An artificial intelligence-based three-dimensional multi-dimensional ocean swimming animal monitoring method

Abstract

The application provides a kind of three-dimensional multi-dimensional ocean swimming animal monitoring method based on artificial intelligence, relating to the technical field of ocean swimming animal monitoring, comprising: deploying a temperature-salinity-depth profiler and an acoustic Doppler current profiler, collecting data and configuring environmental parameter data sets, obtaining three-dimensional environmental parameter data sets, and generating a multi-dimensional environmental parameter database. By integrating various ocean fishery resource survey methods and using advanced artificial intelligence methods, three-dimensional multi-dimensional dynamic monitoring of ocean swimming animals is achieved. On the one hand, the number composition and spatio-temporal distribution of ocean swimming animals, especially the vertical distribution characteristics, can be monitored, revealing the relationship between population dynamics and environmental factors. On the other hand, through monitoring of swimming animals, basic data for population resource assessment can be provided, serving the formulation of ocean economic swimming animal catch quotas and the delineation of marine protected areas.

Description

Technical Field

[0001] This invention relates to the field of ocean swimming animal monitoring technology. Specifically, it relates to a three-dimensional, multi-dimensional ocean swimming animal monitoring method based on artificial intelligence. Background Technology

[0002] As an important component of the global marine ecosystem, the oceanic fishery ecosystem has formed unique fishery biogeographical characteristics due to its vast spatial scale and strong environmental heterogeneity. Swimming animals are consumers in the oceanic food chain and play an important role in maintaining the material cycle and energy flow of the ecosystem. They are also an important part of human development and utilization of oceanic fishery resources. The distribution and behavior of oceanic swimming animals are affected by factors such as ocean currents, water temperature, and food environment, and their dynamic changes are also an important indicator for monitoring the oceanic ecosystem.

[0003] Currently, ocean swimming animal monitoring faces multiple limitations in terms of technology and ecological factors, which restricts the accurate understanding of their dynamic changes. For example, traditional monitoring methods such as trawl sampling have a certain degree of sampling bias. The distribution of swimming animals is significantly affected by ocean currents and climate change. Traditional fixed-point monitoring is difficult to reflect the rapid changes in their spatiotemporal dynamics. In addition, the distribution of swimming animals is highly correlated with factors such as plankton, topography, and water temperature, but existing monitoring focuses on single species and lacks systematic observation of the linkage mechanism of "environment-food-organism". Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a three-dimensional, multi-dimensional monitoring method for ocean swimming animals based on artificial intelligence. This method integrates various ocean fisheries resource survey techniques and utilizes advanced artificial intelligence to achieve three-dimensional, multi-dimensional dynamic monitoring of ocean swimming animals. On the one hand, it can monitor the population composition and spatiotemporal distribution of ocean swimming animals, especially their vertical distribution characteristics, revealing their population dynamics and their relationship with environmental factors. On the other hand, monitoring swimming animals can provide basic data for population resource assessment, serving the formulation of fishing quotas for economically important ocean swimming animals and the delineation of marine protected areas.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A three-dimensional, multi-dimensional monitoring method for ocean swimming animals based on artificial intelligence includes the following:

[0007] Step S100: Deploy the temperature, salinity, depth profiler and the acoustic Doppler current profiler, collect data and configure the environmental parameter data set, obtain the three-dimensional environmental parameter data set, and generate a multi-dimensional environmental parameter database.

[0008] Step S200: Deploy the multi-dimensional acquisition device for swimming animals and collect data to obtain a set of data on the species and quantity distribution of swimming animals in the middle and upper layers and a set of multi-dimensional acoustic inversion data, and configure the multi-dimensional swimming animal composition and distribution data structure.

[0009] Step S300: Based on the three-dimensional environmental parameter data set and the multi-dimensional swimming animal composition and distribution data structure, perform data fusion to obtain the fused input data set, configure the random forest regression model and the deep neural network model and train them, configure the multivariate mapping relationship structure between environmental factors and swimming animal distribution, obtain the random forest regression mapping structure and the deep neural network mapping structure, and output the optimal habitat environmental parameter combination and model performance index.

[0010] Step S400: Generate a spatiotemporal distribution map of swimming animal community structure based on random forest regression mapping structure and deep neural network mapping structure, calculate Shannon diversity index and Simpson diversity index, and perform spatial statistical analysis.

[0011] Step S500: Deploy remote sensing satellite platform and UAV survey platform, collect macro-scale environmental parameter data and preprocess it, configure extended area prediction input data set, obtain extended area prediction input data set, predict and calculate the spatial distribution of swimming animals in extended area, identify high probability distribution areas and perform spatiotemporal dynamic analysis.

[0012] As a preferred embodiment of the present invention, step S100 specifically comprises:

[0013] Step S100.1: Deploy the temperature, salinity, and depth profiler and the acoustic Doppler current profiler, and collect the data.

[0014] Within the target ocean area, pre-set survey stations were established, and the CTD (Conductivity, Temperature, Depth) profiler and ADCP (Acoustic Doppler Current Profiler) were fixedly installed on the deck lifting system and the bottom mounting frame of the oceanographic survey vessel, respectively.

[0015] After the oceanographic survey vessel sails to the survey station, the deck lifting system of the oceanographic survey vessel is activated to control the temperature, salinity and depth profiler to be lowered vertically to the set detection depth at a stable rate. During the lowering process, the temperature, salinity and depth parameters of the water body are continuously collected by the integrated temperature sensor, salinity sensor and pressure sensor. The detection depth is set according to the requirements of the marine survey mission.

[0016] The collected environmental parameters are transmitted in real time to the data acquisition and processing module on the oceanographic survey vessel via communication cables, and are recorded and stored according to the corresponding depth coordinates and time tags, thus constructing a continuous set of vertical temperature, salinity, and depth profile data.

[0017] The acoustic Doppler current profiler operates continuously during the voyage of an oceanographic research vessel. It emits multi-beam acoustic signals at a preset frequency downwards, and by receiving the echo frequency shift, it measures the velocity of suspended particles in the water based on the Doppler effect, and then inverts the magnitude of the current velocity and the direction angle of the current in different depth sections.

[0018] The acoustic Doppler current profiler divides the entire water column into multiple equally spaced depth layers and outputs the velocity vector corresponding to each layer, including: velocity magnitude, direction angle, and vertical depth coordinates.

[0019] Step S100.2: Based on the data collected by the temperature, salinity and depth profiler and the acoustic Doppler current profiler, configure the environmental parameter data set, obtain the three-dimensional environmental parameter data set, and generate a multi-dimensional environmental parameter database.

[0020] The temperature, salinity, and depth parameters collected by the temperature, salinity, and depth profiler are converted to a unified spatial reference frame and synchronized with the velocity, direction, and depth parameters collected by the acoustic Doppler current profiler. The data is then structured and encoded in a unified format in the data processing module to construct a three-dimensional environmental parameter data set covering five physical factors: temperature, salinity, depth, velocity, and direction.

[0021] As a preferred embodiment of the present invention, step S200 specifically comprises:

[0022] Step S200.1: Deploy a multi-dimensional collection device for swimming animals, which includes: a surface environment DNA sampling device, a mid-to-upper layer trawl system, and a Simrad EK80 broadband fishery acoustic detection system.

[0023] Data was collected using a surface environment DNA sampling device, a mid-to-upper layer trawl system, and a Simrad EK80 broadband fishery acoustic detection system to obtain a set of data on the species and quantity distribution of mid-to-upper layer swimming animals and a set of multi-dimensional acoustic inversion data.

[0024] Within the target ocean area, multi-dimensional sampling devices for swimming animals are deployed according to the survey stations. At each sampling point, the biological sampling range of the surface, upper middle and multi-depth layers is set, forming a biological sampling grid covering vertical and horizontal dimensions.

[0025] The surface environment DNA sampling device collects 3 to 5 liters of seawater samples at each surface sampling point and immediately filters the samples on-site through a filter membrane with a pore size of 0.45 micrometers to obtain a filter membrane sample rich in the genetic material of swimming animals.

[0026] The filter membrane samples were transported to the laboratory via cold chain preservation, where DNA was extracted using a commercial environmental DNA extraction kit, and the extracted DNA fragments were amplified and sequenced using the Illumina NovaSeq high-throughput sequencing platform.

[0027] The sequencing results were aligned to the reference databases NCBINT or FISH-BOL to obtain the species taxonomic units and corresponding sequence depths of swimming animals, and to generate a dataset of species composition of mid-to-upper-level swimming animals containing species names, taxonomic levels, relative abundance, and sampling point identifiers.

[0028] A mid-level trawl was used for horizontal trawl operations within a set water depth range. A four-piece mid-level trawl was used, with a main size of 916 mesh × 400 mm, a net opening circumference of 366.4 m, and a net body length of 120 m. The net had a single-bag structure, with a large mesh at the net opening and a machine-woven net body. The net had double-leaf panels and was connected by a single hand rope. The towing speed was set at 3.5-4.5 nautical miles per hour, and the towing time was 1 hour.

[0029] The samples of swimming animals caught by trawling were classified, counted and weighed, and the number and mass data of each species were recorded. The trawling depth, sampling latitude and longitude coordinates and sampling time were also marked to form a data set of the species and quantity distribution of swimming animals in the middle and upper layers of the water with three-dimensional labels of survey point, water depth and species.

[0030] The Simrad EK80 broadband fisheries acoustic detection system was controlled to continuously transmit broadband acoustic signals in the 38kHz, 70kHz, 120kHz and 200kHz frequency bands during the navigation of the ocean survey vessel, and the vertical echo scattering intensity was recorded.

[0031] The Simrad EK80 broadband fishery acoustic detection system acquires data at a pulse transmission frequency of 5 times per second. It divides the entire water column into multiple depth segments with a depth resolution of 0.5 meters. The system performs background noise removal and target intensity inversion processing on the raw acoustic echo data. Combined with system calibration data and acoustic scattering characteristic parameters of swimming animal target species, the system inverts the number density value of swimming animals per unit volume at each depth layer, forming a multi-dimensional acoustic inversion data set including depth coordinates, latitude and longitude coordinates, number density, and acoustic frequency band identifiers.

[0032] Step S200.2: Configure the multi-dimensional swimming animal composition and distribution data structure.

[0033] Based on the obtained datasets of species and number distribution of swimming animals in the upper and middle layers and the multidimensional acoustic inversion dataset, the data were processed for format unification, spatial coordinate and depth label standardization, time synchronization and structured coding. A multidimensional swimming animal composition and distribution data structure oriented towards the survey station was constructed, which includes labels for species name, number density, taxonomic level, spatial location and sampling method.

[0034] As a preferred embodiment of the present invention, step S300 specifically comprises:

[0035] Step S300.1: Based on the three-dimensional environmental parameter data set and the multi-dimensional swimming animal composition and distribution data structure, perform data fusion to obtain the fused input data set, and configure the random forest regression model and the deep neural network model.

[0036] Step S300.2: Train the random forest regression model and the deep neural network model based on the fused input data set.

[0037] The random forest regression model and the deep neural network model were trained using a fused input dataset, and the prediction performance of the random forest regression model and the deep neural network model was verified using the five-fold cross-validation method.

[0038] The fused input dataset is randomly divided into five subsets. Four subsets are used to train the random forest regression model and the deep neural network model, and one subset is used to validate the random forest regression model and the deep neural network model. The training and validation are repeated five times, and the model prediction error is recorded for each validation. The process stops when the trained random forest regression model and deep neural network model show stable prediction performance on unseen data.

[0039] Step S300.3: Based on the trained random forest regression model and deep neural network model, configure the multivariate mapping relationship structure between environmental factors and the distribution of swimming animals, obtain the random forest regression mapping structure and the deep neural network mapping structure, and output the optimal combination of habitat environmental parameters and model performance indicators.

[0040] The parameters of the trained random forest regression model and deep neural network model are fixed to obtain the random forest regression mapping structure and the deep neural network mapping structure, respectively.

[0041] The random forest regression mapping structure includes: input environmental factor dimension, number of decision trees, maximum tree depth, feature weight parameters, and prediction error boundary.

[0042] The deep neural network mapping structure includes: input feature dimension, number of hidden layer nodes, weight matrix, bias vector, activation function type, and output species existence probability threshold.

[0043] The optimal combination of habitat environmental parameters and model performance indicators are recorded as: the output file for feature selection and prediction performance evaluation.

[0044] As a preferred embodiment of the present invention, step S400 specifically includes:

[0045] Step S400.1: Generate a spatiotemporal distribution map of swimming animal community structure based on random forest regression mapping structure and deep neural network mapping structure.

[0046] By inputting the environmental parameters within the spatial analysis unit into the random forest regression mapping structure and the deep neural network mapping structure, the resource density value and the species existence probability value of swimming animals in each spatial analysis unit are predicted.

[0047] The predicted resource density values ​​and species presence probability values ​​obtained in each spatial analysis unit are spatially interpolated and smoothed to output spatial distribution maps of swimming animal population density and species presence probability in the target ocean area.

[0048] Step S400.2: Calculate the Shannon diversity index and Simpson diversity index based on the spatiotemporal distribution map of swimming animal community structure, obtain the spatial distribution map of Shannon diversity index and Simpson diversity index, and perform spatial statistical analysis.

[0049] Spatial statistical analysis was conducted on the spatiotemporal distribution characteristics and changing trends of high-diversity and low-diversity regions. The spatial diffusion characteristics of high-diversity and low-diversity regions over time were analyzed, and the main environmental factors leading to changes in high-diversity and low-diversity regions were analyzed based on temperature, salinity, depth, velocity, and direction parameters.

[0050] The analysis results file includes: spatiotemporal distribution maps of Shannon and Simpson diversity indices, spatial statistical analysis data tables, and trend change maps, comprehensively characterizing the spatial distribution characteristics, spatiotemporal dynamic change trends, community stability, and ecosystem diversity of swimming animal communities in the target ocean area.

[0051] As a preferred embodiment of the present invention, step S500 specifically comprises:

[0052] Step S500.1: Deploy remote sensing satellite platform and UAV survey platform, and collect macro-scale environmental parameter data.

[0053] In the extended observation area beyond the target ocean area, remote sensing satellite platforms and unmanned aerial vehicle (UAV) survey platforms will be deployed. The remote sensing satellite platforms include: MODIS remote sensing satellite, Sentinel-3 remote sensing satellite, and VIIRS remote sensing satellite.

[0054] The drone survey platform includes: multi-rotor drones and fixed-wing drones.

[0055] The remote sensing satellite platform conducts remote sensing observations of sea surface temperature, sea surface chlorophyll concentration, and sea surface wind speed parameters within the extended observation area with a spatial resolution of 1 km × 1 km and a sampling period of 12 hours, thereby obtaining macro-scale environmental parameter data.

[0056] The UAV survey platform, within the extended observation area, surveys the sea area along a pre-planned gridded survey route at a survey flight altitude of 300 meters, obtaining visible light images of the sea surface with a spatial resolution of no less than 10 meters, infrared temperature maps, local wind speed parameters, and local wind direction parameters, supplementing the acquisition of high spatial accuracy local environmental parameter data.

[0057] Step S500.2: Perform data preprocessing based on macro-scale environmental parameter data, and configure the extended region prediction input data set to obtain the extended region prediction input data set.

[0058] The macro-scale environmental parameter data obtained from remote sensing satellite platforms and the local environmental parameter data obtained from UAV survey platforms are processed by spatial coordinate reprojection, time synchronization resampling, and data format unification.

[0059] The sea surface temperature parameters, sea surface chlorophyll concentration parameters, sea surface wind speed parameters, local wind speed parameters, and local wind direction parameters, which have undergone spatial coordinate reprojection processing, are spatially interpolated and edge-stitched with the three-dimensional environmental parameter data set in the original survey area to obtain the extended area prediction input data set.

[0060] Step S500.3: Perform prediction calculations based on the extended region prediction input data set to generate an extended region swimming animal spatial distribution map and identify high-probability distribution areas.

[0061] The extended region prediction input dataset is fed into a random forest regression mapping structure and a deep neural network mapping structure to predict the density of swimming animals and the probability of their existence within the extended region.

[0062] The random forest regression mapping structure outputs the swimming animal resource density value in each spatial analysis unit of the extended region, while the deep neural network mapping structure outputs the swimming animal species existence probability value in each spatial analysis unit of the extended region.

[0063] The predicted swimming animal resource density values ​​and swimming animal species existence probability values ​​within each spatial analysis unit of the extended region are stored in a gridded data structure and labeled with the corresponding prediction timestamp, latitude and longitude spatial coordinates, and depth level label.

[0064] Spatial interpolation and smoothing are performed on the predicted density values ​​and species probability values ​​of swimming animals in the extended area to generate spatial distribution maps of swimming animal population density and species probability in the extended area.

[0065] Based on the prediction calculation results, the threshold for the density value of swimming animals is defined as 10 animals / cubic meter, and the threshold for the probability of the existence of swimming animal species is defined as 0.6. Spatial analysis units with a resource density value greater than 10 animals / cubic meter and a species existence probability value greater than 0.6 are marked as high probability distribution areas.

[0066] Output a spatial distribution map of swimming animals in the extended region, a mask map of high-probability distribution areas, a data table of swimming animal resource density values ​​and swimming animal species existence probability values.

[0067] Step S500.4: Perform spatiotemporal dynamic analysis based on the spatial distribution map of swimming animals in the extended region, and output a large-scale prediction result file.

[0068] Based on the spatial distribution map of swimming animals in the extended region and the mask map of high probability distribution areas, spatial statistical analysis tools were used to analyze the spatial migration characteristics of high probability distribution areas under different prediction timestamps.

[0069] The spatial distribution maps of swimming animal population density, spatial distribution maps of swimming animal species probability, mask maps of high-probability distribution areas, and data tables of swimming animal resource density and species probability values ​​within the extended area are summarized to generate an extended area prediction result file that includes: spatiotemporal dynamic analysis maps and trend change analysis.

[0070] Compared with the prior art, the beneficial effects of the present invention are:

[0071] 1. By deploying a temperature, salinity, and depth profiler and an acoustic Doppler current profiler, and combining this with the spatial mobility characteristics of an oceanographic survey vessel, a three-dimensional environmental parameter dataset covering both vertical and horizontal dimensions was systematically constructed. This dataset not only covers key physical factors such as temperature, salinity, depth, current velocity, and current direction, but also ensures the consistency of the data across spatial reference frames and time axes, effectively improving the applicability and accuracy of environmental factor data in modeling the distribution of swimming animals.

[0072] 2. By integrating surface environmental DNA sampling, trawl sampling, and broadband fisheries acoustic detection, the system acquired multi-dimensional composition and distribution information of swimming animals in the surface and upper layers of the water, and constructed a unified data structure. This approach comprehensively integrates biological information at both macro and micro scales, significantly enhancing the ability to characterize species diversity, spatial distribution, and population density, and solving the problems of single data dimensions and missing distribution information in traditional monitoring methods.

[0073] 3. By constructing a random forest regression model and a deep neural network model that integrate three-dimensional environmental parameters and multi-dimensional distribution data of swimming animals, and training them through multiple rounds of cross-validation, a stable and efficient prediction mapping structure was established. This mapping structure not only improves the accuracy of predicting the spatial distribution of swimming animals in the target area, but also outputs the optimal combination of environmental factors for habitats based on feature importance scores, providing data support for environmental regulation and resource assessment. Attached Figure Description

[0074] Figure 1 A flowchart illustrating a three-dimensional, multi-dimensional ocean swimming animal monitoring method based on artificial intelligence, provided for embodiments of this application. Detailed Implementation

[0075] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0076] Please see Figure 1 , Figure 1 This application provides a flowchart of a three-dimensional, multi-dimensional ocean swimming animal monitoring method based on artificial intelligence.

[0077] In this embodiment, a three-dimensional, multi-dimensional ocean swimming animal monitoring method based on artificial intelligence may include steps S100, S200, S300, S400 and S500.

[0078] Step S100: Deploy the temperature, salinity, depth profiler and the acoustic Doppler current profiler, collect data and configure the environmental parameter data set, obtain the three-dimensional environmental parameter data set, and generate a multi-dimensional environmental parameter database.

[0079] Step S200: Deploy the multi-dimensional acquisition device for swimming animals and collect data to obtain a set of data on the species and quantity distribution of swimming animals in the middle and upper layers and a set of multi-dimensional acoustic inversion data, and configure the multi-dimensional swimming animal composition and distribution data structure.

[0080] Step S300: Based on the three-dimensional environmental parameter data set and the multi-dimensional swimming animal composition and distribution data structure, perform data fusion to obtain the fused input data set, configure the random forest regression model and the deep neural network model and train them, configure the multivariate mapping relationship structure between environmental factors and swimming animal distribution, obtain the random forest regression mapping structure and the deep neural network mapping structure, and output the optimal habitat environmental parameter combination and model performance index.

[0081] Step S400: Generate a spatiotemporal distribution map of swimming animal community structure based on random forest regression mapping structure and deep neural network mapping structure, calculate Shannon diversity index and Simpson diversity index, and perform spatial statistical analysis.

[0082] Step S500: Deploy remote sensing satellite platform and UAV survey platform, collect macro-scale environmental parameter data and preprocess it, configure extended area prediction input data set, obtain extended area prediction input data set, predict and calculate the spatial distribution of swimming animals in extended area, identify high probability distribution areas and perform spatiotemporal dynamic analysis.

[0083] In some specific embodiments, step S100 specifically includes:

[0084] Step S100.1: Deploy the temperature, salinity, and depth profiler and the acoustic Doppler current profiler, and collect the data.

[0085] Within the target ocean area, pre-set survey stations were established, and the CTD (Conductivity, Temperature, Depth) profiler and ADCP (Acoustic Doppler Current Profiler) were fixedly installed on the deck lifting system and the bottom mounting frame of the oceanographic survey vessel, respectively.

[0086] After the oceanographic survey vessel sails to the survey station, the deck lifting system of the oceanographic survey vessel is activated to control the temperature, salinity and depth profiler to be lowered vertically to the set detection depth at a stable rate. During the lowering process, the temperature, salinity and depth parameters of the water body are continuously collected by the integrated temperature sensor, salinity sensor and pressure sensor. The detection depth is set according to the requirements of the marine survey mission.

[0087] The collected environmental parameters are transmitted in real time to the data acquisition and processing module on the oceanographic survey vessel via communication cables, and are recorded and stored according to the corresponding depth coordinates and time tags, thus constructing a continuous set of vertical temperature, salinity, and depth profile data.

[0088] The acoustic Doppler current profiler operates continuously during the voyage of an oceanographic research vessel. It emits multi-beam acoustic signals at a preset frequency downwards, and by receiving the echo frequency shift, it measures the velocity of suspended particles in the water based on the Doppler effect, and then inverts the magnitude of the current velocity and the direction angle of the current in different depth sections.

[0089] The acoustic Doppler current profiler divides the entire water column into multiple equally spaced depth layers and outputs the velocity vector corresponding to each layer, including: velocity magnitude, direction angle, and vertical depth coordinates.

[0090] Step S100.2: Based on the data collected by the temperature, salinity and depth profiler and the acoustic Doppler current profiler, configure the environmental parameter data set, obtain the three-dimensional environmental parameter data set, and generate a multi-dimensional environmental parameter database.

[0091] The temperature, salinity, and depth parameters collected by the temperature, salinity, and depth profiler are converted to a unified spatial reference frame and synchronized with the velocity, direction, and depth parameters collected by the acoustic Doppler current profiler. The data is then structured and encoded in a unified format in the data processing module to construct a three-dimensional environmental parameter data set covering five physical factors: temperature, salinity, depth, velocity, and direction.

[0092] It should be noted that since unified spatial reference frame conversion, time synchronization processing, and unified format structured coding are technologies well known to those skilled in the art, they will not be elaborated upon further here.

[0093] The three-dimensional environmental parameter data set is archived by survey station and labeled with multi-dimensional tags according to time series, depth level and geographic coordinates to generate a multi-dimensional environmental parameter database for the target sea area.

[0094] In some specific embodiments, step S200 specifically includes:

[0095] Step S200.1: Deploy a multi-dimensional collection device for swimming animals, which includes: a surface environment DNA sampling device, a mid-to-upper layer trawl system, and a Simrad EK80 broadband fishery acoustic detection system.

[0096] Data was collected using a surface environment DNA sampling device, a mid-to-upper layer trawl system, and a Simrad EK80 broadband fishery acoustic detection system to obtain a set of data on the species and quantity distribution of mid-to-upper layer swimming animals and a set of multi-dimensional acoustic inversion data.

[0097] Within the target ocean area, multi-dimensional sampling devices for swimming animals are deployed according to the survey stations. At each sampling point, the biological sampling range of the surface, upper middle and multi-depth layers is set, forming a biological sampling grid covering vertical and horizontal dimensions.

[0098] The surface environment DNA sampling device collects 3 to 5 liters of seawater samples at each surface sampling point and immediately filters the samples on-site through a filter membrane with a pore size of 0.45 micrometers to obtain a filter membrane sample rich in the genetic material of swimming animals.

[0099] The filter membrane samples were transported to the laboratory via cold chain preservation, where DNA was extracted using a commercial environmental DNA extraction kit, and the extracted DNA fragments were amplified and sequenced using the Illumina NovaSeq high-throughput sequencing platform.

[0100] The sequencing results were aligned to the reference databases NCBINT or FISH-BOL to obtain the species taxonomic units and corresponding sequence depths of swimming animals, and to generate a dataset of species composition of mid-to-upper-level swimming animals containing species names, taxonomic levels, relative abundance, and sampling point identifiers.

[0101] A mid-level trawl was used for horizontal trawl operations within a set water depth range. A four-piece mid-level trawl was used, with a main size of 916 mesh × 400 mm, a net opening circumference of 366.4 m, and a net body length of 120 m. The net had a single-bag structure, with a large mesh at the net opening and a machine-woven net body. The net had double-leaf panels and was connected by a single hand rope. The towing speed was set at 3.5-4.5 nautical miles per hour, and the towing time was 1 hour.

[0102] The samples of swimming animals caught by trawling were classified, counted and weighed, and the number and mass data of each species were recorded. The trawling depth, sampling latitude and longitude coordinates and sampling time were also marked to form a data set of the species and quantity distribution of swimming animals in the middle and upper layers of the water with three-dimensional labels of survey point, water depth and species.

[0103] The Simrad EK80 broadband fisheries acoustic detection system was controlled to continuously transmit broadband acoustic signals in the 38kHz, 70kHz, 120kHz and 200kHz frequency bands during the navigation of the ocean survey vessel, and the vertical echo scattering intensity was recorded.

[0104] The Simrad EK80 broadband fishery acoustic detection system acquires data at a pulse transmission frequency of 5 times per second. It divides the entire water column into multiple depth segments with a depth resolution of 0.5 meters. The system performs background noise removal and target intensity inversion processing on the raw acoustic echo data. Combined with system calibration data and acoustic scattering characteristic parameters of swimming animal target species, the system inverts the number density value of swimming animals per unit volume at each depth layer, forming a multi-dimensional acoustic inversion data set including depth coordinates, latitude and longitude coordinates, number density, and acoustic frequency band identifiers.

[0105] Step S200.2: Configure the multi-dimensional swimming animal composition and distribution data structure.

[0106] Based on the obtained datasets of species and number distribution of swimming animals in the upper and middle layers and the multidimensional acoustic inversion dataset, the data were processed for format unification, spatial coordinate and depth label standardization, time synchronization and structured coding. A multidimensional swimming animal composition and distribution data structure oriented towards the survey station was constructed, which includes labels for species name, number density, taxonomic level, spatial location and sampling method.

[0107] It should be noted that since format unification, spatial coordinates, depth label standardization, time synchronization, and structured coding are already well-known technologies in this field, they will not be elaborated upon further here.

[0108] In some specific embodiments, step S300 specifically includes:

[0109] Step S300.1: Based on the three-dimensional environmental parameter data set and the multi-dimensional swimming animal composition and distribution data structure, perform data fusion to obtain the fused input data set, and configure the random forest regression model and the deep neural network model.

[0110] The temperature, salinity, depth, flow velocity, and flow direction data from the three-dimensional environmental parameter dataset are fused with the species name, resource density, taxonomic level, spatial location, and sampling method labels from the multi-dimensional swimming animal composition and distribution data structure. During the fusion process, the latitude and longitude coordinates, depth level, and sampling time of the survey station are used as unique indexes to match the environmental parameter data with the corresponding swimming animal distribution data, and the fused input dataset is obtained.

[0111] Based on the obtained fused input dataset, a random forest regression model and a deep neural network model are constructed respectively. The input feature vector X of the random forest regression model is: Where T is the temperature parameter, S is the salinity parameter, D is the depth parameter, and V is the flow velocity parameter. The flow direction parameter is used to output the density of swimming animals at the corresponding location. The model uses an error optimization objective function:

[0112] ;

[0113] In the formula: MSE is the mean squared error loss function of the random forest regression model. It is the true value of the observed swimming animal resource density. is the predicted swimming animal resource density value by the random forest regression model, and n is the total number of training samples for the random forest regression model.

[0114] The input feature vector of the deep neural network model is consistent with that of the random forest regression model. The network structure is set as follows: the input layer has 5 nodes, the first and second hidden layers each have 128 nodes and use the ReLU activation function, and the output layer uses the Sigmoid function to output the probability value of the species' existence. The deep neural network model training process uses a binary cross-entropy loss function for optimization, specifically:

[0115] ;

[0116] In the formula: It is a binary cross-loss function for deep neural network models. The observed species has a true marker value, which can be either 0 or 1. is the probability of species existence predicted by the deep neural network model, ranging from 0 to 1, and m is the total number of training samples for the deep neural network model.

[0117] Step S300.2: Train the random forest regression model and the deep neural network model based on the fused input data set.

[0118] The random forest regression model and the deep neural network model were trained using a fused input dataset, and the prediction performance of the random forest regression model and the deep neural network model was verified using the five-fold cross-validation method.

[0119] The fused input dataset is randomly divided into five subsets. Four subsets are used to train the random forest regression model and the deep neural network model, and one subset is used to validate the random forest regression model and the deep neural network model. The training and validation are repeated five times, and the model prediction error is recorded for each validation. The process stops when the trained random forest regression model and deep neural network model show stable prediction performance on unseen data.

[0120] Step S300.3: Based on the trained random forest regression model and deep neural network model, configure the multivariate mapping relationship structure between environmental factors and the distribution of swimming animals, obtain the random forest regression mapping structure and the deep neural network mapping structure, and output the optimal combination of habitat environmental parameters and model performance indicators.

[0121] The parameters of the trained random forest regression model and deep neural network model are fixed to obtain the random forest regression mapping structure and the deep neural network mapping structure, respectively.

[0122] The random forest regression mapping structure includes: input environmental factor dimension, number of decision trees, maximum tree depth, feature weight parameters, and prediction error boundary.

[0123] The deep neural network mapping structure includes: input feature dimension, number of hidden layer nodes, weight matrix, bias vector, activation function type, and output species existence probability threshold.

[0124] Random forest regression mapping structure and deep neural network mapping structure jointly characterize the spatial response characteristics of the number density and species probability of swimming animals under different combinations of environmental parameters.

[0125] Feature importance scores for temperature, salinity, depth, flow velocity, and flow direction parameters are extracted from the random forest regression mapping structure. These scores are then sorted in descending order to form the optimal combination of habitat environmental parameters. The model's accuracy, recall, and F1 score are output from the deep neural network mapping structure. The specific performance metrics are calculated as follows:

[0126] ;

[0127] ;

[0128] ;

[0129] In the formula: Accuracy is the prediction accuracy in the performance metrics of deep neural network models; TP is the number of samples in the model prediction results where the true class is positive and the model predicts it as positive; TN is the number of samples in the model prediction results where the true class is negative and the model predicts it as negative; FP is the number of samples in the model prediction results where the true class is negative but the model predicts it as positive; FN is the number of samples in the model prediction results where the true class is positive but the model predicts it as negative; Recall is the recall rate in the performance metrics of deep neural network models; F1-score is the overall evaluation score in the performance metrics of deep neural network models; and Precision is the precision rate in the model prediction results, that is, the proportion of samples that the model predicts as positive but are actually positive.

[0130] The optimal combination of habitat environmental parameters and model performance indicators are recorded as: the output file for feature selection and prediction performance evaluation.

[0131] In some specific embodiments, step S400 specifically includes:

[0132] Step S400.1: Generate a spatiotemporal distribution map of swimming animal community structure based on random forest regression mapping structure and deep neural network mapping structure.

[0133] Based on the random forest regression mapping structure and the deep neural network mapping structure, the target ocean area is divided into horizontal grids with a latitude and longitude scale of 1°×1°, and multiple vertical water layers are divided at 500-meter intervals along the vertical direction, forming a spatial analysis unit where the horizontal grid and vertical water layers intersect. The environmental parameters in each spatial analysis unit are composed of temperature parameters, salinity parameters, depth parameters, current velocity parameters and current direction parameters.

[0134] By inputting the environmental parameters within the spatial analysis unit into the random forest regression mapping structure and the deep neural network mapping structure, the resource density value and the species existence probability value of swimming animals in each spatial analysis unit are predicted.

[0135] The predicted resource density values ​​and species presence probability values ​​obtained in each spatial analysis unit are spatially interpolated and smoothed to output spatial distribution maps of swimming animal population density and species presence probability in the target ocean area.

[0136] Step S400.2: Calculate the Shannon diversity index and Simpson diversity index based on the spatiotemporal distribution map of swimming animal community structure, obtain the spatial distribution map of Shannon diversity index and Simpson diversity index, and perform spatial statistical analysis.

[0137] Based on the generally accepted standards in the fields of current ecology and environmental biological monitoring, areas with a Shannon diversity index value >3 and a Simpson diversity index value >0.7 are defined as high diversity areas, while areas with a Shannon diversity index value <1 and a Simpson diversity index value <0.3 are defined as low diversity areas.

[0138] Spatial statistical analysis was conducted on the spatiotemporal distribution characteristics and changing trends of high-diversity and low-diversity regions. The spatial diffusion characteristics of high-diversity and low-diversity regions over time were analyzed, and the main environmental factors leading to changes in high-diversity and low-diversity regions were analyzed based on temperature, salinity, depth, velocity, and direction parameters.

[0139] The analysis results file includes: spatiotemporal distribution maps of Shannon and Simpson diversity indices, spatial statistical analysis data tables, and trend change maps, comprehensively characterizing the spatial distribution characteristics, spatiotemporal dynamic change trends, community stability, and ecosystem diversity of swimming animal communities in the target ocean area.

[0140] In some specific embodiments, step S500 specifically includes:

[0141] Step S500.1: Deploy remote sensing satellite platform and UAV survey platform, and collect macro-scale environmental parameter data.

[0142] In the extended observation area beyond the target ocean area, remote sensing satellite platforms and unmanned aerial vehicle (UAV) survey platforms will be deployed. The remote sensing satellite platforms include: MODIS remote sensing satellite, Sentinel-3 remote sensing satellite, and VIIRS remote sensing satellite.

[0143] The drone survey platform includes: multi-rotor drones and fixed-wing drones.

[0144] The remote sensing satellite platform conducts remote sensing observations of sea surface temperature, sea surface chlorophyll concentration, and sea surface wind speed parameters within the extended observation area with a spatial resolution of 1 km × 1 km and a sampling period of 12 hours, thereby obtaining macro-scale environmental parameter data.

[0145] The UAV survey platform, within the extended observation area, surveys the sea area along a pre-planned gridded survey route at a survey flight altitude of 300 meters, obtaining visible light images of the sea surface with a spatial resolution of no less than 10 meters, infrared temperature maps, local wind speed parameters, and local wind direction parameters, supplementing the acquisition of high spatial accuracy local environmental parameter data.

[0146] Step S500.2: Perform data preprocessing based on macro-scale environmental parameter data, and configure the extended region prediction input data set to obtain the extended region prediction input data set.

[0147] The macro-scale environmental parameter data obtained from remote sensing satellite platforms and the local environmental parameter data obtained from UAV survey platforms are processed by spatial coordinate reprojection, time synchronization resampling, and data format unification.

[0148] The sea surface temperature parameters, sea surface chlorophyll concentration parameters, sea surface wind speed parameters, local wind speed parameters, and local wind direction parameters, which have undergone spatial coordinate reprojection processing, are spatially interpolated and edge-stitched with the three-dimensional environmental parameter data set in the original survey area to obtain the extended area prediction input data set.

[0149] Step S500.3: Perform prediction calculations based on the extended region prediction input data set to generate an extended region swimming animal spatial distribution map and identify high-probability distribution areas.

[0150] The extended region prediction input dataset is fed into a random forest regression mapping structure and a deep neural network mapping structure to predict the density of swimming animals and the probability of their existence within the extended region.

[0151] The random forest regression mapping structure outputs the swimming animal resource density value in each spatial analysis unit of the extended region, while the deep neural network mapping structure outputs the swimming animal species existence probability value in each spatial analysis unit of the extended region.

[0152] The predicted swimming animal resource density values ​​and swimming animal species existence probability values ​​within each spatial analysis unit of the extended region are stored in a gridded data structure and labeled with the corresponding prediction timestamp, latitude and longitude spatial coordinates, and depth level label.

[0153] Spatial interpolation and smoothing are performed on the predicted density values ​​and species probability values ​​of swimming animals in the extended area to generate spatial distribution maps of swimming animal population density and species probability in the extended area.

[0154] Based on the prediction calculation results, the threshold for the density value of swimming animals is defined as 10 animals / cubic meter, and the threshold for the probability of the existence of swimming animal species is defined as 0.6. Spatial analysis units with a resource density value greater than 10 animals / cubic meter and a species existence probability value greater than 0.6 are marked as high probability distribution areas.

[0155] Output a spatial distribution map of swimming animals in the extended region, a mask map of high-probability distribution areas, a data table of swimming animal resource density values ​​and swimming animal species existence probability values.

[0156] Step S500.4: Perform spatiotemporal dynamic analysis based on the spatial distribution map of swimming animals in the extended region, and output a large-scale prediction result file.

[0157] Based on the spatial distribution map of swimming animals in the extended region and the mask map of high probability distribution areas, spatial statistical analysis tools were used to analyze the spatial migration characteristics of high probability distribution areas under different prediction timestamps.

[0158] The spatial distribution maps of swimming animal population density, spatial distribution maps of swimming animal species probability, mask maps of high-probability distribution areas, and data tables of swimming animal resource density and species probability values ​​within the extended area are summarized to generate an extended area prediction result file that includes: spatiotemporal dynamic analysis maps and trend change analysis.

[0159] In practical application, the above involves first setting up multiple survey stations within the target ocean area. The CTD (Conductivity, Temperature, Depth) profiler is fixedly mounted on the deck hoisting system of the oceanographic research vessel, while the ADCP (Acoustic Doppler Current Profiler) is fixedly mounted on a mounting frame at the bottom of the hull. Once the research vessel arrives at the designated survey station, the deck hoisting system is activated to lower the CTD vertically to the detection depth at a stable rate. During the lowering process, temperature, salinity, and depth parameters of the water body are collected in real time using temperature, salinity, and pressure sensors. The ADCP continuously emits acoustic signals and measures the velocity of suspended particles in the water during the research vessel's voyage, obtaining data on the magnitude and direction of currents at different depths.

[0160] Subsequently, the temperature, salinity, and depth data obtained by the temperature, salinity, and depth profiler and the velocity, direction, and depth data obtained by the acoustic Doppler current profiler are subjected to unified spatial reference system conversion and time synchronization processing to construct a three-dimensional environmental parameter data set covering temperature, salinity, depth, velocity, and direction parameters. The data is then archived by survey station to form a multi-dimensional environmental parameter database.

[0161] Subsequently, a multi-dimensional data collection device for swimming animals was deployed at each survey station, including a surface environmental DNA sampling device, a mid-to-upper layer trawl system, and a Simrad EK80 broadband fisheries acoustic detection system. The surface environmental DNA sampling device collected seawater and obtained filter membrane samples after on-site filtration. DNA extraction and sequencing were used to obtain species composition data of mid-to-upper layer swimming animals. The mid-to-upper layer trawl system carried out horizontal trawl operations and caught swimming animals. The captured swimming animals were classified, counted, and weighed to form data on the species and quantity distribution of mid-to-upper layer swimming animals. The Simrad EK80 broadband fisheries acoustic detection system emitted sound wave signals of different frequency bands. The quantity density of swimming animals at each depth layer was obtained through target intensity inversion, forming a multi-dimensional acoustic inversion data set.

[0162] Next, data fusion is performed based on the three-dimensional environmental parameter dataset and the multi-dimensional swimming animal composition and distribution data structure to obtain the fused input dataset. The fused input dataset is then used to construct a random forest regression model and a deep neural network model, respectively, and trained. The five-fold cross-validation method is used to verify the predictive performance of the models. The trained random forest regression mapping structure and deep neural network mapping structure are obtained, the optimal combination of habitat environmental parameters is output, and the model's performance indicators, including the model's prediction accuracy, recall, and F1 score, are calculated to form a feature selection and prediction performance evaluation output file.

[0163] Then, based on the random forest regression mapping structure and the deep neural network mapping structure, the target sea area is divided into horizontal and vertical grids to obtain the swimming animal resource density and species presence probability data in each spatial analysis unit. Spatial interpolation and smoothing are then performed to generate spatial distribution maps of swimming animal population density and species presence probability. The Shannon diversity index and Simpson diversity index of each spatial analysis unit are calculated, and the corresponding spatial distribution maps are output. High-diversity areas and low-diversity areas are defined and identified, and spatial statistical analysis is performed to generate analysis result files including trend change maps.

[0164] Finally, MODIS, Sentinel-3, and VIIRS remote sensing satellites, along with multi-rotor and fixed-wing UAV survey platforms, were deployed in the extended observation area of ​​the target ocean region to collect macro-scale environmental parameter data. Spatial coordinate reprojection and temporal synchronization processing were then performed to obtain the extended region prediction input dataset. This dataset was then input into a random forest regression mapping structure and a deep neural network mapping structure for prediction calculations, generating a spatial distribution map of swimming animals in the extended region and identifying high-probability distribution areas. Combined with remote sensing satellite and UAV survey data, spatiotemporal dynamic analysis of the spatial distribution of swimming animals within the extended region was conducted, ultimately generating an extended region prediction result file containing dynamic analysis maps and trend change analysis.

[0165] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A three-dimensional, multi-dimensional monitoring method for ocean swimming animals based on artificial intelligence, characterized in that, Includes the following steps: The S100 system deploys a temperature, salinity, and depth profiler and an acoustic Doppler current profiler, collects data, configures environmental parameter datasets, obtains a three-dimensional environmental parameter dataset, and generates a multi-dimensional environmental parameter database. S200: Deploy a multi-dimensional acquisition device for swimming animals and collect data to obtain a set of data on the species and quantity distribution of swimming animals in the middle and upper layers and a set of multi-dimensional acoustic inversion data, and configure a multi-dimensional data structure for the composition and distribution of swimming animals. S300: Based on the three-dimensional environmental parameter data set and the multi-dimensional swimming animal composition and distribution data structure, data fusion is performed to obtain the fused input data set. Random forest regression model and deep neural network model are configured and trained. The multivariate mapping relationship structure between environmental factors and swimming animal distribution is configured to obtain the random forest regression mapping structure and the deep neural network mapping structure. The optimal combination of habitat environmental parameters and model performance indicators are output. S400: Based on the random forest regression mapping structure and the deep neural network mapping structure, a spatiotemporal distribution map of swimming animal community structure is generated, the Shannon diversity index and the Simpson diversity index are calculated, and spatial statistical analysis is performed. The S500 system deploys remote sensing satellite platforms and UAV survey platforms, collects and preprocesses macro-scale environmental parameter data, configures extended area prediction input data sets, obtains extended area prediction input data sets, predicts and calculates the spatial distribution of swimming animals in the extended area, identifies high-probability distribution areas, and performs spatiotemporal dynamic analysis.

2. The method for monitoring ocean swimming animals in a three-dimensional, multi-dimensional manner based on artificial intelligence as described in claim 1, characterized in that, Specifically, S100 is as follows: S100.1 Deploy the temperature, salinity, and depth profiler and the acoustic Doppler current profiler, and collect the data; Within the target ocean area, pre-set survey stations were established, and the CTD (Conductivity, Temperature, Depth) profiler and ADCP (Acoustic Doppler Current Profiler) were fixedly installed on the deck lifting system and the bottom mounting frame of the oceanographic survey vessel, respectively. After the oceanographic survey vessel sails to the survey station, the deck lifting system of the oceanographic survey vessel is activated to control the temperature, salinity and depth profiler to be lowered vertically to the set detection depth at a stable rate. During the lowering process, the temperature, salinity and depth parameters of the water body are continuously collected by the integrated temperature sensor, salinity sensor and pressure sensor. The detection depth is set according to the requirements of the marine survey mission. The collected environmental parameters are transmitted in real time to the data acquisition and processing module on the oceanographic survey vessel via communication cables, and are recorded and stored according to the corresponding depth coordinates and time tags, and a continuous set of vertical temperature, salinity and depth profile data is constructed. The acoustic Doppler current profiler works continuously during the voyage of the oceanographic research vessel. It uses a multi-beam acoustic signal of a preset frequency to be emitted downwards. By receiving the echo frequency shift, it determines the velocity of suspended particles in the water body based on the Doppler effect and inverts the magnitude of the current velocity and the angle of the current direction in different depth sections. The acoustic Doppler current profiler divides the entire water column into multiple equally spaced depth layers and outputs the velocity vector corresponding to each layer, including: velocity magnitude, direction angle, and vertical depth coordinates. S100.

2. Based on the data collected by the temperature, salinity and depth profiler and the acoustic Doppler current profiler, configure the environmental parameter data set, obtain the three-dimensional environmental parameter data set, and generate a multi-dimensional environmental parameter database. The temperature, salinity, and depth parameters collected by the temperature, salinity, and depth profiler are converted to a unified spatial reference frame and synchronized with the velocity, direction, and depth parameters collected by the acoustic Doppler current profiler. The data is then structured and encoded in a unified format in the data processing module to construct a three-dimensional environmental parameter data set covering five physical factors: temperature, salinity, depth, velocity, and direction.

3. The method for monitoring ocean swimming animals based on artificial intelligence in a three-dimensional, multi-dimensional manner as described in claim 1, characterized in that, Specifically, S200 is as follows: S200.1 Deploy a multi-dimensional collection device for swimming animals, which includes: a surface environment DNA sampling device, a mid-to-upper layer trawl system, and a Simrad EK80 broadband fishery acoustic detection system; Data was collected using a surface environment DNA sampling device, a mid-to-upper layer trawl system, and a Simrad EK80 broadband fishery acoustic detection system to obtain a set of data on the species and quantity distribution of mid-to-upper layer swimming animals and a set of multi-dimensional acoustic inversion data. Within the target ocean area, multi-dimensional collection devices for swimming animals are deployed according to the survey stations. At each sampling point, the biological sampling range of the surface, middle and upper layers and multiple depths is set, forming a biological sampling grid covering vertical and horizontal dimensions. The surface environment DNA sampling device collects 3 to 5 liters of seawater samples at each surface sampling point and immediately filters the samples on-site through a filter membrane with a pore size of 0.45 micrometers to obtain filter membrane samples rich in the genetic material of swimming animals. The filter membrane samples were transported to the laboratory via cold chain preservation, where DNA was extracted using a commercial environmental DNA extraction kit, and the extracted DNA fragments were amplified and sequenced using the Illumina NovaSeq high-throughput sequencing platform. The sequencing results were compared with the reference databases NCBINT and FISH-BOL to obtain the species taxonomic units and corresponding sequence depths of swimming animals, and a dataset of species composition of mid-to-upper-level swimming animals containing species names, taxonomic levels, relative abundance and sampling point identifiers was generated. A mid-level trawl net is used for horizontal trawl operations within a set water depth range. A four-piece mid-level trawl net is used, with a main size of 916 mesh × 400 mm, a net opening circumference of 366.4 m, and a net body length of 120 m. The net adopts a single-bag structure, with a large mesh at the net opening and a machine-woven net body. The double-leaf mesh panel uses a single-handle connection method, with a towing speed of 3.5-4.5 nautical miles per hour and a towing time of 1 hour. The samples of swimming animals caught by trawling devices are classified, counted and weighed, and the number and mass data of each species are recorded. The trawling depth, sampling latitude and longitude coordinates and sampling time are also marked to form a data set of the species and quantity distribution of swimming animals in the middle and upper layers with three-dimensional labels of survey points, water depth and species. The Simrad EK80 broadband fishery acoustic detection system was controlled to continuously transmit broadband acoustic signals in the 38kHz, 70kHz, 120kHz and 200kHz frequency bands during the navigation of the ocean survey vessel, and the vertical echo scattering intensity was recorded. The Simrad EK80 broadband fishery acoustic detection system acquires data at a pulse transmission frequency of 5 times per second. It divides the entire water column into multiple depth segments with a depth resolution of 0.5 meters. The system performs background noise elimination and target intensity inversion processing on the raw acoustic echo data. Combined with system calibration data and acoustic scattering characteristic parameters of swimming animal target species, the system inverts the number density value of swimming animals per unit volume at each depth layer, forming a multi-dimensional acoustic inversion data set including depth coordinates, latitude and longitude coordinates, number density, and acoustic frequency band identifiers. S200.2 Configure a multi-dimensional data structure for the composition and distribution of swimming animals; Based on the obtained datasets of species and number distribution of swimming animals in the upper and middle layers and the multidimensional acoustic inversion dataset, the data were processed for format unification, spatial coordinate and depth label standardization, time synchronization and structured coding. A multidimensional swimming animal composition and distribution data structure oriented towards the survey station was constructed, which includes labels for species name, number density, taxonomic level, spatial location and sampling method.

4. The method for monitoring three-dimensional, multi-dimensional ocean swimming animals based on artificial intelligence as described in claim 1, characterized in that, Specifically, S300 is as follows: S300.

1. Based on the three-dimensional environmental parameter data set and the multi-dimensional swimming animal composition and distribution data structure, data fusion is performed to obtain the fused input data set, and a random forest regression model and a deep neural network model are configured. The temperature, salinity, depth, flow velocity, and flow direction data in the three-dimensional environmental parameter dataset are fused with the species name, resource density, taxonomic level, spatial location, and sampling method labels in the multi-dimensional swimming animal composition and distribution data structure. During the fusion process, the latitude and longitude coordinates, depth level, and sampling time of the survey station are used as unique indexes to match the environmental parameter data with the corresponding swimming animal distribution data and obtain the fused input dataset. Based on the obtained fused input dataset, a random forest regression model and a deep neural network model are constructed respectively. The input feature vector X of the random forest regression model is: Where T is the temperature parameter, S is the salinity parameter, D is the depth parameter, and V is the flow velocity parameter. The flow direction parameter is used to output the density of swimming animals at the corresponding location. ; The input feature vector of the deep neural network model is consistent with that of the random forest regression model. The network structure is set as follows: the input layer has 5 nodes, the first and second hidden layers each have 128 nodes and use the ReLU activation function, and the output layer uses the Sigmoid function to output the probability value of the species' existence. The training process of deep neural network models is optimized using the binary cross-entropy loss function; S300.2 Training a random forest regression model and a deep neural network model based on a fused input dataset; The random forest regression model and the deep neural network model were trained using a fused input dataset, and the prediction performance of the random forest regression model and the deep neural network model was verified by the five-fold cross-validation method. The fused input dataset is randomly divided into five subsets. Four subsets are used to train the random forest regression model and the deep neural network model, and one subset is used to validate the random forest regression model and the deep neural network model. The training and validation are repeated five times, and the model prediction error is recorded for each validation. The process is stopped when the trained random forest regression model and deep neural network model show stable prediction performance on unseen data. S300.

3. Based on the trained random forest regression model and deep neural network model, configure the multivariate mapping relationship structure between environmental factors and the distribution of swimming animals, obtain the random forest regression mapping structure and the deep neural network mapping structure, and output the optimal combination of habitat environmental parameters and model performance indicators. The parameters of the trained random forest regression model and deep neural network model are fixed to obtain the random forest regression mapping structure and the deep neural network mapping structure, respectively. The random forest regression mapping structure includes: input environmental factor dimension, number of decision trees, maximum tree depth, feature weight parameters, and prediction error boundary; The deep neural network mapping structure includes: input feature dimension, number of hidden layer nodes, weight matrix, bias vector, activation function type, and output species existence probability threshold; Feature importance scores for temperature, salinity, depth, flow velocity, and flow direction parameters are extracted from the random forest regression mapping structure. These scores are then sorted in descending order to form the optimal combination of habitat environmental parameters. The model's accuracy, recall, and F1 score are output from the deep neural network mapping structure. The specific performance metrics are calculated as follows: ; ; ; In the formula: Accuracy is the prediction accuracy in the performance metrics of deep neural network models; TP is the number of samples in the model prediction results where the true class is positive and the model predicts it as positive; TN is the number of samples in the model prediction results where the true class is negative and the model predicts it as negative; FP is the number of samples in the model prediction results where the true class is negative but the model predicts it as positive; FN is the number of samples in the model prediction results where the true class is positive but the model predicts it as negative; Recall is the recall rate in the performance metrics of deep neural network models; F1-score is the overall evaluation score in the performance metrics of deep neural network models; Precision is the precision rate in the model prediction results, that is, the proportion of samples that the model predicts as positive but are actually positive. The optimal combination of habitat environmental parameters and model performance indicators are recorded as: the output file for feature selection and prediction performance evaluation.

5. The method for monitoring three-dimensional, multi-dimensional ocean swimming animals based on artificial intelligence as described in claim 1, characterized in that, Specifically, S400 is: S400.

1. Generating a spatiotemporal distribution map of swimming animal community structure based on random forest regression mapping structure and deep neural network mapping structure; Based on the random forest regression mapping structure and the deep neural network mapping structure, the target ocean area is divided into horizontal grids with a latitude and longitude grid scale of 1°×1°, and multiple vertical water layers are divided at 500-meter intervals along the vertical direction, forming a spatial analysis unit where the horizontal grid and vertical water layers intersect. The environmental parameters in each spatial analysis unit are composed of temperature parameters, salinity parameters, depth parameters, current velocity parameters and current direction parameters. By inputting the environmental parameters within the spatial analysis unit into the random forest regression mapping structure and the deep neural network mapping structure respectively, the resource density value and the species existence probability value of swimming animals in each spatial analysis unit are predicted. Spatial interpolation and smoothing are performed on the predicted resource density values ​​and species existence probability values ​​obtained in each spatial analysis unit to output spatial distribution maps of swimming animal population density and species existence probability in the target ocean area. S400.

2. Based on the spatiotemporal distribution map of swimming animal community structure, calculate the Shannon diversity index and the Simpson diversity index, obtain the spatial distribution map of the Shannon diversity index and the Simpson diversity index, and perform spatial statistical analysis. Based on the generally accepted standards in the fields of current ecology and environmental biological monitoring, areas with a Shannon diversity index value >3 and a Simpson diversity index value >0.7 are defined as high diversity areas, while areas with a Shannon diversity index value <1 and a Simpson diversity index value <0.3 are defined as low diversity areas. Spatial statistical analysis was conducted on the spatiotemporal distribution characteristics and changing trends of high-diversity and low-diversity regions. The spatial diffusion characteristics of high-diversity and low-diversity regions over time were analyzed, and the main environmental factors leading to changes in high-diversity and low-diversity regions were analyzed based on temperature, salinity, depth, velocity, and direction parameters. The analysis results file includes: spatiotemporal distribution maps of Shannon and Simpson diversity indices, spatial statistical analysis data tables, and trend change maps, comprehensively characterizing the spatial distribution characteristics, spatiotemporal dynamic change trends, community stability, and ecosystem diversity of swimming animal communities in the target ocean area.

6. The method for monitoring ocean swimming animals in a three-dimensional, multi-dimensional manner based on artificial intelligence as described in claim 1, characterized in that, Specifically, S500 is as follows: S500.1 deploys remote sensing satellite platforms and UAV survey platforms to collect macro-scale environmental parameter data; In the extended observation area beyond the target ocean area, remote sensing satellite platforms and unmanned aerial vehicle (UAV) survey platforms are deployed. The remote sensing satellite platforms include: MODIS remote sensing satellite, Sentinel-3 remote sensing satellite and VIIRS remote sensing satellite. The drone survey platform includes: multi-rotor drones and fixed-wing drones; The remote sensing satellite platform conducts remote sensing observations of sea surface temperature parameters, sea surface chlorophyll concentration parameters, and sea surface wind speed parameters within the extended observation area with a spatial resolution of 1 km × 1 km and a sampling period of 12 hours, thereby obtaining macro-scale environmental parameter data. The UAV survey platform, within the extended observation area, surveys the sea area along a pre-planned gridded survey route at a survey flight altitude of 300 meters, obtaining visible light images of the sea surface with a spatial resolution of no less than 10 meters, infrared temperature maps, local wind speed parameters, and local wind direction parameters, supplementing the acquisition of high spatial accuracy local environmental parameter data. S500.

2. Based on macro-scale environmental parameter data, perform data preprocessing and configure the extended region prediction input dataset to obtain the extended region prediction input dataset. The macro-scale environmental parameter data obtained from remote sensing satellite platforms and the local environmental parameter data obtained from UAV survey platforms are processed by spatial coordinate reprojection, time synchronization resampling and data format unification. The sea surface temperature parameters, sea surface chlorophyll concentration parameters, sea surface wind speed parameters, local wind speed parameters, and local wind direction parameters that have undergone spatial coordinate reprojection are spatially interpolated and edge-stitched with the three-dimensional environmental parameter data set in the original survey area to obtain the extended area prediction input data set. S500.

3. Based on the extended region prediction input data set, perform prediction calculations to generate an extended region swimming animal spatial distribution map and identify high-probability distribution areas; The extended region prediction input dataset is input into a random forest regression mapping structure and a deep neural network mapping structure to perform prediction calculations of swimming animal resource density and swimming animal species existence probability within the extended region. The random forest regression mapping structure outputs the swimming animal resource density value in each spatial analysis unit of the extended region, while the deep neural network mapping structure outputs the swimming animal species existence probability value in each spatial analysis unit of the extended region. Spatial interpolation and smoothing are performed on the predicted density values ​​of swimming animals and the predicted probability values ​​of swimming animal species in the extended area to generate spatial distribution maps of swimming animal population density and swimming animal species probability in the extended area. Based on the prediction calculation results, the threshold for the density value of swimming animals is defined as 10 animals / cubic meter, and the threshold for the probability of the existence of swimming animal species is defined as 0.

6. Spatial analysis units with a resource density value greater than 10 animals / cubic meter and a species existence probability value greater than 0.6 are marked as high probability distribution areas. S500.4: Perform spatiotemporal dynamic analysis based on the spatial distribution map of swimming animals in the extended region, and output a large-scale prediction result file; Based on the spatial distribution map of swimming animals in the extended region and the mask map of high probability distribution areas, spatial statistical analysis tools were used to analyze the spatial migration characteristics of high probability distribution areas under different prediction timestamps. The spatial distribution maps of swimming animal population density, spatial distribution maps of swimming animal species probability, mask maps of high-probability distribution areas, and data tables of swimming animal resource density and species probability values ​​within the extended area are summarized to generate an extended area prediction result file that includes: spatiotemporal dynamic analysis maps and trend change analysis.

Images