A marine ecological environment detection method, device, equipment and storage medium

Abstract

The application discloses a marine ecological environment detection method, device, equipment and storage medium, and relates to the technical fields of artificial intelligence and marine ecological monitoring. In the method, a target task scene is determined, scene description information of the target task scene is obtained, various types of marine environment data collected under the target task scene are obtained, a comprehensive quality evaluation value of an actual marine environment is obtained based on the scene description information and attribute information of the various types of marine environment data, a data fusion mode matched with the comprehensive quality evaluation value is selected, the various types of marine environment data are fused based on the data fusion mode, fused environment data with multiple modes are obtained, and the fused environment data is analyzed to obtain a detection result of the target task scene. In this way, multi-modal marine environment data can be fused for the target task scene, the data collaboration advantage can be fully utilized, the limitation of single data can be avoided, and the accuracy and adaptability of marine ecological environment detection can be improved.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and marine ecological monitoring technology, and in particular to a method, device, equipment and storage medium for detecting marine ecological environment. Background Technology

[0002] With the continuous deepening of marine scientific research and the acceleration of marine resource development, the importance of marine ecological monitoring has become increasingly prominent. Currently, marine ecological monitoring systems are developing towards intelligence and automation.

[0003] In response to the limitations of single-sensor detection capabilities and adaptability, marine ecological monitoring has gradually adopted multi-sensor collaboration to address these limitations. By integrating multiple sensors (such as sonar, visible light cameras, and lidar) into underwater sensing terminals, multimodal data of underwater targets (such as acoustic data, optical image data, and 3D point cloud data) can be collected to obtain multidimensional information about targets under different water conditions. This approach compensates for the limitations of single sensors in terms of detection range, imaging accuracy, and environmental adaptability.

[0004] In the above scheme, each sensor collects corresponding modal data according to preset parameters, and then the collected multimodal data is simply classified and stored. In subsequent monitoring and analysis, a certain modal data is selected according to specific scenario requirements (e.g., acoustic data is used in deep-sea environments, optical image data is used in clear water bodies, and lidar data is used in shallow and medium-depth water bodies), and then monitoring tasks are carried out based on the modal data.

[0005] However, because the above-mentioned solutions independently call upon data from each modality, relying on the advantages of a single modality to adapt to corresponding scenarios (such as long-range detection of acoustic data, detailed presentation of optical image data, or high-precision morphological capture of lidar data), the advantages of each modality cannot be synergistically complementary when facing dynamic environmental changes or diverse monitoring tasks in marine ecological monitoring.

[0006] For example, when monitoring shallow-sea fish populations, optical image data is initially used to identify fish species in clear water. However, if a sudden increase in plankton causes turbidity, the optical image data becomes ineffective, and acoustic data is then used instead. Although the system collects multimodal data, including optical and acoustic data, during the entire monitoring process, in practical applications, it only calls upon data according to the specific scenario, thus only acquiring single-modal information. This affects the comprehensiveness of monitoring and the accuracy of analysis, reducing the adaptability to the task.

[0007] In view of this, there is a need to provide a new method for detecting the marine ecological environment in order to overcome the above-mentioned shortcomings. Summary of the Invention

[0008] This application provides a method, apparatus, equipment, and storage medium for detecting marine ecological environment, in order to avoid the problem of insufficient detection comprehensiveness and poor task adaptability caused by independent access to multimodal data and the inability to coordinate and complement each other.

[0009] In a first aspect, embodiments of this application provide a method for detecting marine ecological environment, the method comprising:

[0010] The target mission scenario is determined, and the scenario description information of the target mission scenario is obtained; the scenario description information is used to describe the actual marine environment under the target mission scenario.

[0011] Acquire various types of marine environmental data collected under the target task scenario; among them, different types of marine environmental data have different perception modalities;

[0012] Based on the scenario description information and the attribute information of the various marine environmental data, a comprehensive quality assessment value of the actual marine environment is obtained, and a data fusion method that matches the comprehensive quality assessment value is selected.

[0013] Based on the data fusion method described above, the various types of marine environmental data are fused to obtain fused environmental data with multiple modes.

[0014] The fused environment data is analyzed to obtain the detection results of the target task scene.

[0015] In an optional embodiment, the method further includes:

[0016] After acquiring various types of marine environmental data collected under the target task scenario, and before obtaining the comprehensive quality assessment value of the actual marine environment based on the scenario description information and the attribute information of the various types of marine environmental data, the method further includes:

[0017] For each type of marine environmental data, the following steps are performed: delete the portion of the marine environmental data whose correlation with the target task scenario does not meet the preset correlation conditions, and obtain updated marine environmental data.

[0018] In an optional embodiment, the method further includes:

[0019] The target task scenario is: to detect hydrological parameters of the marine environment; the actual marine environment described by the scenario description information is: comprehensive hydrological disturbance.

[0020] The process of obtaining a comprehensive quality assessment value for the actual marine environment based on the scene description information and the attribute information of various marine environmental data, and selecting a data fusion method that matches the comprehensive quality assessment value, includes:

[0021] For each type of marine environmental data, the following is performed: the quality of the hydrological parameters represented by the hydrological-related data contained in the marine environmental data is used as the attribute information of the marine environmental data;

[0022] Based on the comprehensive hydrological disturbance and the quality of the hydrological parameters, a corresponding hydrological detection and evaluation value is obtained, and the hydrological detection and evaluation value is used as the comprehensive quality evaluation value.

[0023] Based on the testing accuracy of the hydrological parameters represented by the hydrological detection and evaluation values, a suitable data fusion method is selected.

[0024] In an optional embodiment, the method further includes:

[0025] The target task scenario is: to identify marine organisms in the marine environment; the scenario description information describes the actual marine environment as: the overall survival status of marine organisms.

[0026] The process of obtaining a comprehensive quality assessment value for the actual marine environment based on the scene description information and the attribute information of various marine environmental data, and selecting a data fusion method that matches the comprehensive quality assessment value, includes:

[0027] For each type of marine environmental data, the following steps are performed: the target identifiability represented by the marine organism-related data contained in the marine environmental data is used as the attribute information of the marine environmental data;

[0028] Based on the overall survival status of marine organisms and the identifiability of the target, a corresponding marine organism identification assessment value is obtained, and the marine organism identification assessment value is used as the overall quality assessment value.

[0029] Based on the accuracy of marine organism identification as represented by the marine organism identification evaluation value, an appropriate data fusion method is selected.

[0030] In an optional embodiment, the method further includes:

[0031] The target task scenario is: to issue an abnormal early warning for the density of marine organisms in the marine environment; the actual marine environment described by the scenario description information is: the overall survival status of marine organisms.

[0032] The process of obtaining a comprehensive quality assessment value for the actual marine environment based on the scene description information and the attribute information of various marine environmental data, and selecting a data fusion method that matches the comprehensive quality assessment value, includes:

[0033] For each type of marine environmental data, the following steps are performed: the credibility of marine organism density represented by the marine organism-related data contained in the marine environmental data is used as the attribute information of the marine environmental data;

[0034] Based on the overall survival status of marine organisms and the reliability of marine organism density, a marine organism density early warning assessment value is obtained, and the marine organism density early warning assessment value is used as the overall quality assessment value.

[0035] Based on the accuracy of the early warning judgment represented by the marine organism density early warning assessment value, an appropriate data fusion method is selected.

[0036] In an optional embodiment, the method further includes:

[0037] The process of fusing various types of marine environmental data based on the aforementioned data fusion method to obtain multimodal fused environmental data includes:

[0038] Under the target task scenario, obtain the first fusion weight of each type of marine environmental data in the data fusion method;

[0039] When the data fusion method is data-level fusion, the various types of marine environmental data are aligned in data format and then fused using the corresponding first fusion weight to obtain fused environmental data with multiple modes.

[0040] When the data fusion method is feature-level fusion, the environmental features of each type of marine environmental data are extracted, and the obtained environmental features are unified in dimension. Then, the corresponding first fusion weight is used to fuse them to obtain fused environmental data with multiple modes.

[0041] In an optional embodiment, the method further includes:

[0042] When the data fusion method is decision-level fusion, the second fusion weight of each type of marine environmental data under the decision-level fusion method is obtained in the target task scenario;

[0043] The various types of marine environmental data are analyzed to obtain corresponding sub-detection results;

[0044] The confidence level of each sub-detection result is evaluated separately to obtain the corresponding confidence level evaluation result;

[0045] Each sub-detection result whose confidence assessment result meets the preset assessment criteria is taken as a positive sub-detection result, and its corresponding second fusion weight is maintained.

[0046] Each sub-detection result whose confidence assessment result does not meet the preset assessment criteria is taken as a negative sub-detection result, and its corresponding second fusion weight is reduced and adjusted to obtain the adjusted third fusion weight.

[0047] Based on the positive sub-detection results and their corresponding second fusion weights, and combined with the negative detection results and their corresponding third fusion weights, the sub-detection results are fused to obtain the corresponding detection results.

[0048] In an optional embodiment, the method further includes:

[0049] The method is implemented by an object detection model. After obtaining the detection results of the target task scene, it further includes:

[0050] The target task scenario, the various marine environmental data, and the detection results are used as a new sample data to update the sample dataset, and the real environmental state described by the detection results is used as the sample label of the sample data; each other sample data in the sample dataset includes: an other task scenario, various marine environmental data collected under the other task scenario, and corresponding other detection results;

[0051] The sample dataset is divided into a training dataset and a test dataset;

[0052] The detection model to be trained is trained based on the training dataset to obtain an initial detection model;

[0053] The initial detection model is tested and trained using the test dataset to obtain the target detection model.

[0054] In an optional embodiment, the method further includes:

[0055] If the test accuracy of the target detection model does not reach a preset accuracy threshold, the method further includes:

[0056] Filter out the task scenarios corresponding to test data whose test results do not match the real environment from the test dataset;

[0057] For each of the aforementioned task scenarios, the following steps are performed: acquiring various new marine environmental data and corresponding new detection results under the aforementioned task scenario; and adding the task scenario, the various new marine environmental data, and the new detection results as a new sample data to the aforementioned sample dataset, and using the real environmental state described by the new detection results as the sample label of the new sample data.

[0058] A new training dataset is derived from the sample dataset, and the object detection model is trained based on the new training dataset to obtain a new object detection model.

[0059] Secondly, embodiments of this application provide a marine ecological environment detection device, comprising:

[0060] Acquisition module: used to determine the target task scenario and acquire scenario description information of the target task scenario; the scenario description information is used to describe the actual marine environment under the target task scenario, and to acquire various types of marine environmental data collected under the target task scenario; wherein, different types of marine environmental data have different perception modalities;

[0061] Processing module: used to obtain the comprehensive quality assessment value of the actual marine environment based on the scene description information and the attribute information of the various types of marine environmental data, and select a data fusion method that matches the comprehensive quality assessment value; and to perform fusion processing on the various types of marine environmental data based on the data fusion method to obtain fused environmental data with multiple modes;

[0062] Parsing module: Parses the fused environment data to obtain the detection results of the target task scene.

[0063] Thirdly, this application provides an electronic device, comprising:

[0064] Memory, used to store computer programs;

[0065] When a processor executes a computer program stored in the memory, it implements the steps of the above-described data update method.

[0066] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described data update method.

[0067] In this embodiment, the target task scenario for marine ecological environment detection is first determined and scenario description information and multimodal marine environmental data are obtained. After filtering out low-correlation data, a comprehensive quality assessment value is obtained by combining scenario information and data attribute information, and an appropriate data fusion method is selected accordingly. Subsequently, the data is fused according to the corresponding fusion method (data level, feature level) to obtain multimodal fused environmental data. After parsing, the detection results are output. The method also includes fusing the sub-detection results of various types of marine environmental data (decision level) to obtain the final detection result.

[0068] Meanwhile, the sample dataset is updated with the detection results and the real environment status for the training and testing of the target detection model. If the model accuracy does not meet the standard, data will be added for mismatched scenarios and the model will be retrained. This effectively avoids the problems of insufficient detection comprehensiveness and poor task adaptation caused by independent calls to multimodal data and the inability to coordinate and complement each other, ensuring detection accuracy and model adaptability. Attached Figure Description

[0069] Figure 1 A schematic diagram of a system architecture provided in an embodiment of this application;

[0070] Figure 2 A flowchart illustrating a method for detecting marine ecological environment provided in this application embodiment;

[0071] Figure 3 This is a schematic diagram illustrating a method for acquiring various types of marine environmental data provided in an embodiment of this application;

[0072] Figure 4 A flowchart illustrating a method for filtering various types of marine environmental data provided in this application embodiment;

[0073] Figure 5 A schematic diagram illustrating a method for filtering various types of marine environmental data provided in this application embodiment;

[0074] Figure 6 A flowchart illustrating a method for selecting data fusion methods for different target task scenarios, provided in this application embodiment;

[0075] Figure 7 A flowchart illustrating a method for data fusion of various types of marine environmental data provided in this application embodiment;

[0076] Figure 8 A flowchart illustrating a method for fusing detection results of various types of marine data, provided in this application embodiment;

[0077] Figure 9A A schematic diagram of the detection logic for a marine ecological environment, provided as an embodiment of this application, where the detection device is a sensing device.

[0078] Figure 9B A schematic diagram of the detection logic for a marine ecological environment, provided as an embodiment of this application, where the detection device is a background device.

[0079] Figure 10 A schematic diagram of the structure of a marine ecological environment detection device provided in an embodiment of this application;

[0080] Figure 11 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0081] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The specific operational methods in the method embodiments can also be applied to the device embodiments or system embodiments. It should be noted that in the description of this application, "multiple" is understood as "at least two". "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. A connected to B can represent: A and B directly connected, and A and B connected through C. Furthermore, in the description of this application, terms such as "first" and "second" are used only for distinguishing the purpose of description and should not be construed as indicating or implying relative importance or order.

[0082] With the continuous deepening of marine scientific research and the acceleration of marine resource development, the importance of marine ecological monitoring has become increasingly prominent. Currently, marine ecological monitoring systems are developing towards intelligence and automation.

[0083] In response to the limitations of single-sensor detection capabilities and adaptability, marine ecological monitoring has gradually adopted multi-sensor collaboration to address these limitations. By integrating multiple sensors (such as sonar, visible light cameras, and lidar) into underwater sensing terminals, multimodal data of underwater targets (such as acoustic data, optical image data, and 3D point cloud data) can be collected to obtain multidimensional information about targets under different water conditions. This approach compensates for the limitations of single sensors in terms of detection range, imaging accuracy, and environmental adaptability.

[0084] In the above scheme, each sensor collects corresponding modal data according to preset parameters, and then the collected multimodal data is simply classified and stored. In subsequent monitoring and analysis, a certain modal data is selected according to specific scenario requirements (e.g., acoustic data is used in deep-sea environments, optical image data is used in clear water bodies, and lidar data is used in shallow and medium-depth water bodies), and then monitoring tasks are carried out based on the modal data.

[0085] However, the above schemes only call up each modal data independently to adapt to the corresponding scenario. Once marine ecological monitoring faces dynamic environmental changes or diverse tasks, the advantages of each modal data cannot be coordinated and complemented.

[0086] To address the aforementioned technical issues, this application embodiment determines the target task scenario and obtains its scenario description information. It also acquires various types of marine environmental data collected within the target task scenario. Based on the scenario description information and the attribute information of the various marine environmental data, a comprehensive quality assessment value of the actual marine environment is obtained. A data fusion method matching the comprehensive quality assessment value is selected, and the various marine environmental data are fused using this method to obtain multimodal fused environmental data. This fused environmental data is then analyzed to obtain the detection results for the target task scenario. This approach avoids the problems of independent multimodal data access and lack of collaborative complementarity, which can lead to insufficient detection comprehensiveness and poor task adaptation.

[0087] The following is a brief introduction to the system architecture diagram applicable to the technical solutions of the embodiments of this application. It should be noted that the system architecture diagram described below is only used to illustrate the embodiments of this application and is not intended to limit the scope of the application.

[0088] For example, see Figure 1 As shown, it is a system architecture diagram applicable to the embodiments of this application. The system architecture includes at least a sonar 100, a visible light camera 101, a lidar 102, and a marine back-end processor 103.

[0089] Sonar 100: As an acoustic sensing device for detecting the marine ecological environment, Sonar 100 is subjected to the circumferential forces of water surface density, ocean currents and topography on the sound wave path, thereby generating acoustic echo data to detect the marine ecological environment.

[0090] Visible light camera 101: As a visual perception device for detecting marine ecological environment, the visible light camera 101 is subjected to the comprehensive modulation of the light path by sea surface illumination, water transparency and atmospheric disturbance, thereby generating optical image data to detect marine ecological environment in real time.

[0091] LiDAR 102: As an optical sensing device for detecting the marine ecological environment, LiDAR 102 is affected by the scattering and absorption of laser pulses by sea waves, water optical properties and aerosols, thereby generating three-dimensional point cloud data to detect the marine ecological environment in real time.

[0092] Marine back-end processor 103: As the core hub for marine ecological environment monitoring, the marine back-end processor 103 is used to uniformly receive and process marine environmental data from sonar 100, visible light camera 101 and lidar 102.

[0093] In addition, if the sonar 100, visible light camera 101, and lidar 102 have data analysis capabilities, they can each directly analyze the acquired acoustic data, optical image data, and 3D point cloud data to obtain the corresponding detection results. If the sonar 100, visible light camera 101, and lidar 102 do not have data analysis capabilities, the marine backend processor 103 processes and analyzes the acoustic data, optical image data, and 3D point cloud data uploaded by the sonar 100, visible light camera 101, and lidar 102 to obtain the detection results.

[0094] For example, see Figure 2 As shown in the embodiments of this application, the specific process of the marine ecological environment detection method is as follows:

[0095] Step 200: Determine the target task scenario and obtain the scenario description information of the target task scenario.

[0096] Specifically, in this embodiment, the access sensing device is configured according to the target task, and the target task scenario is determined and the scenario description information of the target task scenario is obtained. The scenario description information describes the actual marine environment under the target task scenario.

[0097] For example, when shallow-sea organism detection is the target task scenario, the target task configuration integrates two sensing devices: lidar and visible light camera. When the scene description information under the target task scenario, that is, the actual marine environment does not have good lighting and water quality conditions, the data from both lidar and visible light camera sensing devices are fused to detect shallow-sea organisms.

[0098] By following the target task scenario and scenario description information, we can gain insight into the actual situation of the marine ecological environment, provide factual support for the dynamic configuration of sensing devices, and enable the task scenario-driven marine ecological environment detection method to have adaptive capabilities in the complex and ever-changing marine environment.

[0099] Step 201: Acquire various marine environmental data collected in the target mission scenario.

[0100] For details, please refer to Figure 3 As shown in the embodiments of this application, by configuring access sensing devices according to the target task scenario and using the sensing devices to collect marine environmental data, various types of marine environmental data can be obtained, and different types of marine environmental data have different sensing modalities.

[0101] For example, when the target task is to identify fish schools in the lower and middle layers of the water, sonar, visible light cameras, and lidar are specifically configured to address the characteristics of weak light and poor visible light penetration in these waters. Sonar can penetrate the water to obtain information on fish distribution, quantity, and size (acoustic data), but lacks color information; visible light cameras can identify fish density, morphology, and distinguish species by color (optical image data), but only cover the surface layer; lidar can construct the three-dimensional outline of fish schools, distinguish organisms, and correct sonar size errors (3D point cloud data). Based on these three sensing devices, acoustic data, optical image data, and 3D point cloud data of the marine environment can be collected, and the three types of marine environmental data have different sensing modalities, allowing for complementary advantages between different modalities.

[0102] By collecting and acquiring multimodal marine environmental data, we can capture multi-source information about the marine environment under different perception dimensions, providing data support for subsequent multimodal data fusion.

[0103] Furthermore, after performing step 201 and before performing step 202, in order to eliminate redundant data that is weakly related to the target task scenario and ensure that the data input to the target detection model is highly related to the target task, optionally, data filtering can also be performed on each type of marine environmental data.

[0104] For details, please refer to Figure 4 As shown in the embodiments of this application, the specific steps for filtering each type of marine environmental data are as follows:

[0105] Step 4000: For each type of marine environmental data, perform the following: Delete the data in the marine environmental data whose correlation with the target task scenario does not meet the preset correlation conditions, and obtain the updated marine environmental data.

[0106] For details, please refer to Figure 5 As shown in the embodiments of this application, for each type of marine environmental data obtained, there may be some data whose correlation with the target task scenario does not meet the preset correlation conditions. In this case, it is necessary to delete this part of the data from each type of marine environmental data, so as to obtain marine environmental data with a high correlation with the target task scenario, that is, to obtain the updated marine environmental data.

[0107] For example, in the target task scenario of detecting coral reef bleaching, the data in the deep-water scattered echo data collected by sonar that is too far away from the coral reef will be removed, the images of coral reefs with yellowing due to clouds or obscured by ships collected by visible light cameras will be removed, and the non-coral reef point cloud data collected by lidar will be removed. Finally, the data that is highly correlated with the coral reef surface and its surrounding waters in each type of marine environmental data will be retained as the updated marine environmental data.

[0108] By removing redundant information that is weakly related to the target mission from each type of marine environmental data, the subsequent processing load can be significantly reduced, and the updated marine environmental data can be ensured to be consistent with the target mission scenario, thereby providing reliable data support for the execution of subsequent target missions.

[0109] Step 202: Based on the scene description information and the attribute information of various marine environmental data, obtain the comprehensive quality assessment value of the actual marine environment, and select a data fusion method that matches the comprehensive quality assessment value.

[0110] Specifically, in the embodiments of this application, different target task scenarios have different scenario description information, and the attribute information of various marine environmental data obtained under different target task scenarios is different. The attribute information of various marine environmental data describes the comprehensive quality of each type of marine environmental data in the corresponding task scenario. Based on the scenario description information and the attribute information of various marine environmental data, the comprehensive quality assessment value of the actual marine environment can be obtained. Based on the comprehensive quality assessment value, a data fusion method that matches it can be selected.

[0111] For details, please refer to Figure 6 As shown, different data fusion methods are selected for different target task scenarios, and the technical process for selecting data fusion methods also varies, which may include the following operations:

[0112] Step 202-1: The target task scenario is to detect the hydrological parameters of the marine environment. The actual marine environment described in the scenario description information is the comprehensive hydrological disturbance situation. Based on the scenario description information and the attribute information of various marine environmental data, the comprehensive quality assessment value of the actual marine environment is obtained, and a data fusion method that matches the comprehensive quality assessment value is selected.

[0113] Specifically, for each type of marine environmental data, the following steps are performed: the quality of hydrological parameters represented by the hydrological data contained in the marine environmental data is used as the attribute information of the marine environmental data; then, based on the comprehensive hydrological interference and the quality of hydrological parameters, the corresponding hydrological detection and evaluation values ​​are obtained, and these hydrological detection and evaluation values ​​are used as the comprehensive quality evaluation values; finally, based on the test accuracy of the hydrological parameters represented by the hydrological detection and evaluation values, an appropriate data fusion method is selected.

[0114] For example, in the task of hydrological parameter detection in nearshore aquaculture areas (the core requirements of the task are: detecting flow velocity, sediment concentration, and water depth to support the prevention of water flow impact, sediment deposition control, and cage placement adaptation for aquaculture organisms), firstly, considering the comprehensive hydrological interference situation of the area being close to the estuary, with rapid water flow and a sharp increase in sediment concentration during high tide, and calm water but fluctuating light during low tide, sonar, visible light cameras, and lidar are configured to collect various marine environmental data. Based on the target task, the weight of sonar is set to 0.6 (sonar obtains water flow velocity data through the acoustic Doppler principle, which can monitor the flow velocity in the middle and lower layers but is subject to error due to turbulence, resulting in moderate hydrological parameter quality), the weight of visible light cameras is set to 0.5 (cameras calculate sediment concentration through image grayscale comparison, which can capture changes in surface sediment concentration but is greatly affected by light, resulting in moderate to low hydrological parameter quality), and the weight of lidar is set to 0.8 (lidar obtains water depth data through laser ranging, is less affected by water transparency and has stable accuracy, resulting in high hydrological parameter quality).

[0115] Then, sonar-collected water flow velocity data is used to predict whether the water flow will impact aquaculture organisms, visible light camera-collected water sediment content data is used to assess the risk of sediment deposition, and lidar-collected water depth data is used to adapt the placement of aquaculture cages. Based on the quality of the collected hydrological data parameters, the weights of water flow velocity data (associated with core requirements for aquaculture organism safety), water sediment content data (associated with requirements for sediment deposition control in aquaculture areas), and water depth data (associated with requirements for precise placement of aquaculture cages) are set to 0.4, resulting in a hydrological detection assessment value of 0.6 × 0.4 + 0.5 × 0.3 + 0.8 × 0.3, which is 0.7. In this case, the hydrological detection assessment value represents a precision P of 0.7 for the hydrological parameter detection.

[0116] Based on the task configuration for hydrological parameter detection, the computing power level C of the equipment is set to 0.8. Considering the significant interference of high tide but stable low tide in the hydrological environment, the environmental dynamics E is set to 0.2. Based on the requirement of generating a hydrological detection report once a day, the real-time requirement L is set to 0.1.

[0117] Different data fusion methods exhibit varying sensitivities to accuracy requirements P, device computing power level C, environmental dynamism E, and real-time requirements L. The preset weights of these three data fusion methods for accuracy requirements P, device computing power level C, environmental dynamism E, and real-time requirements L are shown in Table 1.

[0118] Table 1

[0119]

[0120] The system calculates the adaptation score for each fusion method using a weighted scoring formula: Then, according to the preset precision weights under each data fusion method in Table 1 Computing power weight Environmental dynamics weight and real-time weight Furthermore, the actual accuracy (P), equipment computing power (C), environmental dynamics (E), and real-time requirements (L) under any hydrological parameter detection scenario can be used to obtain a data-level fusion score. The feature-level fusion score is 1.32. The decision-level fusion score is 1.11. The value is 0.78. Therefore, when conducting hydrological parameter monitoring in nearshore aquaculture areas, the selected data fusion method is data-level fusion.

[0121] Step 202-2: The target task scenario is to identify marine organisms in the marine environment. The actual marine environment described by the scenario description information is the comprehensive survival status of marine organisms. Based on the scenario description information and the attribute information of various marine environmental data, the comprehensive quality assessment value of the actual marine environment is obtained, and a data fusion method that matches the comprehensive quality assessment value is selected.

[0122] Specifically, in this application example, for each type of marine environmental data, the following steps are performed: the target identifiability represented by the marine organism-related data contained in the marine environmental data is used as the attribute information of the marine environmental data; then, based on the comprehensive survival status of marine organisms and the target identifiability, the corresponding marine organism identification assessment value is obtained, and the marine organism identification assessment value is used as the comprehensive quality assessment value; finally, based on the identification accuracy of marine organisms represented by the marine organism identification assessment value, an appropriate data fusion method is selected.

[0123] For example, in the task of target identification of marine life in the waters surrounding nearshore aquaculture areas (the core requirements of the task are: identifying the species and quantity of fish, the density of plankton, and the location of jellyfish to support the assessment of feed resources, water quality monitoring, and pest control in aquaculture areas), firstly, considering the comprehensive survival situation of marine life in the area, such as the concentration of fish in the middle and lower layers, the light sensitivity of plankton, and the tidal activity of jellyfish, sonar, visible light cameras, and lidar are configured to collect various marine environmental data. Based on the degree of dependence of the target task on the data of each device, the weight of sonar is set to 0.6 (sonar acquires the two-dimensional outline and echo intensity of fish schools, which can initially identify the distribution of fish schools but lacks color information, and the target identification is moderate), the weight of visible light camera is set to 0.55 (camera captures the color and shape of surface organisms, which can distinguish plankton from jellyfish but is greatly affected by light, and the target identification is moderate to low), and the weight of lidar is set to 0.9 (lidar generates three-dimensional outlines of organisms, which can distinguish between living and non-living things and correct misjudgments of body shape, and the target identification is high).

[0124] Then, the acoustic data of the fish school collected by sonar is used to initially determine the distribution and quantity of the fish school; the optical image data of surface organisms collected by the visible light camera is used to identify the density of plankton and the morphology of jellyfish; and the 3D point cloud data of organisms collected by lidar is used to construct the 3D outline of the organisms and eliminate non-biological interference. Next, based on the target identifiability of the collected marine organism data, the weights of the fish school acoustic data (associated with the core requirement of bait resource assessment), the surface organism optical image data (associated with the requirement of water quality and predator control), and the 3D point cloud data of organisms (associated with the requirement of identification accuracy) are set to 0.3. This yields a marine organism identification assessment value of 0.6 × 0.4 + 0.55 × 0.3 + 0.9 × 0.3, which is 0.675. In this case, the marine organism identification assessment value represents the accuracy of target identification of marine organisms, P = 0.675.

[0125] Finally, based on the task configuration for marine biometrics, considering the deployment of on-site shipborne mid-range processors, the computing power level C is set to 0.6; considering the changes in biometric distribution with light and tides, the environmental dynamics E is set to 0.5; and based on the task's requirement to update the identification results hourly, the real-time requirement L is set to 0.4. Combining the relevant preset weights in Table 1, the data-level fusion score can be obtained. The feature-level fusion score is 1.2275. The decision-level fusion score is 1.2425. The value is 1.22. Therefore, for the task of identifying marine biological targets in the waters surrounding nearshore aquaculture areas, feature-level fusion is selected as the data fusion method.

[0126] Step 202-3, the target task scenario is: to issue anomaly warnings for marine organism density in the marine environment; the scenario description information describes the actual marine environment as: the overall survival status of marine organisms; based on the scenario description information and attribute information of various marine environmental data, obtain the comprehensive quality assessment value of the actual marine environment, and select a data fusion method that matches the comprehensive quality assessment value.

[0127] Specifically, in this application example, for each type of marine environmental data, the following steps are performed: the credibility of marine organism density represented by marine organism-related data contained in the marine environmental data is used as the attribute information of the marine environmental data; then, based on the overall survival status of marine organisms and the credibility of marine organism density, a marine organism density early warning assessment value is obtained, and the marine organism density early warning assessment value is used as the comprehensive quality assessment value; finally, based on the early warning judgment accuracy represented by the marine organism density early warning assessment value, an appropriate data fusion method is selected.

[0128] For example, in the task of issuing early warnings for abnormal marine organism densities in the waters surrounding nearshore aquaculture areas (the core requirement of the task is to monitor changes in the density of plankton and fish, triggering an early warning when the plankton density exceeds a preset plankton density threshold and the fish density falls below a preset fish density threshold, in order to prevent red tide disasters and ensure the supply of feed for aquaculture organisms), firstly, considering the overall survival situation of marine organisms in the area, such as tidal fluctuations of plankton, fish aggregation and distribution, and the ease with which impurities can cause confusion, sonar, visible light cameras, and lidar are configured to collect various types of marine environmental data, and then, according to the target task, [the following is a separate, unrelated sentence:] ... Regarding the dependence on equipment data, the weight of sonar is set to 0.65 (sonar inverts fish density through echo intensity and can penetrate water to monitor fish in the middle and lower layers, but is affected by the echo of impurities, so the reliability of density calculation is moderate), the weight of visible light camera is set to 0.5 (cameras count plankton density by the pixel ratio of the image and can capture changes in surface density, but are affected by light reflection, so the reliability of the data is moderate to low), and the weight of lidar is set to 0.85 (lidar distinguishes between living and non-living things by point cloud density, can eliminate the interference of impurities, accurately count biological density, and has high data reliability).

[0129] Then, the acoustic data of fish schools collected by sonar is used to invert the density of fish schools in the middle and lower layers, the optical image data of surface organisms collected by visible light cameras is used to statistically analyze the density of plankton, and the three-dimensional point cloud data of organisms collected by lidar is used to distinguish between living and non-living things and correct density statistical errors. Next, based on the density reliability of the collected marine biological data, the weight of the acoustic data of fish schools is set to 0.4 (associating with the core needs of fish school density early warning and ensuring food supply), the weight of the optical image data of surface organisms is set to 0.4 (associating with the core needs of red tide early warning), and the weight of the three-dimensional point cloud data of organisms is set to 0.2 (associating with the needs of density data interference removal and improving early warning accuracy). Thus, the marine biological density early warning assessment value is 0.65×0.4+0.5×0.4+0.85×0.2, which is 0.61. At this point, the marine biological density early warning assessment value represents an accuracy P of 0.61 for the early warning of abnormal marine biological density.

[0130] Finally, based on the mission configuration for marine organism density early warning, considering the deployment of shipborne embedded processors (supporting only lightweight data processing), the equipment computing power level C is set to 0.4; considering the dynamic changes in organism density with tides, the environmental dynamics E is set to 0.8 (low environmental stability); and based on the mission's requirement to output early warning results within a short time, the real-time requirement L is set to 0.9 (high real-time performance required). Combining the relevant preset weights in Table 1, the data-level fusion score can be obtained. The feature-level fusion score is 1.119. The decision-level fusion score is 1.427. The value is 1.804. Therefore, for the early warning task of abnormal marine organism density in the waters surrounding nearshore aquaculture areas, the selected data fusion method is decision-level fusion.

[0131] By combining scenario descriptions of different target tasks with attribute information of various marine environmental data in the corresponding scenarios (such as hydrological parameter quality, target identifiability, and biological density reliability), a comprehensive quality assessment value of the actual marine environment is obtained. Based on the accuracy requirements reflected by the assessment value, and combined with the requirements of equipment computing power, environmental dynamics, and real-time performance, an appropriate data fusion method is selected. This approach can specifically match the core needs of different detection tasks, fully leverage the synergistic advantages of multimodal data, avoid detection bias caused by mismatch between fusion methods and tasks, and effectively improve the accuracy and adaptability of marine ecological environment detection.

[0132] Step 203: Based on the data fusion method, perform fusion processing on various types of marine environmental data to obtain fused environmental data with multiple modes.

[0133] For details, please refer to Figure 7 As shown, the process of fusing various types of marine environmental data to obtain multimodal fused environmental data includes the following technical steps:

[0134] Step 2030: Obtain the first fusion weight of each type of marine environmental data under the data fusion method in the target task scenario.

[0135] Specifically, in this embodiment of the application, the weight of each type of marine environmental data obtained under the target mission scenario is different. Therefore, when performing data fusion, it is necessary to obtain the first fusion weight of each type of marine environmental data under the data fusion method. The setting of the first fusion weight needs to be combined with the core requirements of the target mission, the contribution of each type of data in the fusion method, and the actual data quality to ensure that the weight allocation is highly adapted to the mission objectives.

[0136] Step 2030-1: When the data fusion method is data-level fusion, after aligning the data formats of various marine environmental data, the corresponding first fusion weight is used for fusion to obtain fused environmental data with multiple modes.

[0137] Specifically, in this embodiment of the application, if a data-level fusion method is selected for the target task scenario, the sensing device will align the formats of the various types of marine data collected, and then fuse the various types of marine data that have been format aligned according to the first fusion weight of each modal data, thereby obtaining fused environmental data with multiple modalities.

[0138] For example, in step 202-1, targeting the hydrological parameter detection scenario in nearshore aquaculture areas, a data-level fusion method has been selected, and the first fusion weights for the water flow velocity data collected by sonar, the sediment content data collected by visible light cameras, and the water depth and height data collected by lidar are determined to be 0.4, 0.3, and 0.3, respectively. First, the three types of hydrological data formats are aligned and uniformly converted into a structured data format containing timestamps, spatial coordinates, and parameter values; then, the aligned data is weighted according to the first fusion weight. For example, for hydrological data at 10:00:00 (a certain moment) and 30°15′N, 120°20′E (a certain spatial coordinate system), the aligned sonar water velocity data is 0.3 m / s at 10:00:00, 30°15′N, 120°20′E; the aligned visible light camera sediment concentration data is 0.2 kg / m³ at 10:00:00, 30°15′N, 120°20′E; and the aligned lidar water depth data is 10:00:00, 30°15′N. At 10:00:00, at 30°15′N, 120°20′E, the water flow velocity fusion value is 0.3×0.35, the sediment concentration fusion value is 0.2×0.25, and the water depth fusion value is 8.1×0.4. The final fused environmental data for this spatiotemporal point are: water flow velocity 0.12 m / s, water depth sediment concentration 0.06-0.2 kg / m³, and water depth 2.43 m.

[0139] Step 2030-2: When the data fusion method is feature-level fusion, extract the environmental features of each type of marine environmental data, unify the dimensions of the obtained environmental features, and then use the corresponding first fusion weight to fuse them to obtain fused environmental data with multiple modes.

[0140] Specifically, in this embodiment of the application, if a feature-level fusion method is selected for the target task scenario, environmental features that can characterize the core requirements of the task should be extracted from various marine environmental data first. Then, by unifying the dimensions of each environmental feature, the environmental features whose feature dimensions have been unified are fused to obtain multimodal fused environmental data.

[0141] For example, in step 202-2, in the target scenario for marine biological target identification in nearshore aquaculture areas, a feature-level fusion method has been selected, and the first fusion weights for the sonar-acquired fish acoustic data, the visible light camera-acquired surface biological optical image data, and the lidar-acquired biological 3D point cloud data are determined to be 0.4, 0.3, and 0.3, respectively. First, environmental features of the three types of data are extracted: from the sonar fish acoustic data, features such as the mean echo intensity of the fish school and the aspect ratio of the target contour are extracted (to characterize the fish school distribution density and body size); from the visible light camera-acquired surface biological optical image data, features such as the color histogram and biological contour edge gradient are extracted (to distinguish the morphological and color differences between plankton and jellyfish); from the lidar biological 3D point cloud data, features such as point cloud density and the target 3D volume ratio are extracted (to eliminate non-biological interference and correct the biological body size judgment); then, the obtained environmental features are processed... The dimensions are unified by mapping the mean sonar echo intensity (0-255), visible light color histogram features (0-1), and lidar point cloud density (0-100 points / cm²) to a 0-1 feature dimension (e.g., the mean sonar echo intensity of 200 is normalized to 0.78, the visible light color feature remains unchanged at 0.6, and the lidar point cloud density of 50 points / cm² is normalized to 0.5), ensuring that different modal features can directly participate in the fusion calculation. Then, the feature data of the same spatiotemporal point after dimension unification are weighted according to the first fusion weight. For example, for the spatiotemporal point 10:30:00, 30°16′N, 120°21′E, the dimensionally unified feature data are: sonar fish swarm feature value 0.6, visible light surface organism feature value 0.55, and lidar 3D feature value 0.8. Using the corresponding weights, the fused fish swarm feature value is calculated to be 0.6×0.4, the surface organism feature value to be 0.55×0.35, and the 3D feature value to be 0.8×0.25. The final fused environmental data for this spatiotemporal point is: fish swarm feature value 0.24, surface organism feature value 0.1925, and 3D feature value 0.2.

[0142] By obtaining the first fusion weight of various marine environmental data under the corresponding data fusion method for different target mission scenarios, and then processing them separately according to the differences in fusion method, the data format is aligned and weighted according to the first fusion weight during data-level fusion. During feature-level fusion, the environmental features of various types of data are extracted and the dimensions are unified before being weighted according to the first fusion weight. Finally, the fusion processing of various types of marine environmental data can be completed to obtain multimodal fused environmental data that meets the mission requirements.

[0143] Step 204: Analyze the fused environment data to obtain the detection results of the target task scene.

[0144] Specifically, in the embodiments of this application, under the target task scenario, after various marine environmental data are processed by the corresponding fusion method to obtain multimodal fused environmental data, the appropriate analytical model and judgment rules are called in combination with the core requirements of the task, key features in the fused data are extracted and matched with classification standards, interference information is removed and comprehensive judgment is made, and the final corresponding detection result is obtained.

[0145] For example, in the target task scenario of step 202-3, assuming that fused data for a certain spatiotemporal point (11:00:00, 30°15′N, 120°20′E) is obtained through decision-level fusion: fish density fused value 0.3 fish / m³, plankton density fused value 0.4 kg / m³. The biological density early warning analysis model is invoked to extract key features of fish / plankton density, match preset standards (fish density < 0.5 fish / m³ requires early warning, plankton density > 0.5 kg / m³ requires early warning) and determine: the fish density at this spatiotemporal point is too low and attention should be paid to food replenishment; the plankton density has not reached the early warning threshold. Finally, the detection results of fish density early warning and plankton safety for this sea area at 11:00:00 are output, meeting the requirements for density anomaly early warning.

[0146] By analyzing fused environmental data and combining it with the core requirements of the task to call upon adapted analytical models and judgment rules, key features can be accurately extracted and matched with classification standards, thereby efficiently obtaining detection results for the target task scenario. This process can specifically meet the needs of different tasks, ensuring the accuracy and reliability of detection results, providing a strong basis for subsequent decision-making, and effectively improving the overall efficiency of marine ecological environment monitoring.

[0147] In one optional embodiment, when the data fusion method is decision-level fusion, the sub-detection results of various types of marine data are fused to obtain the final detection result.

[0148] For details, please refer to Figure 8 As shown in the embodiment of this application, when the data fusion method is decision-level fusion, the second fusion weights of various types of marine environmental data under the decision-level fusion method are obtained in the target task scenario. Then, the various types of marine environmental data are parsed to obtain corresponding sub-detection results, and the confidence of each sub-detection result is evaluated to obtain corresponding confidence evaluation results. Next, each sub-detection result whose confidence evaluation result meets the preset evaluation criteria is taken as a positive sub-detection result, and its corresponding second fusion weight is maintained. Each sub-detection result whose confidence evaluation result does not meet the preset evaluation criteria is taken as a negative sub-detection result, and its corresponding second fusion weight is reduced and adjusted to obtain an adjusted third fusion weight. Finally, based on each positive sub-detection result and its corresponding second fusion weight, combined with each negative detection result and its corresponding third fusion weight, the sub-detection result fusion processing is performed to obtain the corresponding detection result.

[0149] For example, in step 202-3, in the target task scenario of early warning of abnormal marine organism density in nearshore aquaculture areas, a decision-level fusion method has been selected. First, the second fusion weights of sonar (fish density), visible light camera (plankton density), and lidar (non-biological interference) data are obtained, with weights of 0.4, 0.4, and 0.2, respectively. Next, the various types of data are analyzed: the acoustic data shows a fish density of 0.3 fish / m³ (below the warning threshold of 0.5 fish / m³, requiring a warning), the optical image data shows a plankton density of 0.6 kg / m³ (above the warning threshold of 0.5 kg / m³, requiring a warning), and the 3D point cloud data shows a non-biological impurity ratio of 5% (no interference), resulting in three sub-detection results. Subsequently, the confidence levels are evaluated: the confidence level of the sonar result is 0.85, the confidence level of the visible light result is 0.6, and the confidence level of the lidar result is 0.9. The positive / negative sub-detection results are then determined: sonar and lidar results meet the criteria and are considered positive sub-detections, maintaining their second fusion weight; visible light results do not meet the preset evaluation criteria and are considered negative sub-detections, with their second fusion weight reduced from 0.4 to 0.2 (third fusion weight). Finally, the fused confidence level is calculated as 0.4×0.85+0.2×0.6+0.2×0.9=0.52, resulting in a fusion confidence level of 0.52. Therefore, the detection results indicate that abnormal fish density in a nearshore aquaculture area requires a strong warning, and abnormal plankton density requires careful monitoring.

[0150] By fusing sub-detection results from various marine environmental data at the decision level and adjusting weights based on confidence assessment, the impact of high-confidence results can be highlighted while the interference of low-confidence information can be mitigated. This approach can integrate judgments from multiple data sources, improve the reliability and accuracy of the final detection results, effectively integrate the advantages of different types of data, provide more comprehensive and accurate detection conclusions for target mission scenarios, and meet the needs of effective fusion and efficient analysis of complex data in marine environmental monitoring.

[0151] In one optional embodiment, the method for detecting the marine ecological environment is implemented by a target detection model. After obtaining the detection results of the target task scene, the target detection model needs to be trained.

[0152] Specifically, in this embodiment, firstly, the target task scenario, various marine environmental data, and detection results are added to the sample dataset as a new sample data entry, and the real environmental state described by the detection results is used as the sample label for the sample data. Each other sample data entry in the sample dataset includes: an other task scenario, various marine environmental data collected under the other task scenario, and corresponding other detection results. Then, the sample dataset is divided into a training dataset and a test dataset. The detection model to be trained is trained based on the training dataset to obtain an initial detection model. The initial detection model is then tested and trained using the test dataset to obtain the target detection model.

[0153] For example, to monitor nearshore plankton and understand the marine ecological status, the target scenario is first determined as a summer nearshore plankton monitoring area. Various marine environmental data, such as seawater temperature, salinity, and plankton images, are collected using sensors. Preliminary detection tools then yield results showing a plankton density of 280 organisms per liter, containing three dominant species. Next, the summer nearshore plankton monitoring area, the collected marine environmental data, and the aforementioned detection results are used as new sample data to update the sample dataset. Simultaneously, the actual environmental state of the plankton density per liter and the number of dominant species in the area is used as the sample label. Other samples in the sample dataset include other scenarios such as a "spring nearshore plankton monitoring area," corresponding environmental data, and detection results. The sample dataset is then divided into training and testing datasets. The training dataset is used to train the detection model, obtaining an initial detection model. The testing dataset is then used to test the training, ultimately resulting in a target detection model capable of accurately detecting nearshore plankton.

[0154] By integrating the target task scenario, relevant marine environmental data, and detection results into new samples, and combining them with real-world environmental conditions as labels to update the dataset, a target detection model can be built through training and testing. This approach continuously enriches the sample diversity and scenario coverage for model learning. It allows the model to constantly absorb new scenario information, improving its adaptability and detection accuracy to different task scenarios, ensuring that the model's output highly matches the actual environmental conditions, and providing reliable model support for marine ecological environment monitoring.

[0155] In one optional embodiment, if the test accuracy of the target detection model does not reach a preset accuracy threshold, the target detection model needs to be retrained.

[0156] Specifically, in this embodiment, the test data whose test results do not match the real environment state are selected from the test dataset for each task scenario; for each task scenario, the following steps are performed: acquiring various new marine environment data and corresponding new detection results under the task scenario, and adding the task scenario, various new marine environment data and new detection results as a new sample data to the sample dataset, and using the real environment state described by the new detection results as the sample label of the new sample data; dividing a new training dataset from the sample dataset, and training the target detection model based on the new training dataset to obtain a new target detection model.

[0157] It should be noted that before acquiring various new marine environmental data and corresponding new detection results in the mission scenario, the test data in the test dataset that do not match the real environmental conditions should be cleaned and enhanced to ensure that the data quality meets the standards. Then, it should be determined whether the sample data in the existing sample dataset is sufficient. If the sample data is insufficient, data collection should be triggered to supplement the sample dataset (i.e., to acquire various new marine environmental data and corresponding new detection results in the mission scenario).

[0158] For example, in testing a nearshore plankton detection model, if its accuracy fails to reach a preset threshold, scenarios where the detection results do not match the actual situation are selected from the test data. The test data for these scenarios is then processed: blurry plankton images are removed, abnormal seawater temperature values ​​are corrected, and the images are rotated and their brightness adjusted to enhance data quality. For instance, if the existing sample size for a scenario is only 12, far from meeting the requirements, new data collection is initiated to supplement the area with seven consecutive days of seawater monitoring data and high-resolution plankton images. Then, the new data for the target scenario and the actual situation labels are added to the sample set, the training set is re-divided, and the original model is retrained. The final result is a new target detection model.

[0159] By targeting task scenarios that do not match the model's performance during testing, and through data cleaning, enhancement, and supplementary sample collection followed by retraining, the model's detection shortcomings in specific scenarios can be accurately addressed. This approach can specifically improve the quality and quantity of samples, strengthen the model's adaptability to complex scenarios, and effectively improve the accuracy of the retrained model to reach a preset threshold. This ensures that the model continuously outputs reliable results in marine ecological environment monitoring, enhancing its practicality and stability.

[0160] In the above embodiments, the process of acquiring, processing and analyzing various types of marine environmental data is completed by the detection device. In different application scenarios, the detection device can correspond to different functional entities.

[0161] On the one hand, when the sensing device has data analysis capabilities, it can be used directly as a detection device. When placed in seawater, the sensing device can collect multimodal marine environmental data in real time, including acoustic data, optical image data, and 3D point cloud data, using its built-in sensors. Furthermore, it can analyze and generate corresponding detection results directly based on the collected marine environmental data, tailored to the specific task scenario, without relying on external equipment for analysis.

[0162] For example, see Figure 9A As shown, the multimodal perception layer 900 resides on the sensing device, while the global data processing layer 901, the integrated training and push central layer 902, and the intelligent interactive application layer 903 reside on the backend device. When the sensing device has data parsing capabilities, it uploads the obtained detection results to the integrated training and push central layer 902 via the multi-source data access gateway of the global data processing layer 901. The integrated training and push central layer 902 analyzes the received test results according to the configured business logic, generates the analysis results for the final target task, and then uploads the analysis results to the intelligent interactive layer 903 for real-time display. If the current scene model training task requires new data, the sensing device uploads various types of marine environmental data collected to the global data governance layer 901. Then, according to the scene task configuration, it processes the various types of marine environmental data into training data and adds them to the sample dataset. After the training dataset is partitioned, it is uploaded to the integrated training and push central layer 902 for training the new scene model.

[0163] For example, in the target scenario of detecting shallow-sea organisms, this task is equipped with two modes of sensing devices: LiDAR and visible light cameras. An environmental detection model can be trained based on the collected optical image data to determine underwater video acquisition or target detection. The model outputs three categories: 1) Good environment: refers to an underwater environment with good lighting and water quality, where the optical image data quality is good. Target detection using only optical image data yields relatively accurate results; that is, only video images captured by the optical camera are used to call the corresponding model for real-time analysis, saving computing power; 2) Moderate environment: Detection is performed by fusing the LiDAR and visible light camera models. The advantage of LiDAR in detecting targets with poor visibility is utilized to improve detection accuracy. Appropriate image enhancement processing can be enabled for the optical image data; 3) Poor environment: Optical image data quality is too poor to be recovered through image enhancement. Analysis can be performed using only the 3D point cloud detection model of the LiDAR. Environmental conditions can be determined using a lightweight image quality assessment method with strong real-time performance. Assessment parameters include: a) Brightness L: the average pixel value of the image; b) Contrast C: the standard deviation of the image pixels. The judgment logic is as follows: ① Good environment: L≥120 and C≥40; ② Average environment: 80≤L<120 or 20≤C<40; ③ Poor environment: L<80 or C<20. The evaluation method, parameters, and thresholds can be defined and adjusted in the target task scenario configuration.

[0164] On the other hand, when the sensing devices lack data analysis capabilities, the detection devices act as back-end equipment. In this case, the sensing devices deployed in seawater are only responsible for collecting various types of marine environmental data, such as acoustic data, optical image data, and 3D point cloud data. After collection, these raw data are uploaded to the back-end equipment located at sea, where the back-end equipment performs unified data analysis and processing to obtain the detection results for the target task scenario.

[0165] For example, see Figure 9B As shown, when the sensing device lacks data parsing capabilities, after collecting raw marine environmental data such as acoustic data, optical image data, and 3D point cloud data, the sensing device directly uploads it to the multi-source data access gateway of the full-domain data processing layer 901. The gateway performs data format verification and preliminary straightening. Subsequently, the full-domain data processing layer 901 transmits the processed raw data to the training and propulsion integrated central layer 902. This layer calls the appropriate parsing model to analyze the data, generate detection results for the target task scenario, and then optimizes them according to business logic before uploading them to the intelligent interactive application layer 903 for display. If additional model training data is needed, the full-domain data processing layer 901 will filter data that meets the scenario requirements from the stored raw data, trigger the intelligent preprocessing module to convert it into training data and integrate it into the sample dataset. After dividing the training dataset, it is uploaded to the training and propulsion integrated central layer 902 for training the model in the new scenario.

[0166] The detection device's acquisition, processing, and analysis modules work together, combining scene information and data attributes to select an appropriate fusion method. It can also update the sample training model. Furthermore, the detection device can correspond to different functional entities (sensing devices with analytical capabilities or back-end devices). It can enable sensing devices to directly generate detection results or have back-end devices process the data, thus fully adapting to the needs of multiple scenarios and improving the accuracy, flexibility, and model adaptability of marine ecological environment detection.

[0167] Based on the same inventive concept, this application also provides a marine ecological environment detection device, for example, see [reference]. Figure 10 As shown, the marine ecological environment monitoring device 1000 includes:

[0168] Acquisition Module 10000: Used to determine the target task scenario and acquire the scenario description information of the target task scenario; the scenario description information is used to describe the actual marine environment under the target task scenario, and to acquire various types of marine environmental data collected under the target task scenario; among them, different types of marine environmental data have different perception modalities;

[0169] Processing module 10001: is used to obtain the comprehensive quality assessment value of the actual marine environment based on scene description information and attribute information of various marine environmental data, and select a data fusion method that matches the comprehensive quality assessment value; and to perform fusion processing on various marine environmental data based on the data fusion method to obtain fused environmental data with multiple modes;

[0170] Parsing module 10002: Parses the fused environment data to obtain the detection results of the target task scene.

[0171] In an optional embodiment, the processing module 10001 is specifically used for:

[0172] After acquiring various marine environmental data collected in the target mission scenario, and before obtaining the comprehensive quality assessment value of the actual marine environment based on the scenario description information and the attribute information of various marine environmental data, the following steps are also included:

[0173] For each type of marine environmental data, the following steps are performed: delete the data from the marine environmental data that does not meet the preset correlation conditions with the target task scenario, and obtain the updated marine environmental data.

[0174] In an optional embodiment, the processing module 10001 is further configured to:

[0175] When the target task scenario is: to detect hydrological parameters of the marine environment, the actual marine environment described by the scenario description information is: comprehensive hydrological disturbances.

[0176] Based on scene description information and attribute information from various marine environmental data, a comprehensive quality assessment value for the actual marine environment is obtained, and a data fusion method matching the comprehensive quality assessment value is selected, including:

[0177] For each type of marine environmental data, the following steps are taken: the quality of hydrological parameters represented by the hydrological data contained in the marine environmental data is used as the attribute information of the marine environmental data;

[0178] Based on the comprehensive hydrological disturbances and the quality of hydrological parameters, the corresponding hydrological monitoring and evaluation values ​​are obtained, and the hydrological monitoring and evaluation values ​​are used as the comprehensive quality evaluation values.

[0179] Based on the testing accuracy of hydrological parameters expressed by hydrological detection and evaluation values, an appropriate data fusion method is selected.

[0180] In an optional embodiment, the processing module 10001 is further configured to:

[0181] The target task scenario is: to identify marine organisms in the marine environment; the scenario description information describes the actual marine environment as: the overall survival status of marine organisms.

[0182] Based on scene description information and attribute information from various marine environmental data, a comprehensive quality assessment value for the actual marine environment is obtained, and a data fusion method matching the comprehensive quality assessment value is selected, including:

[0183] For each type of marine environmental data, the following steps are taken: the target identifiability represented by the marine biological-related data contained in the marine environmental data is used as the attribute information of the marine environmental data;

[0184] Based on the overall survival status of marine organisms and the identifiability of targets, corresponding marine organism identification assessment values ​​are obtained, and these marine organism identification assessment values ​​are used as comprehensive quality assessment values.

[0185] Based on the accuracy of marine organism identification expressed by the marine organism identification assessment value, an appropriate data fusion method is selected.

[0186] In an optional embodiment, the processing module 10001 is further configured to:

[0187] The target task scenario is: to issue anomaly warnings for the density of marine organisms in the marine environment; the scenario description information describes the actual marine environment as: the overall survival status of marine organisms.

[0188] Based on scene description information and attribute information from various marine environmental data, a comprehensive quality assessment value for the actual marine environment is obtained, and a data fusion method matching the comprehensive quality assessment value is selected, including:

[0189] For each type of marine environmental data, the following steps are performed: the reliability of marine organism density represented by the marine organism-related data contained in the marine environmental data is used as the attribute information of the marine environmental data;

[0190] Based on the overall survival status of marine organisms and the reliability of marine organism density, a marine organism density early warning assessment value is obtained, and the marine organism density early warning assessment value is used as the comprehensive quality assessment value.

[0191] Based on the accuracy of early warning judgment expressed by the early warning assessment value of marine organism density, an appropriate data fusion method is selected.

[0192] In an optional embodiment, the processing module 10001 is further configured to:

[0193] Various types of marine environmental data are fused using data fusion methods to obtain multimodal fused environmental data, including:

[0194] Under the target task scenario, obtain the first fusion weight of each type of marine environmental data under the data fusion method;

[0195] When the data fusion method is data-level fusion, various marine environmental data are aligned in data format and then fused using the corresponding first fusion weight to obtain fused environmental data with multiple modes.

[0196] When the data fusion method is feature-level fusion, the environmental features of each type of marine environmental data are extracted separately, and the obtained environmental features are unified in dimensions and then fused using the corresponding first fusion weight to obtain fused environmental data with multiple modes.

[0197] In an optional embodiment, the processing module 10001 is further configured to:

[0198] After obtaining the detection results for the target task scene, the following also applies:

[0199] The target mission scenario, various marine environmental data, and detection results are added to the sample dataset as a new sample data, and the real environmental state described by the detection results is used as the sample label of the sample data. Each other sample data in the sample dataset contains: an other mission scenario, various marine environmental data collected under the other mission scenario, and corresponding other detection results.

[0200] The sample dataset is divided into a training dataset and a test dataset;

[0201] The initial detection model is obtained by training the detection model to be trained based on the training dataset.

[0202] The initial detection model was tested and trained using a test dataset to obtain the target detection model.

[0203] In an optional embodiment, the processing module 10001 is further configured to:

[0204] If the test accuracy of the target detection model does not reach the preset accuracy threshold, the method also includes:

[0205] Filter out the task scenarios corresponding to test data whose test results do not match the real environment from the test dataset;

[0206] For each task scenario, the following steps are performed: acquire various new marine environmental data and corresponding new detection results under the task scenario; add the task scenario, various new marine environmental data and new detection results as a new sample data to the sample dataset; and use the real environmental state described by the new detection results as the sample label of the new sample data.

[0207] A new training dataset is created by dividing the sample dataset and training the object detection model on the new training dataset to obtain a new object detection model.

[0208] In an optional embodiment, the parsing module 10002 can be used to:

[0209] When the data fusion method is decision-level fusion, obtain the second fusion weight of each type of marine environmental data under the decision-level fusion method in the target task scenario;

[0210] Various types of marine environmental data are analyzed to obtain corresponding sub-detection results;

[0211] The confidence level of each sub-detection result is evaluated separately to obtain the corresponding confidence level evaluation result;

[0212] Each sub-detection result whose confidence assessment result meets the preset assessment criteria is taken as a positive sub-detection result, and its corresponding second fusion weight is maintained.

[0213] Each sub-detection result whose confidence assessment result does not meet the preset assessment criteria is regarded as a negative sub-detection result, and its corresponding second fusion weight is reduced and adjusted to obtain the adjusted third fusion weight.

[0214] Based on the positive sub-detection results and their corresponding second fusion weights, and combined with the negative detection results and their corresponding third fusion weights, the sub-detection results are fused to obtain the corresponding detection results.

[0215] The acquisition, processing, and analysis modules of this marine ecological environment monitoring device work in tandem to not only filter out low-correlation marine environmental data and select appropriate fusion methods to process multimodal data by combining scene information and data attributes, but also update the sample training model, supplement data for retraining when the model accuracy is insufficient, and adjust the weights of sub-detection results for decision-level fusion. This fully leverages the advantages of multimodal data, improves detection accuracy and model adaptability, and provides reliable support for marine ecological environment monitoring.

[0216] Based on the same inventive concept, this application provides a computer device that can perform the functions of the aforementioned marine ecological environment detection device. (Refer to...) Figure 11 Electronic devices include:

[0217] At least one processor 11001 and a memory 11002 connected to at least one processor 11001. In this embodiment, the specific connection medium between the processor 11001 and the memory 11002 is not limited. Figure 11 The example shown is the connection between processor 11001 and memory 11002 via bus 11000. Bus 11000... Figure 11 The connections between other components are shown in thick lines only and are not intended to be limiting. The 11000 bus can be divided into address bus, data bus, control bus, etc., for ease of representation. Figure 11 The term is represented by a single thick line, but this does not imply that there is only one bus or one type of bus. Alternatively, the processor 11001 can also be called a controller; there are no restrictions on the name.

[0218] In this embodiment, the memory 11002 stores instructions executable by at least one processor 11001. By executing the instructions stored in the memory 11002, the at least one processor 11001 can perform the marine ecological environment detection method discussed above. The processor 11001 can implement... Figure 10 The functions of each module in the device are shown. Among them, the processor 11001 is the control center of the device. It can connect to various parts of the entire control device through various interfaces and lines. By running or executing instructions stored in memory 11002 and calling data stored in memory 11002, the various functions of the device and the processing of data are performed, thereby monitoring the device as a whole.

[0219] In one possible design, processor 11001 may include one or more processing units. Processor 11001 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 processor 11001. In some embodiments, processor 11001 and memory 11002 may be implemented on the same chip; in some embodiments, they may also be implemented on separate chips.

[0220] The processor 11001 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the marine ecological environment detection method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0221] Memory 11002, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 11002 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage, magnetic disk, optical disk, etc. Memory 11002 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. Memory 11002 in the embodiments of this application may also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0222] By designing and programming the processor 11001, the code corresponding to the marine ecological environment detection method described in the foregoing embodiments can be embedded into the chip, thereby enabling the chip to execute the code during operation. Figure 2 The steps of the marine ecological environment detection method shown in the embodiment are described. How to design and program the processor 11001 is a technique well-known to those skilled in the art and will not be elaborated here.

[0223] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device. When the program is run on the computer device, it causes the computer device to perform the steps of the above-described method for detecting the marine ecological environment.

[0224] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0225] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0226] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0227] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0228] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting marine ecological environment, characterized in that, The method includes: The target mission scenario is determined, and scenario description information of the target mission scenario is obtained; the scenario description information is used to describe the actual marine environment under the target mission scenario. Acquire various types of marine environmental data collected under the target task scenario; among them, different types of marine environmental data have different perception modalities; Based on the scenario description information and the attribute information of the various types of marine environmental data, a comprehensive quality assessment value of the actual marine environment is obtained. Furthermore, the device computing power level, environmental dynamics, and real-time requirements under the target task scenario are acquired and weighted with preset precision weights, computing power weights, environmental dynamics weights, and real-time weights under each data fusion method to obtain an adaptation score for each data fusion method. A data fusion method is selected based on each adaptation score. The attribute information of the various types of marine environmental data is used to describe the comprehensive quality of the various types of marine environmental data in the target task scenario. Based on the data fusion method described above, the various types of marine environmental data are fused to obtain fused environmental data with multiple modes. The fused environment data is analyzed to obtain the detection results of the target task scene.

2. The method as described in claim 1, characterized in that, After acquiring various types of marine environmental data collected under the target task scenario, and before obtaining the comprehensive quality assessment value of the actual marine environment based on the scenario description information and the attribute information of the various types of marine environmental data, the method further includes: For each type of marine environmental data, the following steps are performed: delete the portion of the marine environmental data whose correlation with the target task scenario does not meet the preset correlation conditions, and obtain updated marine environmental data.

3. The method as described in claim 1, characterized in that, The target task scenario is: to detect hydrological parameters of the marine environment; the actual marine environment described by the scenario description information is: comprehensive hydrological disturbance. The process of obtaining a comprehensive quality assessment value for the actual marine environment based on the scene description information and the attribute information of various marine environmental data, and selecting a data fusion method that matches the comprehensive quality assessment value, includes: For each type of marine environmental data, the following is performed: the quality of the hydrological parameters represented by the hydrological-related data contained in the marine environmental data is used as the attribute information of the marine environmental data; Based on the comprehensive hydrological disturbance and the quality of the hydrological parameters, a corresponding hydrological detection and evaluation value is obtained, and the hydrological detection and evaluation value is used as the comprehensive quality evaluation value. Based on the testing accuracy of the hydrological parameters represented by the hydrological detection and evaluation values, a suitable data fusion method is selected.

4. The method as described in claim 1, characterized in that, The target task scenario is: to identify marine organisms in the marine environment; the scenario description information describes the actual marine environment as: the overall survival status of marine organisms. The process of obtaining a comprehensive quality assessment value for the actual marine environment based on the scene description information and the attribute information of various marine environmental data, and selecting a data fusion method that matches the comprehensive quality assessment value, includes: For each type of marine environmental data, the following steps are performed: the target identifiability represented by the marine organism-related data contained in the marine environmental data is used as the attribute information of the marine environmental data; Based on the overall survival status of marine organisms and the identifiability of the target, a corresponding marine organism identification assessment value is obtained, and the marine organism identification assessment value is used as the overall quality assessment value. Based on the accuracy of marine organism identification as represented by the marine organism identification evaluation value, an appropriate data fusion method is selected.

5. The method as described in claim 1, characterized in that, The target task scenario is: to issue an abnormal early warning for the density of marine organisms in the marine environment; the actual marine environment described by the scenario description information is: the overall survival status of marine organisms. The process of obtaining a comprehensive quality assessment value for the actual marine environment based on the scene description information and the attribute information of various marine environmental data, and selecting a data fusion method that matches the comprehensive quality assessment value, includes: For each type of marine environmental data, the following steps are performed: the credibility of marine organism density represented by the marine organism-related data contained in the marine environmental data is used as the attribute information of the marine environmental data; Based on the overall survival status of marine organisms and the reliability of marine organism density, a marine organism density early warning assessment value is obtained, and the marine organism density early warning assessment value is used as the overall quality assessment value. Based on the accuracy of the early warning judgment represented by the marine organism density early warning assessment value, an appropriate data fusion method is selected.

6. The method according to any one of claims 1-5, characterized in that, The process of fusing various types of marine environmental data based on the aforementioned data fusion method to obtain multimodal fused environmental data includes: Under the target task scenario, obtain the first fusion weight of each type of marine environmental data in the data fusion method; When the data fusion method is data-level fusion, the various types of marine environmental data are aligned in data format and then fused using the corresponding first fusion weight to obtain fused environmental data with multiple modes. When the data fusion method is feature-level fusion, the environmental features of each type of marine environmental data are extracted, and the obtained environmental features are unified in dimension. Then, the corresponding first fusion weight is used to fuse them to obtain fused environmental data with multiple modes.

7. The method according to any one of claims 1-5, characterized in that, The method further includes: When the data fusion method is decision-level fusion, the second fusion weight of each type of marine environmental data under the decision-level fusion method is obtained in the target task scenario; The various types of marine environmental data are analyzed to obtain corresponding sub-detection results; The confidence level of each sub-detection result is evaluated separately to obtain the corresponding confidence level evaluation result; Each sub-detection result whose confidence assessment result meets the preset assessment criteria is taken as a positive sub-detection result, and its corresponding second fusion weight is maintained. Each sub-detection result whose confidence assessment result does not meet the preset assessment criteria is taken as a negative sub-detection result, and its corresponding second fusion weight is reduced and adjusted to obtain the adjusted third fusion weight. Based on the positive sub-detection results and their corresponding second fusion weights, and combined with the negative detection results and their corresponding third fusion weights, the sub-detection results are fused to obtain the corresponding detection results.

8. The method according to any one of claims 1-5, characterized in that, The method is implemented by an object detection model. After obtaining the detection results of the target task scene, it further includes: The target task scenario, the various marine environmental data, and the detection results are used as a new sample data to update the sample dataset, and the real environmental state described by the detection results is used as the sample label of the sample data; each other sample data in the sample dataset includes: an other task scenario, various marine environmental data collected under the other task scenario, and corresponding other detection results; The sample dataset is divided into a training dataset and a test dataset; The detection model to be trained is trained based on the training dataset to obtain an initial detection model; The initial detection model is tested and trained using the test dataset to obtain the target detection model.

9. The method as described in claim 8, characterized in that, If the test accuracy of the target detection model does not reach a preset accuracy threshold, the method further includes: Filter out the task scenarios corresponding to test data whose test results do not match the real environment from the test dataset; For each of the aforementioned task scenarios, the following steps are performed: acquiring various new marine environmental data and corresponding new detection results under the aforementioned task scenario; and adding the task scenario, the various new marine environmental data, and the new detection results as a new sample data to the aforementioned sample dataset, and using the real environmental state described by the new detection results as the sample label of the new sample data. A new training dataset is derived from the sample dataset, and the object detection model is trained based on the new training dataset to obtain a new object detection model.

10. A device for detecting marine ecological environment, characterized in that, The device includes: Acquisition module: used to determine the target task scenario and acquire scenario description information of the target task scenario; the scenario description information is used to describe the actual marine environment under the target task scenario, and to acquire various types of marine environmental data collected under the target task scenario; wherein, different types of marine environmental data have different perception modalities; The processing module is used to obtain a comprehensive quality assessment value of the actual marine environment based on the scenario description information and the attribute information of the various types of marine environmental data; and to obtain the equipment computing power level, environmental dynamics, and real-time requirements under the target task scenario, and to perform weighted fusion with preset accuracy weights, computing power weights, environmental dynamics weights, and real-time weights under each data fusion method to obtain an adaptation score under each data fusion method, and to select a data fusion method based on each adaptation score; the attribute information of the various types of marine environmental data is used to describe the comprehensive quality of the various types of marine environmental data in the target task scenario; and to perform fusion processing on the various types of marine environmental data based on the data fusion method to obtain fused environmental data with multimodality; Parsing module: Parses the fused environment data to obtain the detection results of the target task scene.

11. A computer device, characterized in that, include: Memory, used to store computer programs; A processor, when executing a computer program stored in the memory, implements the method steps of any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, It stores a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of any one of the methods described in claims 1-9.

Images