Artificial intelligence-based marine environment monitoring method and device, and storage medium

Abstract

The application discloses an ocean environment monitoring method and device based on artificial intelligence and a storage medium, wherein the satellite sea area image is subjected to image segmentation to determine a plurality of first red tide risk areas on the satellite sea area image; current multi-dimensional ocean data collected by a movable marine buoy is used to determine a buoy pixel point category of a pixel point corresponding to the movable marine buoy; a prompt vector is generated based on the buoy pixel point category of the pixel point corresponding to the movable marine buoy; the satellite sea area image is subjected to secondary image segmentation based on the prompt vector and the satellite sea area image to determine a plurality of second red tide risk areas on the satellite sea area image; a plurality of movable marine buoys are allocated to each second red tide risk area; each movable marine buoy corresponding to the second red tide risk area is controlled to move to the corresponding second red tide risk area; and the plurality of second red tide risk areas on the satellite sea area image are updated. The application can improve the accuracy of ocean environment monitoring.

Description

Technical Field

[0001] This application relates to the field of marine monitoring technology, specifically to a marine environmental monitoring method, device, and storage medium based on artificial intelligence. Background Technology

[0002] Red tides are a frequent marine ecological disaster in nearshore waters. Their outbreaks can lead to oxygen depletion in the sea, death of aquatic organisms, and even the production of toxins that endanger public health. Currently, red tide monitoring mainly relies on a combination of macroscopic early warning via satellite remote sensing and fixed-point monitoring by buoys. However, satellite remote sensing can only achieve macroscopic identification of red tide areas and cannot achieve pixel-level precise positioning. Buoys are mostly in an anchored state, making it impossible to conduct targeted detection based on the location of the red tide area, resulting in low accuracy of marine environmental monitoring. Summary of the Invention

[0003] This application provides an artificial intelligence-based marine environmental monitoring method, device, and storage medium, which can improve the accuracy of marine environmental monitoring.

[0004] Firstly, the marine environmental monitoring method based on artificial intelligence provided in this application includes:

[0005] Satellite images of the target sea area are acquired through a satellite remote sensing system;

[0006] The satellite image of the sea area is segmented to obtain the first prediction confidence level of each pixel in the satellite image of the sea area belonging to the red tide category;

[0007] Based on the first prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category, multiple first red tide risk areas are determined on the satellite sea area image;

[0008] Based on the current multidimensional ocean data collected by the movable ocean buoy, the buoy pixel category of the corresponding pixel of the movable ocean buoy is determined, wherein the buoy pixel category of the corresponding pixel of the movable ocean buoy is either red tide category or normal category;

[0009] A cue vector is generated based on the buoy pixel category corresponding to the movable ocean buoy;

[0010] Based on the cue vector and the satellite sea area image, a second image segmentation is performed on the satellite sea area image to obtain the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category, and multiple second red tide risk areas on the satellite sea area image are determined based on the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0011] Multiple movable marine buoys are allocated to each of the second red tide risk areas;

[0012] Control each of the movable marine buoys corresponding to the second red tide risk area to move to the corresponding second red tide risk area;

[0013] When the movable ocean buoy moves to the corresponding second red tide risk area, the multiple second red tide risk areas on the satellite ocean image are updated.

[0014] Optionally, determining multiple first red tide risk areas on the satellite sea area image based on the first prediction confidence score of each pixel on the satellite sea area image belonging to the red tide category includes:

[0015] Pixels on the satellite sea area image whose first prediction confidence is higher than a preset confidence threshold are identified as candidate pixels;

[0016] Based on the position of each candidate pixel in the satellite sea area image, multiple candidate pixels in the satellite sea area image are clustered to obtain multiple pixel clusters;

[0017] The target polygon region corresponding to the pixel cluster is determined as the first red tide risk region.

[0018] Optionally, the allocation of multiple movable marine buoys to each of the second red tide risk areas includes:

[0019] The area in the satellite sea area image where multiple second red tide risk areas have been removed is determined as a normal area;

[0020] The normal region is refined, and skeleton refinement lines of the normal region are generated on the satellite sea area image;

[0021] Determine the shortest planned path for the movable marine buoy to reach the centroid of the second red tide risk area along the skeleton refinement line;

[0022] The allocation matching degree between the movable marine buoy and the second red tide risk area is determined based on the shortest planning path;

[0023] Based on the allocation matching degree, the target number of the movable marine buoys are allocated to the second red tide risk area.

[0024] Optionally, the step of performing image segmentation on the satellite sea area image to obtain the first prediction confidence score of each pixel in the satellite sea area image belonging to the red tide category includes:

[0025] The system acquires multiple historical sea area data collected at different times in the target sea area by the satellite remote sensing system and the mobile ocean buoy. The historical sea area data includes historical satellite sea area images collected by the satellite remote sensing system and historical multidimensional ocean data collected by each mobile ocean buoy on the historical satellite sea area images. The historical multidimensional ocean data includes water temperature, salinity, nutrient concentration, ocean current velocity, ocean current direction, and chlorophyll a concentration.

[0026] By inputting multiple historical sea area images into a time series prediction model, the third prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category is obtained;

[0027] Based on the first image segmentation model, the satellite sea area image is segmented to obtain the fourth prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category;

[0028] The first prediction confidence is obtained by taking a weighted average of the third prediction confidence and the fourth prediction confidence.

[0029] Optionally, the step of performing image segmentation on the satellite sea area image based on the first image segmentation model to obtain a fourth prediction confidence score for each pixel in the satellite sea area image belonging to the red tide category includes:

[0030] Based on the channel attention submodule of the first image segmentation model, feature extraction is performed on the satellite sea area image to obtain a channel attention feature map;

[0031] Based on the spatial attention submodule of the first image segmentation model, feature extraction is performed on the satellite sea area image to obtain an enhanced red tide feature map;

[0032] The multi-scale feature extraction module based on the first image segmentation model performs multi-scale feature extraction and fusion on the enhanced red tide feature map to obtain a multi-scale red tide feature map.

[0033] Based on the decoding module of the first image segmentation model, the multi-scale red tide feature map is processed to obtain the fourth prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0034] Optionally, generating a cue vector based on the buoy pixel category corresponding to the movable ocean buoy includes:

[0035] The pixels corresponding to the movable ocean buoy and the buoy pixel category corresponding to the pixels are used as cue points and input into the cue encoder of the second image segmentation model to obtain the cue vector.

[0036] The step of performing secondary image segmentation on the satellite sea area image based on the cue vector and the satellite sea area image to obtain the second prediction confidence score of each pixel in the satellite sea area image belonging to the red tide category includes:

[0037] The satellite sea area image is input into the image encoder of the second image segmentation model to obtain the image feature vector;

[0038] The image feature vector and the cue vector are input into the mask decoder of the second image segmentation model to obtain the second prediction confidence that each pixel in the satellite sea area image belongs to the red tide category.

[0039] Secondly, the marine environmental monitoring device based on artificial intelligence provided in this application includes:

[0040] The acquisition module is used to acquire satellite images of the target sea area through a satellite remote sensing system;

[0041] The first segmentation module is used to perform image segmentation on the satellite sea area image to obtain the first prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category.

[0042] The first determining module is used to determine multiple first red tide risk areas on the satellite sea area image based on the first prediction confidence level of each pixel on the satellite sea area image belonging to the red tide category;

[0043] The second determining module is used to determine the buoy pixel category of the pixel corresponding to the movable ocean buoy based on the current multidimensional ocean data collected by the movable ocean buoy, wherein the buoy pixel category of the pixel corresponding to the movable ocean buoy is either a red tide category or a normal category;

[0044] The prompt generation module is used to generate a prompt vector based on the buoy pixel category of the corresponding pixel of the movable ocean buoy.

[0045] The second segmentation module is used to perform secondary image segmentation on the satellite sea area image based on the prompt vector and the satellite sea area image to obtain the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category, and to determine multiple second red tide risk areas on the satellite sea area image based on the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0046] The allocation module is used to allocate multiple movable marine buoys to each of the second red tide risk areas;

[0047] The control module is used to control each of the movable marine buoys corresponding to the second red tide risk area to move to the corresponding second red tide risk area;

[0048] An update module is used to update multiple second red tide risk areas on the satellite sea area image when the movable ocean buoy moves to the corresponding second red tide risk area.

[0049] Thirdly, the electronic device provided in this application includes a memory and a processor. The memory stores a computer program, and the processor is used to run the computer program in the memory to implement the steps in the artificial intelligence-based marine environmental monitoring method provided in this application.

[0050] Fourthly, the computer-readable storage medium provided in this application stores a computer program that is adapted to be loaded by a processor to implement the steps in the artificial intelligence-based marine environment monitoring method provided in this application.

[0051] Fifthly, the computer program product provided in this application includes a computer program that, when executed by a processor, implements the steps in the artificial intelligence-based marine environment monitoring method provided in this application.

[0052] In this application, compared to related technologies, satellite images of the target sea area are acquired through a satellite remote sensing system; image segmentation is performed on the satellite sea area images to obtain a first prediction confidence level for each pixel in the satellite sea area image belonging to the red tide category; multiple first red tide risk areas are determined on the satellite sea area images based on the first prediction confidence level for each pixel in the satellite sea area image belonging to the red tide category; the buoy pixel category of the pixels corresponding to the movable ocean buoy is determined based on the current multidimensional ocean data collected by the movable ocean buoy, wherein the buoy pixel category of the pixels corresponding to the movable ocean buoy is either the red tide category or the normal category; the buoy pixel category of the pixels corresponding to the movable ocean buoy is determined based on the current multidimensional ocean data collected by the movable ocean buoy. A cue vector is generated based on pixel category. Secondary image segmentation is performed on the satellite image based on the cue vector and the satellite imagery to obtain a second predicted confidence level for each pixel belonging to the red tide category. Multiple second red tide risk areas are determined on the satellite imagery based on this second predicted confidence level. Multiple movable ocean buoys are assigned to each second red tide risk area. The movable ocean buoys corresponding to each second red tide risk area are controlled to move to that area. When a movable ocean buoy moves to its corresponding second red tide risk area, the multiple second red tide risk areas on the satellite imagery are updated. This application, on the one hand, preliminarily identifies the first red tide risk area by initial image segmentation, and then determines the buoy pixel category of the corresponding pixels of the movable ocean buoy based on data collected by the movable ocean buoy. Using the buoy pixel category as a clue, a second image segmentation is performed, thereby obtaining a second red tide risk area with higher accuracy. On the other hand, for the second red tide risk area with higher accuracy, the corresponding movable ocean buoy is assigned to move to the second red tide risk area for re-detection, thereby dynamically updating the second red tide risk area and improving the accuracy of marine environmental monitoring. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a schematic diagram of a marine environmental monitoring system based on artificial intelligence provided in an embodiment of this application;

[0055] Figure 2 This is a schematic flowchart of an embodiment of the marine environment monitoring method based on artificial intelligence provided in this application.

[0056] Figure 3This is a schematic diagram of multiple second red tide risk areas and skeleton refinement lines on a satellite sea area image in one embodiment of the marine environment monitoring method based on artificial intelligence provided in this application;

[0057] Figure 4 This is a schematic diagram of the structure of the artificial intelligence-based marine environmental monitoring device provided in the embodiments of this application;

[0058] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0059] It should be noted that the principles of this application are illustrated by example in a suitable computing environment. The following description is based on the specific embodiments of this application that are illustrated, and should not be regarded as limiting other specific embodiments not detailed herein.

[0060] In the following description of this application, "some embodiments" are referred to, which describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same subset or different subset of all possible embodiments, and may be combined with each other without conflict.

[0061] In the following description of this application, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0063] To improve the accuracy of marine environmental monitoring, embodiments of this application provide an artificial intelligence-based marine environmental monitoring method, an artificial intelligence-based marine environmental monitoring device, an electronic device, a computer-readable storage medium, and a computer program product. The artificial intelligence-based marine environmental monitoring method can be executed by the artificial intelligence-based marine environmental monitoring device, or by an electronic device integrating the artificial intelligence-based marine environmental monitoring device.

[0064] The technical solutions of the embodiments 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, and 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.

[0065] This application also provides an artificial intelligence-based marine environmental monitoring system, which includes electronic equipment. The electronic equipment integrates the artificial intelligence-based marine environmental monitoring device provided in this application.

[0066] To better understand the artificial intelligence-based marine environment monitoring method, device, electronic device, and storage medium provided in the embodiments of this application, the application environment applicable to the embodiments of this application will be described below.

[0067] Please see Figure 1 , Figure 1 This diagram illustrates an application environment of an artificial intelligence-based marine environment monitoring method according to an embodiment of this application. As one implementation, the artificial intelligence-based marine environment monitoring method provided in this embodiment can be applied to electronic devices. These electronic devices can be, for example... Figure 1 The server 110 shown can be connected to the terminal device 120 via a network. The network serves as a medium for providing a communication link between the server 110 and the terminal device 120. The network can include various connection types, such as wired communication links, wireless communication links, etc., and this embodiment does not limit this. Optionally, in other embodiments, the electronic device can also be a smartphone, laptop, or other terminal device.

[0068] It should be understood that Figure 1 The server 110, network, and terminal device 120 shown are merely illustrative. Depending on the implementation requirements, any number of servers, networks, and terminal devices can be included. For example, server 110 can be a physical server, a cloud server, or a server cluster composed of multiple servers, etc., and terminal device 120 can be a mobile phone, tablet computer, desktop computer, laptop computer, smartwatch, smart glasses, in-vehicle terminal, smart home terminal, cloud computer, etc. It is understood that embodiments of this application can also allow multiple terminal devices 120 to access server 110 simultaneously.

[0069] In addition, the AI-based marine environmental monitoring system may also include a memory for storing raw data, intermediate data, and result data during the AI-based marine environmental monitoring process.

[0070] In this embodiment of the application, the storage device can be a cloud storage device. Cloud storage is a new concept that is extended and developed from the concept of cloud computing. A distributed cloud storage system (hereinafter referred to as a storage system) refers to a storage system that uses cluster applications, grid technology and distributed storage file system functions to bring together a large number of storage devices of various types in the network (storage devices are also called storage nodes) through application software or application interfaces to work together to provide data storage and business access functions to the outside world.

[0071] Currently, the storage method of storage systems is as follows: Logical volumes are created. During the creation of a logical volume, physical storage space is allocated to each logical volume. This physical storage space may consist of a single storage device or the disks of several storage devices. Clients store data on a logical volume, which means storing the data on the file system. The file system divides the data into many parts, each part being an object. Each object contains not only the data but also additional information such as a data identifier (ID). The file system writes each object to the physical storage space of that logical volume, and it records the storage location information of each object. Therefore, when a client requests access to data, the file system can allow the client to access the data based on the storage location information of each object.

[0072] The process by which a storage system allocates physical storage space to a logical volume is as follows: the physical storage space is pre-divided into strips according to the capacity estimate of the objects stored in the logical volume (this estimate often has a large margin relative to the actual capacity of the objects to be stored) and the grouping of the Redundant Array of Independent Disks (RAID). A logical volume can be understood as a strip, thus allocating physical storage space to the logical volume.

[0073] It should be noted that, Figure 1 The schematic diagram of the AI-based marine environmental monitoring system shown is merely an example. The AI-based marine environmental monitoring system and scenarios described in this application are intended to more clearly illustrate the technical solutions of this application and do not constitute a limitation on the technical solutions provided in this application. As those skilled in the art will know, with the evolution of AI-based marine environmental monitoring systems and the emergence of new business scenarios, the technical solutions provided in this application are also applicable to similar technical problems.

[0074] The following sections provide detailed descriptions of each example. It should be noted that the sequence numbers of the following embodiments are not intended to limit the preferred order of the embodiments.

[0075] Please refer to Figure 2 , Figure 2This is a schematic flowchart of an embodiment of the artificial intelligence-based marine environment monitoring method provided in this application, as shown below. Figure 2 As shown, the process of the artificial intelligence-based marine environment monitoring method provided in this application is as follows:

[0076] 201. Collect satellite images of the target sea area using a satellite remote sensing system.

[0077] Among them, the satellite remote sensing system can be equipped with a panchromatic multispectral imager, which covers four red tide sensitive bands: blue (450-520nm), green (520-590nm), red (630-690nm), and near-infrared (770-890nm). The imaging resolution of the target area is 10m, and the data can be received by the ground station of the National Satellite Ocean Application Center.

[0078] The satellite remote sensing system collects satellite images of the target sea area at a preset frequency, for example, once every hour, which can be set according to the specific situation.

[0079] Multiple mobile ocean buoys are deployed in the target sea area. For example, the core parameters and deployment requirements of the mobile ocean buoys are as follows: Monitoring parameter accuracy: water temperature (accuracy ±0.1℃), salinity (accuracy ±0.1‰), dissolved inorganic nitrogen (accuracy ±0.01mg / L), reactive phosphate (accuracy ±0.001mg / L), chlorophyll a (range 0-50μg / L, accuracy ±0.5μg / L), ocean current velocity (0-2m / s, accuracy ±0.05m / s) and ocean current direction (0-360°, accuracy ±5°).

[0080] The speed of a movable ocean buoy can be determined based on specific circumstances. Its location can be determined using positioning systems such as BeiDou. The movable ocean buoy is powered by its own propulsion system and can navigate to a predetermined location.

[0081] 202. Perform image segmentation on the satellite sea area image to obtain the first prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0082] In this embodiment of the application, the satellite sea area image is preprocessed to obtain a preprocessed satellite sea area image. The preprocessed satellite sea area image is then segmented to obtain the first prediction confidence score of each pixel in the satellite sea area image belonging to the red tide category.

[0083] Specifically, the preprocessing of satellite sea area images includes performing radiometric calibration, atmospheric correction, geometric correction, and cloud removal on the satellite sea area images in sequence, ultimately obtaining cloudless, accurately coordinated preprocessed satellite sea area images.

[0084] In this embodiment of the application, image segmentation is performed on the satellite sea area image to obtain the first prediction confidence score of each pixel in the satellite sea area image belonging to the red tide category, including:

[0085] (1) Acquire multiple historical sea area data collected at different times in the target sea area by satellite remote sensing system and mobile ocean buoy.

[0086] Historical marine data includes historical satellite images of marine areas collected by satellite remote sensing systems, as well as historical multidimensional marine data collected by various movable marine buoys on the historical satellite images. The historical multidimensional marine data includes water temperature, salinity, nutrient concentration, ocean current velocity, ocean current direction, and chlorophyll a concentration.

[0087] For example, a satellite remote sensing system acquires a satellite image of the target sea area every hour and obtains historical multidimensional ocean data from each movable ocean buoy within one hour, forming historical sea area data. Thus, multiple historical sea area data are combined to form sequence data.

[0088] (2) Input multiple historical sea area images into the time series prediction model to obtain the third prediction confidence of each pixel on the satellite sea area image belonging to the red tide category.

[0089] The time-series prediction model can be an LSTM time-series prediction model. The core is to uncover the temporal correlation between historical environmental factors and red tide occurrences. The time-series prediction model can be pre-trained.

[0090] (3) Based on the first image segmentation model, the satellite sea area image is segmented to obtain the fourth prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0091] The first image segmentation model can be the U-Net model.

[0092] Furthermore, the first image segmentation model can be an improved model based on the U-Net model.

[0093] Specifically, the first image segmentation model includes a channel attention submodule, a spatial attention submodule, a multi-scale feature extraction module, and a decoding module.

[0094] Specifically, based on the first image segmentation model, the satellite sea area image is segmented to obtain the fourth prediction confidence score of each pixel in the satellite sea area image belonging to the red tide category, including:

[0095] Step 1-1: Based on the channel attention submodule of the first image segmentation model, feature extraction is performed on the satellite sea area image to obtain the channel attention feature map.

[0096] Specifically, feature extraction is performed on satellite sea area images using encoded convolutional blocks to obtain the original feature map.

[0097] The channel attention submodule focuses on weight enhancement for red tide sensitive bands. Its core is to aggregate channel-dimensional information from the original feature map output by the encoding convolutional block. The feature map is processed using Global Average Pooling (GAP) and Global Max Pooling (GMP) to obtain two channel statistical vectors of dimension [C, 1, 1] (where C is the number of channels in the feature map). These two vectors are then input into a shared two-layer perceptron (MLP), which outputs the corresponding channel weight vectors. The learned channel weights are multiplied by the original feature map to obtain the channel attention feature map.

[0098] This submodule can automatically enhance the feature weights of the 670nm red light band, which is sensitive to chlorophyll a, and the 550nm blue-green band, which is sensitive to algal scattering, while suppressing the invalid channel features of the near-infrared band and clouds and white waves.

[0099] For example, the current satellite image of the sea area has 4 channels and the image size is 5000×5000.

[0100] The spatial attention submodule performs global average pooling on the original feature map, compressing the 5000×5000 features of each channel into a 1×1 channel statistical value; it learns the weights of each channel through two fully connected layers (the first layer has 2 neurons and the second layer has 4 neurons), in which the weights of the near-infrared channel (low reflectivity of red tide algae) and the green light channel (high reflectivity of red tide algae) are significantly improved; the learned channel weights are multiplied with the original feature map to obtain the channel attention feature map, thereby achieving feature enhancement of the red tide sensitive band.

[0101] Steps 1-2: Based on the spatial attention submodule of the first image segmentation model, feature extraction is performed on the satellite sea area image to obtain an enhanced red tide feature map.

[0102] The spatial attention submodule further implements spatial dimension region filtering based on the feature map after channel attention weighting. Global average pooling and global max pooling are performed on the spatial dimension of the feature map to obtain two spatial statistical feature maps with dimensions [1,H,W] (H and W are the height and width of the feature map). The two feature maps are concatenated by channel, and the number of channels is compressed to 1 through a 3×3 convolutional layer to obtain the channel statistics. The weights of each channel are learned through two fully connected layers. The learned channel weights are multiplied by the original feature map to obtain the enhanced red tide feature map.

[0103] Specifically, the spatial attention submodule performs a 3×3 convolution operation on the channel attention feature map to extract the spatial neighborhood correlation of pixels; a spatial attention mask is generated through the sigmoid activation function, in which the pixel weights of the red tide region (0.8-1.0) are significantly higher than those of the normal sea area (0.1-0.3); the spatial attention mask is multiplied with the channel attention feature map to obtain the enhanced red tide feature map, which can clearly distinguish the boundary between the red tide region and the normal sea area.

[0104] Steps 1-3: Based on the first image segmentation model, the multi-scale feature extraction module extracts and fuses the enhanced red tide feature map at multiple scales to obtain a multi-scale red tide feature map.

[0105] The multi-scale feature extraction module uses multiple dilated convolution kernels of different sizes to extract features from the enhanced red tide feature map, resulting in multiple single-scale feature maps. These single-scale feature maps are then fused to obtain a multi-scale red tide feature map.

[0106] Specifically, multiple dilated convolutional kernels of different sizes are used, with dimensions of 1×1, 3×3, and 5×5. Parallel dilated convolutional kernels of 1×1, 3×3, and 5×5 are used for multi-scale feature extraction. The 1×1 convolution preserves the detailed information of the original features and is used to capture small-area red tide budding patches; the 3×3 dilated convolution expands the receptive field and is used to identify medium-sized red tide areas; the 5×5 dilated convolution further enhances the receptive field and is used to cover large-scale red tide patches. Furthermore, dilated convolution can achieve feature extraction without reducing resolution.

[0107] To avoid redundant superposition of features at different scales, an information entropy weighting mechanism is introduced. The information entropy is calculated for the single-scale feature maps at the three scales respectively. The higher the information entropy, the richer the feature information, and the greater the weight assigned. Then, the single-scale feature maps at the three scales are weighted and summed to generate a fused multi-scale red tide feature map.

[0108] Steps 1-4: Based on the decoding module of the first image segmentation model, the multi-scale red tide feature map is processed to obtain the fourth prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0109] The decoding module of the first image segmentation model can adopt the decoding module of the U-Net model. The multi-scale red tide feature map is input into the decoding module of the first image segmentation model. Image resolution is gradually restored through upsampling, and a 5000×5000 fourth prediction confidence matrix is ​​finally output. This fourth prediction confidence matrix includes the fourth prediction confidence of each pixel in the satellite ocean image belonging to the red tide category. For example, the fourth prediction confidence of the same pixel in region B is 0.91, indicating that the probability of this point being identified as a red tide based on spatial spectral features is 91%.

[0110] (4) The first prediction confidence is obtained by weighted averaging of the third prediction confidence and the fourth prediction confidence.

[0111] In one specific embodiment, a first accuracy metric obtained by testing the time-series prediction model on a first test set is obtained, and a second accuracy metric obtained by testing the first image segmentation model on a second test set is obtained. A weighted average of the third prediction confidence Y3 and the fourth prediction confidence Y4 based on the first accuracy metric Acc3 and the second accuracy metric Acc4 is used to obtain the first prediction confidence Y1.

[0112] Specifically, the formula for calculating the first prediction confidence level Y1 is as follows:

[0113] Y1=(Acc3*Y3+Acc4*Y4) / 2.

[0114] 203. Based on the first prediction confidence of each pixel in the satellite marine image belonging to the red tide category, determine multiple first red tide risk areas in the satellite marine image.

[0115] In this embodiment of the application, multiple first red tide risk areas on the satellite sea area image are determined based on the first prediction confidence of each pixel on the satellite sea area image belonging to the red tide category, including:

[0116] (1) Pixels on satellite sea area images whose first prediction confidence is higher than the preset confidence threshold are identified as candidate pixels.

[0117] In this embodiment of the application, the preset confidence threshold can be 0.8 or other values, which can be set according to the specific situation.

[0118] (2) Based on the position of each candidate pixel in the satellite sea area image, multiple candidate pixels in the satellite sea area image are clustered to obtain multiple pixel clusters.

[0119] In this embodiment of the application, multiple candidate pixels on the satellite sea area image are clustered based on the position of each candidate pixel on the satellite sea area image to obtain multiple pixel clusters, including: determining the position of each candidate pixel. Multiple candidate pixels are randomly divided into K first pixel sets, where each first pixel set includes at least two pixels, and K is an integer greater than 1. Based on a preset number N, the pixels within each first pixel set are clustered to obtain N second pixel sets and the first position mean information of each second pixel set, where the first position mean information is the average value of each position in the second pixel set. Each second pixel set is determined as a target pixel set. Within each first pixel set, the second pixel set with the highest similarity to the first position mean information of the target pixel set is obtained, resulting in the second position mean information of the K second pixel sets corresponding to the target pixel set. This yields the second position mean information of each target pixel, resulting in N second position mean information. A set is created for each second position mean information, and each candidate pixel is placed into the set corresponding to the second position mean information with the highest similarity among the candidate pixel positions, resulting in N pixel clusters.

[0120] (3) The target polygon region corresponding to the pixel cluster is determined as the first red tide risk area.

[0121] In this context, the vertices of the target polygon region corresponding to the pixel cluster all belong to the pixels in the pixel cluster, and the pixels in the pixel cluster are not located outside the target polygon region.

[0122] Furthermore, multiple candidate polygon regions corresponding to multiple pixel clusters are obtained. The vertices of each candidate polygon region corresponding to a pixel cluster belong to pixels within that pixel cluster, and none of the pixels within the pixel cluster are located outside the candidate polygon region. The candidate polygon region with the smallest area among these multiple candidate polygon regions is determined as the target polygon region.

[0123] 204. Determine the buoy pixel category of the corresponding pixel point based on the current multidimensional ocean data collected by the movable ocean buoy.

[0124] Among them, the buoy pixel category corresponding to the movable ocean buoy is either the red tide category or the normal category.

[0125] Specifically, the threshold values ​​for each dimension of the current multidimensional ocean data are obtained. These dimensions include water temperature, salinity, nutrient concentration, ocean current velocity, ocean current direction, and chlorophyll a concentration. When a dimension indicator is below its corresponding threshold, it is identified as an anomalous indicator. The percentage of anomalous indicators in the current multidimensional ocean data of the movable ocean buoy is calculated. When this percentage exceeds a first preset threshold, the buoy pixel category is determined to be red tide. When the percentage is below a second preset threshold, the buoy pixel category is determined to be red tide. The second preset threshold is less than the first preset threshold. For example, the second preset threshold is 0.1, and the first preset threshold is 0.8.

[0126] In this embodiment of the application, the average number of abnormal indicators of each movable ocean buoy is obtained, and the average number of abnormal indicators is denoted as a. The variance of the number of abnormal indicators of each movable ocean buoy is calculated to obtain the variance of the number of abnormal indicators, and the variance of the number of abnormal indicators is denoted as b.

[0127] The first preset percentage threshold and the second preset percentage threshold are determined based on the weighting coefficient c, the average percentage a, and the percentage variance b. The first preset percentage threshold is (a + cb), and the second preset percentage threshold is (a - cb).

[0128] Specifically, the weighting coefficient c is denoted as [1,2].

[0129] 205. Generate cue vectors based on the buoy pixel category corresponding to the pixels of the movable ocean buoy.

[0130] In this embodiment, the pixels corresponding to the movable ocean buoy and the buoy pixel category corresponding to the pixels are used as cue points and input into the cue encoder of the second image segmentation model to obtain the cue vector.

[0131] In this embodiment, the second image segmentation model is a pre-trained model. The second image segmentation model includes an image encoder, a prompt encoder, and a mask decoder.

[0132] An image encoder is used to transform an input image into a high-dimensional, abstract image feature vector, providing visual information for subsequent segmentation. Image encoders can be ResNet, EfficientNet, etc., and the appropriate one can be selected based on the specific situation.

[0133] The cue encoder is used to convert red tide cue points into embedding vectors compatible with image features, thereby aligning the cue with the image feature vectors.

[0134] The mask decoder is used to fuse cue vectors and image feature vectors to generate the final segmentation mask. The mask decoder can be a Transformer decoder.

[0135] In one specific embodiment, cue points whose buoy pixels belong to the red tide category are determined as positive cue points, and cue points whose buoy pixels belong to the normal category are determined as negative cue points. The positive and negative cue points are input into the cue encoder of the second image segmentation model to obtain the cue vector.

[0136] 206. Based on the cue vector and satellite sea area image, perform secondary image segmentation on the satellite sea area image to obtain the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category, and determine multiple second red tide risk areas on the satellite sea area image based on the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0137] In this embodiment of the application, a second image segmentation is performed on the satellite sea area image based on the cue vector and the satellite sea area image to obtain the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category. This includes: inputting the satellite sea area image into the image encoder of the second image segmentation model to obtain the image feature vector; and inputting the image feature vector and the cue vector into the mask decoder of the second image segmentation model to obtain the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0138] Multiple second red tide risk areas on satellite marine images are determined based on the second prediction confidence level of each pixel belonging to a red tide category. The method for determining the second red tide risk areas based on the second prediction confidence level of pixels is the same as the method for determining the first red tide risk areas based on the first prediction confidence level of pixels, and will not be described in detail here.

[0139] like Figure 3 As shown, the centroids of the multiple secondary red tide risk areas are Q1, Q2, Q3, Q4, Q5, and Q6, respectively.

[0140] 207. Assign multiple movable marine buoys to each second red tide risk area.

[0141] In this embodiment of the application, multiple movable marine buoys are allocated to each second red tide risk area, including:

[0142] (1) The area in the satellite sea area image where multiple second red tide risk areas have been removed is identified as a normal area.

[0143] (2) Refine the normal area and generate skeleton refinement lines for the normal area on the satellite sea area image.

[0144] Specifically, the Zhang-Suen thinning algorithm is used to thin the normal region, generating skeleton thinning lines on the satellite marine image. These skeleton thinning lines are single-pixel thinning lines. For each pixel on the skeleton thinning line, the normal to the thinning line at that pixel is drawn, intersecting the boundary of the normal region at two points. The distance between these two intersection points and the pixel is the same.

[0145] like Figure 3 As shown, the skeleton refinement lines are black line segments between multiple secondary red tide risk areas.

[0146] (3) Determine the shortest planned path for the movable marine buoy to reach the centroid of the second red tide risk area along the skeleton refinement line.

[0147] In this embodiment, starting from a movable marine buoy and ending at the centroid of the second red tide risk area, multiple candidate planning paths are planned along the skeleton refinement line from the starting point to the ending point. The shortest candidate planning path among these multiple paths is determined as the shortest planning path. The candidate planning paths are located on the skeleton refinement line.

[0148] Furthermore, starting from the centroid of the second red tide risk area, multiple target rays are drawn, with adjacent target rays having the same angle, for example, an angle of 30 degrees between adjacent target rays. The first intersection point of the target ray and the skeleton refinement line is determined as the first intersection point. The first line segment formed by the centroid and the first intersection point is added to the skeleton refinement line to obtain an updated skeleton refinement line. The shortest planned path for the movable marine buoy to reach the centroid of the second red tide risk area along the updated skeleton refinement line is then determined.

[0149] like Figure 3 As shown, the centroid of one of the second red tide risk areas is Q1. Starting from Q1, multiple target rays are drawn. The first intersection point of the target ray and the skeleton refinement line is determined as the first intersection point. The multiple first intersection points are P1, P2, P3, and P4.

[0150] like Figure 3 As shown, the centroid of one of the second red tide risk areas is Q1, and one of the movable ocean buoys is B1. The shortest planned path for the movable ocean buoy B1 to reach Q1 along the skeleton refinement line is B1-P1-Q1.

[0151] Since there may be a lot of algae in the second red tide risk area, which may affect the navigation of the mobile ocean buoy, controlling the mobile ocean buoy to reach the second red tide risk area along the skeleton refinement line can avoid the algae and thus ensure that the mobile ocean buoy can reach its destination quickly.

[0152] (4) Determine the matching degree of the allocation of movable marine buoys and the second red tide risk area based on the shortest planning path.

[0153] In a specific implementation, the shorter the shortest planned path of the movable ocean buoy, the greater the matching degree of the allocation of the movable ocean buoy and the second red tide risk area.

[0154] In another specific embodiment, the direction of the movable ocean buoy toward the centroid of the corresponding second red tide risk area is determined as the navigation direction. The ocean current direction within the target area is determined, and the vector angle between the navigation direction and the ocean current direction is determined. The allocation matching degree of the movable ocean buoy and the second red tide risk area is determined based on the vector angle and the shortest planned path. Specifically, the longer the shortest planned path, the smaller the allocation matching degree; the larger the vector angle, the smaller the allocation matching degree.

[0155] Furthermore, the remaining power of each movable ocean buoy is determined, and the allocation matching degree of the movable ocean buoy and the second red tide risk area is determined based on the vector angle, the shortest planned path, and the remaining power. Specifically, the longer the shortest planned path, the smaller the allocation matching degree; the larger the vector angle, the smaller the allocation matching degree; and the larger the remaining power, the higher the allocation matching degree.

[0156] First, the three core indicators—shortest planning path, vector angle, and remaining power—are normalized to eliminate dimensional differences between them. For the shortest planning path, reverse normalization is used (the larger the shortest planning path, the smaller the normalized value), mapping it to the [0,1] interval to obtain the reverse normalized path value, denoted as Ppath. For the vector angle, reverse normalization is performed with a maximum value of 1 corresponding to 0° (buoy orientation perfectly aligned with ocean current direction) and a minimum value of 0 corresponding to 180° (buoy orientation perfectly opposite to ocean current direction) to obtain the reverse normalized direction value, denoted as Pangle. For the remaining power, forward normalization is performed with 1 corresponding to full power and 0 corresponding to depleted power to obtain the forward normalized power value, denoted as Ppower. Then, operational weights are assigned to the three indicators: the weight coefficient for the shortest planning path is w1, the weight coefficient for the vector angle is w2, and the weight coefficient for the remaining power is w3, satisfying w1+w2+w3=1.

[0157] The final matching degree M is calculated using the formula M=w1*Ppath+w2*Pangle+w3*Ppower. This formula integrates multiple factors into a unified matching degree value. The higher the value, the stronger the matching degree between the buoy and the second red tide risk area.

[0158] Suppose there is a movable marine buoy A and a second red tide risk area R in a certain sea area. We set the indicator weights as w1=0.3 (shortest planned path), w2=0.3 (vector angle), and w3=0.4 (remaining power). Buoy A currently has 80% remaining power (full power is 100%), and its shortest planned path to risk area R is 5km (the maximum planned path for buoys in this sea area is 10km), with a vector angle of 30°. First, we normalize the values: path dimension Ppath=0.5, angle dimension Pangle=0.833, and power dimension Ppower=0.8. Then, we substitute these values ​​into the formula to calculate the matching degree M=0.3×0.5+0.3×0.833+0.4×0.8=0.72.

[0159] (5) Allocate the target number of mobile marine buoys to the second red tide risk area based on the allocation matching degree.

[0160] In one specific embodiment, allocating a target number of movable ocean buoys to a second red tide risk area based on allocation matching degree includes: for the second red tide risk area, sorting the movable ocean buoys from largest to smallest based on allocation matching degree to obtain the top-ranked movable ocean buoys with the target allocation degree; and allocating the target number of movable ocean buoys to the second red tide risk area.

[0161] Specifically, the target allocation number of movable marine buoys for the second red tide risk area is determined based on the area of ​​the second red tide risk area. The larger the area of ​​the second red tide risk area, the larger the corresponding target allocation number.

[0162] In one specific embodiment, allocating a target number of movable ocean buoys to a second red tide risk area based on allocation matching degree includes: for the second red tide risk area, sorting movable ocean buoys whose pixel category is normal based on allocation matching degree from largest to smallest, obtaining the top-ranked movable ocean buoys with the target allocation number; and allocating the target allocation number of movable ocean buoys to the second red tide risk area.

[0163] In one specific embodiment, the number of movable ocean buoys located in the second red tide risk zone is determined as the first number of buoys, and the difference between the target allocation number and the first number of buoys is determined as the second number of buoys. For the second red tide risk zone, based on the allocation matching degree from largest to smallest, movable ocean buoys whose pixel category is normal and are not located in the second red tide risk zone are sorted to obtain the movable ocean buoys with the highest sorted second number of buoys. The movable ocean buoys with the first number of buoys located in the second red tide risk zone and the movable ocean buoys with the second number of buoys not located in the second red tide risk zone are determined as the target allocation number of movable ocean buoys, and the movable ocean buoys with the target allocation number are allocated to the second red tide risk zone.

[0164] 208. Control each movable marine buoy corresponding to the second red tide risk area to move to the corresponding second red tide risk area.

[0165] Specifically, the movable ocean buoy is controlled to move along the shortest planned path to the corresponding second red tide risk area.

[0166] Furthermore, a uniformly distributed number of target arrival points are generated within the second red tide risk area, and these target arrival points are assigned to a target number of movable ocean buoys. Using the target arrival points of the movable ocean buoys as endpoints, the buoys are controlled to move along the shortest planned path to the boundary of the corresponding second red tide risk area, and then controlled to move in a straight line to the corresponding target arrival point.

[0167] Specifically, the target arrival points for the second number of movable ocean buoys not located in the second red tide risk zone are assigned to the second number of buoys. The remaining target arrival points are assigned to the first number of movable ocean buoys located in the second red tide risk zone. Specifically, for the second number of movable ocean buoys located in the second red tide risk zone, the movable ocean buoy with the largest shortest planned path is designated as the target ocean buoy. The intersection of the shortest planned path of the target ocean buoy and the boundary of the second red tide risk zone is designated as the third intersection point. The target arrival point closest to the third intersection point is assigned to the target ocean buoy.

[0168] Specifically, the intersection of the target ray and the boundary of the second red tide risk area is defined as the second intersection point, and the line segment between the centroid and the second intersection point is defined as the second line segment. The second line segment is divided equally at preset intervals, generating multiple division points on the second line segment. The second red tide risk area is proportionally reduced, resulting in multiple reduced second red tide risk areas. These multiple reduced second red tide risk areas share the same centroid, and the boundary of each reduced second red tide risk area passes through one division point on the second line segment. The boundary of each reduced second red tide risk area is then equally divided, resulting in multiple division points on the boundary of each reduced second red tide risk area. The total number of division points on the boundaries of the multiple reduced second red tide risk areas is counted, and the preset interval is iteratively updated until the total number of division points equals the target allocation number. The division points of the target allocation number are then defined as the target arrival points of the target allocation number. In this process, after the boundary of each reduced second red tide risk area is equally divided, the difference between the straight-line distance between two adjacent dividing points and the preset interval is less than a preset threshold. For example, the preset interval is 30 pixels, and the preset threshold is 2 pixels or other values, which can be set according to the specific situation. This results in a uniformly distributed target arrival point.

[0169] like Figure 3 As shown, the centroid of one of the second red tide risk areas is Q1, and a movable ocean buoy is B1. The shortest planned path for movable ocean buoy B1 to reach Q1 along the skeleton refinement line is B1-P1-Q1. The third intersection point of the shortest planned path with the boundary of the second red tide risk area is J1. The target arrival point corresponding to movable ocean buoy B1 is D1. Therefore, the movable ocean buoy B1 is controlled to move along the shortest planned path to the boundary of the corresponding second red tide risk area, that is, along B1-P1-J1 to reach J1, and then the movable ocean buoy B1 is controlled to move along a straight line from J1 to the corresponding target arrival point D1.

[0170] Assigning target arrival points to mobile ocean buoys nearby reduces travel time and facilitates timely data collection. Furthermore, the even distribution of target arrival points covers a wider area, improving prediction accuracy.

[0171] 209. When a movable ocean buoy moves to the corresponding second red tide risk area, update multiple second red tide risk areas on the satellite ocean image.

[0172] In this embodiment of the application, when the movable ocean buoy moves to the corresponding second red tide risk area, the movable ocean buoy updates the multidimensional ocean data, updates the buoy pixel category of the corresponding pixel of the movable ocean buoy, and then updates multiple second red tide risk areas on the satellite sea area image, thereby enabling accurate monitoring of the marine environment.

[0173] To facilitate better implementation of the AI-based marine environment monitoring method provided in this application, this application also provides an AI-based marine environment monitoring device based on the aforementioned AI-based marine environment monitoring method. The meanings of the terms used are the same as in the AI-based marine environment monitoring method described above; for specific implementation details, please refer to the descriptions in the above method embodiments.

[0174] Please refer to Figure 4 , Figure 4 This is a schematic diagram of an embodiment of an artificial intelligence-based marine environment monitoring device provided in this application. The artificial intelligence-based marine environment monitoring device may include a data acquisition module 701, a first segmentation module 702, a first determination module 703, a second determination module 704, a prompt generation module 705, a second segmentation module 706, an allocation module 707, a control module 708, and an update module 709.

[0175] The acquisition module 701 is used to acquire satellite images of the target sea area through a satellite remote sensing system;

[0176] The first segmentation module 702 is used to perform image segmentation on the satellite sea area image to obtain the first prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0177] The first determining module 703 is used to determine multiple first red tide risk areas on the satellite sea area image based on the first prediction confidence of each pixel on the satellite sea area image belonging to the red tide category;

[0178] The second determining module 704 is used to determine the buoy pixel category of the pixel corresponding to the movable ocean buoy based on the current multidimensional ocean data collected by the movable ocean buoy, wherein the buoy pixel category of the pixel corresponding to the movable ocean buoy is either a red tide category or a normal category.

[0179] The prompt generation module 705 is used to generate a prompt vector based on the buoy pixel category of the corresponding pixel of the movable ocean buoy;

[0180] The second segmentation module 706 is used to perform secondary image segmentation on the satellite sea area image based on the cue vector and the satellite sea area image, to obtain the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category, and to determine multiple second red tide risk areas on the satellite sea area image based on the second prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

[0181] The allocation module 707 is used to allocate multiple movable marine buoys to each second red tide risk area;

[0182] Control module 708 is used to control each movable marine buoy corresponding to the second red tide risk area to move to the corresponding second red tide risk area;

[0183] Update module 709 is used to update multiple second red tide risk areas on satellite ocean images when a movable ocean buoy moves to the corresponding second red tide risk area.

[0184] Please refer to Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0185] The electronic device may include a radio frequency (RF) circuit 901, a memory 902 including one or more computer-readable storage media, an input unit 903, a display unit 904, a sensor 905, an audio circuit 906, a wireless Fidelity (WiFi) module 907, a processor 908 including one or more processing cores, and a power supply 909, etc. Those skilled in the art will understand that... Figure 5 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0186] RF circuit 901 can be used for receiving and transmitting signals during information transmission or calls. Specifically, it receives downlink information from the base station and hands it over to one or more processors 908 for processing; additionally, it transmits uplink data to the base station. Typically, RF circuit 901 includes, but is not limited to, an antenna, at least one amplifier, a tuner, one or more oscillators, a Subscriber Identity Module (SIM) card, a transceiver, a coupler, a Low Noise Amplifier (LNA), a duplexer, etc. Furthermore, RF circuit 901 can also communicate wirelessly with networks and other devices. Wireless communication can use any communication standard or protocol, including but not limited to GSM, GPRS, CDMA, WCDMA, LTE, email, and SMS services.

[0187] The memory 902 can be used to store software programs and modules. The processor 908 executes various functional applications and stack back operations by running the software programs and modules stored in the memory 902. The memory 902 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device (such as audio data, telephone directory, etc.). In addition, the memory 902 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 902 may also include a memory controller to provide access to the memory 902 for the processor 908 and the input unit 903.

[0188] Input unit 903 can be used to receive input digital or character information, and to generate keyboard, mouse, joystick, optical, or trackball signal inputs related to user settings and function control. Specifically, in one embodiment, input unit 903 may include a touch-sensitive surface and other input devices. A touch-sensitive surface, also known as a touch display or touchpad, can collect user touch operations on or near it (e.g., user operations using fingers, styluses, or any suitable object or accessory on or near the touch-sensitive surface) and drive corresponding connection devices according to a pre-set program. Optionally, the touch-sensitive surface may include a touch detection device and a touch controller. The touch detection device detects the user's touch location and the signal generated by the touch operation, transmitting the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, sends it to the processor 908, and can receive and execute commands from the processor 908. Furthermore, various types of touch-sensitive surfaces, such as resistive, capacitive, infrared, and surface acoustic wave, can be used. In addition to the touch-sensitive surface, input unit 903 may also include other input devices. Specifically, other input devices may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.

[0189] Display unit 904 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of electronic devices. These graphical user interfaces can be composed of graphics, text, icons, video, and any combination thereof. Display unit 904 may include a display panel, optionally configured as a liquid crystal display (LCD), organic light-emitting diode (OLED), or similar form. Furthermore, a touch-sensitive surface may cover the display panel. When the touch-sensitive surface detects a touch operation on or near it, it transmits the information to processor 908 to determine the type of touch event. Subsequently, processor 908 provides corresponding visual output on the display panel according to the type of touch event. Although in Figure 5 In this context, the touch-sensitive surface and the display panel are two separate components for implementing input and output functions. However, in some embodiments, the touch-sensitive surface and the display panel can be integrated to achieve both input and output functions.

[0190] Electronic devices may also include at least one sensor 905, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor can adjust the brightness of the display panel according to the ambient light level, and the proximity sensor can turn off the display panel and / or backlight when the electronic device is moved to the ear. As a type of motion sensor, a gravity acceleration sensor can detect the magnitude of acceleration in various directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that recognize the phone's posture (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometers, taps), etc. Other sensors that may be configured in electronic devices, such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, will not be described in detail here.

[0191] Audio circuitry 906, a speaker, and a microphone provide an audio interface between the user and the electronic device. Audio circuitry 906 converts received audio data into electrical signals, transmits them to the speaker, and the speaker converts them into sound signals for output. Conversely, the microphone converts collected sound signals into electrical signals, which are then received by audio circuitry 906, converted back into audio data, and processed by processor 908. The processed data is then transmitted via RF circuitry 901 to, for example, another electronic device, or output to memory 902 for further processing. Audio circuitry 906 may also include an earphone jack to facilitate communication between external headphones and the electronic device.

[0192] WiFi is a short-range wireless transmission technology. Electronic devices using the WiFi module 907 can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access. Although Figure 5 WiFi module 907 is shown, but it is understood that it is not a necessary component of the electronic device and can be omitted as needed without changing the nature of the invention.

[0193] The processor 908 is the control center of the electronic device. It connects various parts of the phone via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 902, and by calling data stored in the memory 902, thereby performing overall detection of the phone. Optionally, the processor 908 may include one or more processing cores; preferably, the processor 908 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into the processor 908.

[0194] The electronic device also includes a power supply 909 (such as a battery) that supplies power to various components. Preferably, the power supply can be logically connected to the processor 908 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 909 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0195] Although not shown, the electronic device may also include a camera, Bluetooth module, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 908 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 902 according to the following instructions, and the processor 908 runs the applications stored in the memory 902 to realize various functions:

[0196] Satellite images of the target sea area are acquired using a satellite remote sensing system. These images are then segmented to obtain a first predicted confidence level for each pixel belonging to a red tide category. Based on this first predicted confidence level, multiple first red tide risk areas are identified within the satellite images. The buoy pixel category of each pixel corresponding to a mobile ocean buoy is determined based on current multidimensional ocean data collected from mobile ocean buoys. The buoy pixel category of each pixel corresponding to a mobile ocean buoy is either a red tide category or a normal category. Based on the buoy pixel category of each pixel corresponding to a mobile ocean buoy... A cue vector is generated; based on the cue vector and satellite sea area image, secondary image segmentation is performed on the satellite sea area image to obtain the second predicted confidence of each pixel in the satellite sea area image belonging to the red tide category, and multiple second red tide risk areas are determined on the satellite sea area image based on the second predicted confidence of each pixel in the satellite sea area image belonging to the red tide category; multiple movable ocean buoys are assigned to each second red tide risk area; each movable ocean buoy corresponding to the second red tide risk area is controlled to move to the corresponding second red tide risk area; when the movable ocean buoy moves to the corresponding second red tide risk area, the multiple second red tide risk areas on the satellite sea area image are updated.

[0197] It should be noted that the electronic device provided in this application embodiment and the artificial intelligence-based marine environment monitoring method in the above embodiment belong to the same concept. The specific implementation process can be found in the above related embodiments, and will not be repeated here.

[0198] This application also provides a computer-readable storage medium storing a computer program thereon. When the computer program stored thereon is executed on the processor of the electronic device provided in the embodiments of this application, the processor of the electronic device performs the steps in the artificial intelligence-based marine environment monitoring method provided in this application. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0199] This application also provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform various optional implementations of the aforementioned artificial intelligence-based marine environmental monitoring method.

[0200] The above provides a detailed description of an artificial intelligence-based marine environment monitoring method, device, and storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

[0201] It should be noted that when the above embodiments of this application are applied to specific products or technologies, and user-related data is involved, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

Claims

1. A marine environmental monitoring method based on artificial intelligence, characterized in that, The artificial intelligence-based marine environmental monitoring method includes: Satellite images of the target sea area are acquired through a satellite remote sensing system; The satellite image of the sea area is segmented to obtain the first prediction confidence level of each pixel in the satellite image of the sea area belonging to the red tide category; Based on the first prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category, multiple first red tide risk areas are determined on the satellite sea area image; The buoy pixel category of the pixel corresponding to the movable ocean buoy is determined based on the current multidimensional ocean data collected by the movable ocean buoy, wherein the buoy pixel category of the pixel corresponding to the movable ocean buoy is either red tide category or normal category; A cue vector is generated based on the buoy pixel category of the pixel corresponding to the movable ocean buoy. The pixel corresponding to the movable ocean buoy and the buoy pixel category corresponding to the pixel are used as cue points and input into the cue encoder of the second image segmentation model to obtain the cue vector. Based on the cue vector and the satellite sea area image, a secondary image segmentation is performed on the satellite sea area image to obtain a second predicted confidence level that each pixel in the satellite sea area image belongs to the red tide category. Based on the second predicted confidence level that each pixel in the satellite sea area image belongs to the red tide category, multiple second red tide risk areas are determined on the satellite sea area image. Specifically, the satellite sea area image is input into the image encoder of the second image segmentation model to obtain an image feature vector; the image feature vector and the cue vector are input into the mask decoder of the second image segmentation model to obtain the second predicted confidence level that each pixel in the satellite sea area image belongs to the red tide category. Multiple movable marine buoys are allocated to each of the second red tide risk areas; Control each of the movable marine buoys corresponding to the second red tide risk area to move to the corresponding second red tide risk area; When the movable ocean buoy moves to the corresponding second red tide risk area, the multiple second red tide risk areas on the satellite ocean image are updated.

2. The marine environmental monitoring method based on artificial intelligence according to claim 1, characterized in that, The determination of multiple first red tide risk areas on the satellite sea area image based on the first prediction confidence score of each pixel on the satellite sea area image belonging to the red tide category includes: Pixels on the satellite sea area image whose first prediction confidence is higher than a preset confidence threshold are identified as candidate pixels; Based on the position of each candidate pixel in the satellite sea area image, multiple candidate pixels in the satellite sea area image are clustered to obtain multiple pixel clusters; The target polygon region corresponding to the pixel cluster is determined as the first red tide risk region.

3. The marine environmental monitoring method based on artificial intelligence according to claim 1, characterized in that, The allocation of multiple movable marine buoys to each of the second red tide risk areas includes: The area in the satellite sea area image where multiple second red tide risk areas have been removed is determined as a normal area; The normal region is refined, and skeleton refinement lines of the normal region are generated on the satellite sea area image; Determine the shortest planned path for the movable marine buoy to reach the centroid of the second red tide risk area along the skeleton refinement line; The allocation matching degree between the movable marine buoy and the second red tide risk area is determined based on the shortest planning path; Based on the allocation matching degree, the target number of the movable marine buoys are allocated to the second red tide risk area.

4. The marine environmental monitoring method based on artificial intelligence according to claim 1, characterized in that, The step of performing image segmentation on the satellite sea area image to obtain the first predicted confidence level of each pixel in the satellite sea area image belonging to the red tide category includes: The system acquires multiple historical sea area data collected at different times in the target sea area by the satellite remote sensing system and the mobile ocean buoy. The historical sea area data includes historical satellite sea area images collected by the satellite remote sensing system and historical multidimensional ocean data collected by each mobile ocean buoy on the historical satellite sea area images. The historical multidimensional ocean data includes water temperature, salinity, nutrient concentration, ocean current velocity, ocean current direction, and chlorophyll a concentration. By inputting multiple historical sea area images into a time series prediction model, the third prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category is obtained; Based on the first image segmentation model, the satellite sea area image is segmented to obtain the fourth prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category; The first prediction confidence is obtained by taking a weighted average of the third prediction confidence and the fourth prediction confidence.

5. The marine environmental monitoring method based on artificial intelligence according to claim 4, characterized in that, The step of segmenting the satellite sea area image based on the first image segmentation model to obtain the fourth prediction confidence score of each pixel in the satellite sea area image belonging to the red tide category includes: Based on the channel attention submodule of the first image segmentation model, feature extraction is performed on the satellite sea area image to obtain a channel attention feature map; Based on the spatial attention submodule of the first image segmentation model, feature extraction is performed on the satellite sea area image to obtain an enhanced red tide feature map; The multi-scale feature extraction module based on the first image segmentation model performs multi-scale feature extraction and fusion on the enhanced red tide feature map to obtain a multi-scale red tide feature map. Based on the decoding module of the first image segmentation model, the multi-scale red tide feature map is processed to obtain the fourth prediction confidence of each pixel in the satellite sea area image belonging to the red tide category.

6. A marine environmental monitoring device based on artificial intelligence, characterized in that, The artificial intelligence-based marine environmental monitoring device includes: The acquisition module is used to acquire satellite images of the target sea area through a satellite remote sensing system; The first segmentation module is used to perform image segmentation on the satellite sea area image to obtain the first prediction confidence level of each pixel in the satellite sea area image belonging to the red tide category. The first determining module is used to determine multiple first red tide risk areas on the satellite sea area image based on the first prediction confidence level of each pixel on the satellite sea area image belonging to the red tide category; The second determining module is used to determine the buoy pixel category of the pixel corresponding to the movable ocean buoy based on the current multidimensional ocean data collected by the movable ocean buoy, wherein the buoy pixel category of the pixel corresponding to the movable ocean buoy is either a red tide category or a normal category; The prompt generation module is used to generate a prompt vector based on the buoy pixel category of the pixel corresponding to the movable ocean buoy. The pixel corresponding to the movable ocean buoy and the buoy pixel category corresponding to the pixel are used as prompt points and input into the prompt encoder of the second image segmentation model to obtain the prompt vector. The second segmentation module is used to perform secondary image segmentation on the satellite sea area image based on the cue vector and the satellite sea area image, to obtain a second predicted confidence level that each pixel in the satellite sea area image belongs to the red tide category, and to determine multiple second red tide risk areas on the satellite sea area image based on the second predicted confidence level that each pixel in the satellite sea area image belongs to the red tide category. Specifically, the satellite sea area image is input into the image encoder of the second image segmentation model to obtain an image feature vector; the image feature vector and the cue vector are input into the mask decoder of the second image segmentation model to obtain the second predicted confidence level that each pixel in the satellite sea area image belongs to the red tide category. The allocation module is used to allocate multiple movable marine buoys to each of the second red tide risk areas; The control module is used to control each of the movable marine buoys corresponding to the second red tide risk area to move to the corresponding second red tide risk area; An update module is used to update multiple second red tide risk areas on the satellite sea area image when the movable ocean buoy moves to the corresponding second red tide risk area.

7. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor running the computer program in the memory to perform the steps in the artificial intelligence-based marine environment monitoring method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted for loading by a processor to perform the steps of the artificial intelligence-based marine environmental monitoring method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps in the artificial intelligence-based marine environment monitoring method according to any one of claims 1 to 5.

Images