YOLO model to generate the center coordinates and confidence of the bounding box in the feature map; According to the center coordinates and confidence of the bounding box, the category probability of each bounding box in multiple categories is predicted respectively, and normalized; Performing non-maximum suppression on the predicted bounding boxes to filter out bounding boxes with an overlap greater than a preset value and retaining the bounding box with the highest confidence; Among them, in combination with the underwater environment, the underwater image is imported into the convolution layer and pooling layer of the YOLO model, and in the step of extracting the multi-scale feature map, the calculation expression is: ； In the formula, For the The pixel value of the feature map of the layer, is the vertical position offset of the convolution kernel, is the horizontal position offset of the convolution kernel, is the convolution kernel, is the bias, is the activation function.

2. The underwater image data processing method according to claim 1, characterized in that: In the step of importing the feature map into the fully connected layer of the YOLO model to generate the center coordinates and confidence of the bounding box in the feature map, the tuple of the center coordinates and the confidence is expressed as: ； In the formula, is the confidence level of the target existence, are the center coordinates of the bounding box, and are the width and height of the bounding box respectively.

3. The underwater image data processing method according to claim 1, characterized in that: According to the center coordinates and confidence of the bounding box, the category probability of each bounding box in multiple categories is predicted respectively, and in the step of normalizing, the calculation expression is: ； In the formula, The target in the given bounding box belongs to The probability of the class, For the The score of the class.

4. The underwater image data processing method according to claim 1, characterized in that: The predicted bounding box is subjected to non-maximum suppression to filter out bounding boxes with an overlap greater than a preset value, and in the step of retaining the bounding box with the largest confidence, the calculation expression is: ； In the formula, is the intersection-and-union ratio, and Two bounding boxes respectively.

5. The underwater image data processing method according to any one of claims 1 to 4, characterized in that: In the step of performing image processing and target detection on the underwater image based on the deep learning algorithm of the YOLO model in combination with the underwater environment, the step of performing image processing on the underwater image comprises: The underwater image is subjected to denoising, color correction and format conversion.

6. The underwater image data processing method according to claim 5, characterized in that: The step of denoising the underwater image is implemented using a non-local means algorithm, and the calculation expression is: ； In the formula, is the pixel value after processing, is the neighboring pixel value, After denoising The pixel value of is the pixel pitch, and are filtering parameters, is the normalization factor.

7. The underwater image data processing method according to claim 5, characterized in that: The step of color correction of the underwater image is implemented by using a white balance algorithm, and the calculation expression is: ； In the formula, is the corrected pixel value, is the channel gain coefficient, is the original pixel value.

8. An underwater image data processing system, characterized in that: The system is applied to the method described in any one of claims 1 to 7, and the system comprises: A data acquisition module, used to collect underwater images and underwater environment in real time through a preset underwater acquisition device, and transmit them to obtain the underwater images and underwater environment; A data processing module, used for performing image processing and target detection on the underwater image based on a deep learning algorithm of a YOLO model in combination with the underwater environment; The classification storage module is used to store the sub-images of different categories detected in the underwater image in different storage paths respectively, and to create an image index for each sub-image accordingly.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of being executed on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images