Marine fishing ground monitoring system based on sensor network

Abstract

The invention discloses an ocean fishery monitoring system based on a sensor network, which relates to the technical field of ocean fishery, and comprises a data acquisition unit, a data processing unit, a behavior analysis unit, an anomaly detection unit and a response feedback unit, the data acquisition unit comprises a node configuration module, an optical monitoring module and a water quality monitoring module, the node configuration module is used for acquiring a fishing ground monitoring system distribution diagram from a database, establishing a polar coordinate system taking the center of a fishing ground as an original point, converting GPS latitude and longitude coordinates of each monitoring device into polar coordinates to form a monitoring node topology network, and transmitting the monitoring node topology network to the optical monitoring module; by integrating multi-source data of optical monitoring and water quality sensing and combining node coordinates to construct a space-time change model, the limitation that single sonar or radar equipment cannot capture behavior details of fish schools is broken through, and meanwhile, the problem that single-point monitoring of a water quality sensor is difficult to reflect the overall condition of a fishery is solved; and multi-dimensional and high-precision monitoring of the fishing ground environment is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of ocean fisheries, and in particular to an ocean fisheries monitoring system based on a sensor network. Background Art

[0002] Marine fishing grounds are areas of the ocean where aquatic resources such as fish are concentrated and suitable for fishing or aquaculture. Their formation is closely related to water temperature, salinity, ocean currents, and bait organisms. Areas such as the continental shelf, upwelling zones, and the confluence of cold and warm currents often become high-quality fishing grounds, supporting the development of the fishery economy.

[0003] Most of the existing marine fishery monitoring systems use a single monitoring method, such as detecting fish schools only through sonar or radar equipment, or monitoring water quality parameters only through water quality sensors. Although sonar or radar equipment can detect the presence and location of fish schools, it is difficult to provide detailed behavioral characteristics of fish schools, such as movement trajectory, aggregation degree, etc., and they are easily affected by environmental noise and interference. Although water quality sensors can monitor changes in water quality parameters in real time, they can usually only provide single-point data and cannot fully reflect the overall water quality status of the fishery. In addition, the temporal and spatial correlation between the data is weak, and there are problems such as limited monitoring range, insufficient data accuracy, and weak anomaly detection capabilities. It is difficult to meet the requirements of modern marine fisheries for comprehensive and accurate monitoring of fishery environment.

[0004] In view of this, the present invention proposes a marine fishery monitoring system based on a sensor network to remedy and improve the shortcomings of the existing technology. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a marine fishery monitoring system based on a sensor network to solve the corresponding technical problems raised in the above background technology.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: a marine fishery monitoring system based on a sensor network, comprising: a data acquisition unit, a data processing unit, a behavior analysis unit, an anomaly detection unit and a response feedback unit;

[0007] The data acquisition unit includes a node configuration module, an optical monitoring module, and a water quality monitoring module. The node configuration module is used to obtain a distribution map of the fishery monitoring system from a database, establish a polar coordinate system with the center of the fishery as the origin, convert the GPS latitude and longitude coordinates of each monitoring device into polar coordinates, and form a monitoring node topology network;

[0008] The optical monitoring module is used to collect a dynamic image sequence of a school of fish to generate image data with node coordinate marks, and the water quality monitoring module is used to collect water quality parameters to generate real-time monitoring data with node coordinate marks;

[0009] The data processing unit is used to acquire image data and real-time monitoring data, perform spatiotemporal alignment of image sequences at different times and perspectives based on timestamp synchronization and image registration technology, perform spatiotemporal calibration between nodes of the real-time monitoring data based on linear interpolation, and perform spatial fusion of the real-time monitoring data;

[0010] The behavior analysis unit includes an image processing module and a water quality analysis module. The image processing module is used to pre-process the dynamic image sequence of the fish school, identify and track the fish school in the image sequence based on the target detection and tracking algorithm, extract the motion trajectory of the fish school at each node by combining image processing technology, and calculate the characteristic parameters of the fish school behavior, including the aggregation index S and the abnormal behavior frequency F;

[0011] The water quality analysis module is used to establish a spatiotemporal variation model based on time series analysis and machine learning algorithms in combination with real-time monitoring data, and obtain the water quality parameter deviation δ by comparing the difference between the actual measurement value and the model prediction value;

[0012] The anomaly detection unit is used to obtain the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ of each monitoring node, preset multi-level threshold conditions according to the configuration file, database and expert system method, and combine the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ data of each monitoring node, as well as the correlation between the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ to determine the anomaly level λ∈{Ⅰ, Ⅱ, Ⅲ}.

[0013] Preferably, the response feedback unit is used to obtain the abnormality level λ and trigger a differentiated response according to the abnormality level λ, including:

[0014] Ⅰ: Mark abnormal nodes in the monitoring distribution map and generate logs;

[0015] II: Push warning information containing abnormal node coordinates to the management terminal;

[0016] III: Locate the abnormal node and control the feeding equipment within the radius of the node for targeted intervention.

[0017] Preferably, the optical monitoring module includes an underwater camera array arranged at each monitoring node, the water quality monitoring module includes a sensor array consisting of a dissolved oxygen sensor, a pH sensor, a temperature sensor, an ammonia nitrogen sensor and a turbidity sensor arranged at each monitoring node, and the water quality parameters include dissolved oxygen concentration DO, temperature T, pH, ammonia nitrogen concentration NH and turbidity NT.

[0018] Preferably, the specific process of performing spatiotemporal alignment on the image sequence is as follows:

[0019] S101, obtaining the polar coordinate position and acquisition timestamp of each underwater camera array, and mapping each image data and node coordinate to the corresponding position in the monitoring node topology network;

[0020] S102: For image data of different nodes at the same timestamp, spatial alignment of the image data is performed according to the node coordinates and preset alignment parameters;

[0021] S103. Based on underwater optical characteristics, an optical distortion model is established. Distortion compensation is performed on the graphics using the node coordinates and the optical distortion model. The formula is: m1sin(θ1) = m2sin(θ2), where m1 and m2 are the refractive indices of the incident and exiting media, respectively, and θ1 and θ2 are the incident and refraction angles, respectively.

[0022] S104: Perform a temporal consistency check on the aligned image sequence to ensure that the image data is continuous in time and has no obvious jumps.

[0023] As an example, the specific process of performing spatiotemporal calibration between nodes for real-time monitoring data is as follows:

[0024] S201, obtaining the polar coordinate position and acquisition timestamp of the sensor array consisting of the dissolved oxygen sensor, pH sensor, temperature sensor, ammonia nitrogen sensor, and turbidity sensor of each monitoring node, and mapping each real-time monitoring data and node coordinate to the corresponding position in the monitoring node topology network;

[0025] S202: Perform quality checks on real-time monitoring data, remove outliers, and clean the data;

[0026] S203. For the real-time monitoring data of different nodes at the same timestamp, perform time and space calibration of the real-time monitoring data according to the node coordinates and preset standard parameters. The formula is: Where e is the calibrated value, e1 and e2 are the original data points, t1 and t2 are the corresponding timestamps, and t is the target timestamp;

[0027] S204: Fusing the calibrated real-time monitoring data to generate a unified real-time monitoring data set, the formula of which is: in, is the fused value, k i is the original data point, w i is the corresponding weight.

[0028] Preferably, the specific process of calculating the aggregation index S is as follows:

[0029] S301, pre-processing the fish school dynamic image sequence, detecting and segmenting the fish school target from the image according to image processing technology, and calculating the centroid coordinates (x1, y1), the formula is: Where x and y are the coordinates of the image pixel, and I(x,y) is the grayscale value of the pixel;

[0030] S302, using the centroid of each fish school target as a vertex, perform Delaunay triangulation, and calculate the average perimeter of all triangles to measure the aggregation index S of the fish school. The formula is: Among them, n is the number of Delaunay triangles in the current frame image, P i is the perimeter of the i-th triangle. The smaller the calculated aggregation index S value, the higher the degree of aggregation of the fish school. The larger the calculated aggregation index S value, the lower the degree of aggregation of the fish school.

[0031] As an example, the specific process of calculating the abnormal behavior frequency F is as follows:

[0032] S401, extracting the movement trajectory of the fish school from the image sequence by tracking the change of the center of mass position of the fish school target over time based on image processing technology, outputting the actual movement trajectory, and classifying the actual movement trajectory into a behavior pattern;

[0033] S402: Compare the actual motion trajectories of different categories with the corresponding preset normal behavior patterns. If a trajectory is significantly different from the normal behavior pattern, it is considered an abnormal behavior;

[0034] S403. Count the number of abnormal behaviors that occur during the monitoring period, and regard it as the abnormal behavior frequency F.

[0035] As a preferred method, the specific process of obtaining the water quality parameter deviation δ is as follows:

[0036] S501. Based on the real-time monitoring data and the coordinate information of each node, a spatiotemporal variation model of water quality parameters is established, the formula of which is: q = a + bs + cd + ∈;

[0037] Among them, q is a water quality parameter;

[0038] s is time, which is used to describe the changes of water quality parameters over time;

[0039] d is the spatial coordinate, which is used to describe the changes of water quality parameters at different spatial locations;

[0040] a is the intercept term of the regression model, which represents the predicted value of q when both s and d are 0;

[0041] b and c are the regression coefficients of time s and spatial coordinate d, respectively, which are used to quantify the rate of change of water quality parameters with time and space;

[0042] ∈ is the error term;

[0043] S502. Based on the spatiotemporal variation model, the water quality parameter deviation δ between the actual measured value of the water quality parameter at each monitoring node and the model predicted value is calculated according to the relative error method. The calculation formula is:

[0044]

[0045] S503. Comprehensively evaluate the water quality status at each monitoring node based on the water quality parameter deviation δ. If the water quality parameter deviation δ exceeds a preset threshold, the abnormality detection unit is triggered to generate an abnormal alarm signal, and corresponding measures are taken through the corresponding feedback unit.

[0046] As an example, the specific process of determining the abnormality level λ∈{I, II, III} is as follows:

[0047] S601. Based on historical data of marine fisheries, industry standards and expert experience, preset multi-level threshold conditions for joint judgment of fish school behavior characteristic parameters and water quality parameter deviation δ to determine the severity of the anomaly. For the aggregation index S, set the threshold S 低 、S 中 、S 高 ;

[0048] For the frequency of abnormal behavior F, set the threshold F 低 、F 中 、F 高 ;

[0049] For the water quality parameter deviation δ, set the threshold δ 低 , δ 中 , δ 高 ;

[0050] S602: Compare the received fish school behavior characteristic parameters and water quality parameter deviation δ with the preset multi-level threshold conditions, where if S>S 高 or F>F 高 or δ>δ 高 , then it is preliminarily judged as abnormal level III;

[0051] If S 中 <S≤S 高 or F 中 <F≤F 高 or δ 中 <δ≤δ 高 , then it is preliminarily judged as abnormal level II;

[0052] If S≤S中 And F≤F 中 And δ≤δ 中 , but at least one parameter is close to its corresponding medium threshold, or there are multiple parameters that deviate slightly at the same time, then it is preliminarily judged as abnormal level I;

[0053] S603. Based on parameter comparison and judgment, the data of each monitoring node, as well as the correlation between the fish school behavior characteristic parameters and the water quality parameter deviation δ are comprehensively considered to finally determine the abnormality level λ. The formula is: λ = f(S, F, δ), where f is a function that determines the abnormality level based on the values of S, F, δ and their comparison with the threshold.

[0054] Compared with the existing technology, the beneficial effects of the present invention are: by integrating multi-source data of optical monitoring and water quality sensing, and combining node coordinates to construct a spatiotemporal change model, it not only breaks through the limitation that a single sonar or radar device cannot capture the details of fish behavior, but also solves the problem that single-point monitoring of water quality sensors is difficult to reflect the overall situation of the fishery, and realizes multi-dimensional and high-precision monitoring of the fishery environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic diagram of the overall structure of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] Embodiments of the present invention:

[0058] Please refer to Figure 1 As shown, the marine fishery monitoring system based on the sensor network includes: a data acquisition unit, a data processing unit, a behavior analysis unit, an anomaly detection unit and a response feedback unit;

[0059] The data acquisition unit includes a node configuration module, an optical monitoring module, and a water quality monitoring module. The node configuration module is used to obtain the distribution map of the fishery monitoring system from the database, establish a polar coordinate system with the center of the fishery as the origin, and convert the GPS longitude and latitude coordinates of each monitoring device into polar coordinates to form a monitoring node topology network.

[0060] The optical monitoring module is used to collect dynamic image sequences of fish schools to generate image data with node coordinate labels. The water quality monitoring module is used to collect water quality parameters to generate real-time monitoring data with node coordinate labels.

[0061] The data processing unit is used to acquire image data and real-time monitoring data. Based on timestamp synchronization and image registration technology, it performs spatiotemporal alignment of image sequences at different times and perspectives. Based on linear interpolation, it performs spatiotemporal calibration between nodes of the real-time monitoring data and spatially fuses the real-time monitoring data.

[0062] The behavior analysis unit includes an image processing module and a water quality analysis module. The image processing module is used to pre-process the dynamic image sequence of the fish school, identify and track the fish school in the image sequence based on the target detection and tracking algorithm, and extract the motion trajectory of the fish school at each node by combining image processing technology. The characteristic parameters of the fish school behavior, including the aggregation index S and the frequency of abnormal behavior F, are calculated.

[0063] The water quality analysis module is used to establish a spatiotemporal variation model based on time series analysis and machine learning algorithms combined with real-time monitoring data. By comparing the difference between the actual measured values and the model predicted values, the water quality parameter deviation δ is obtained;

[0064] The anomaly detection unit is used to obtain the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ of each monitoring node. According to the configuration file, database and expert system, the multi-level threshold conditions are preset, and the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ data of each monitoring node, as well as the correlation between the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ, are combined to determine the anomaly level λ∈{Ⅰ, Ⅱ, Ⅲ};

[0065] The response feedback unit is used to obtain the abnormality level λ and trigger differentiated responses based on the abnormality level λ, including:

[0066] Ⅰ: Mark abnormal nodes in the monitoring distribution map and generate logs;

[0067] II: Push warning information containing abnormal node coordinates to the management terminal;

[0068] III: Locate the abnormal node and control the feeding equipment within the radius of the node for targeted intervention;

[0069] The optical monitoring module includes an underwater camera array installed at each monitoring node. The water quality monitoring module includes a sensor array consisting of a dissolved oxygen sensor, a pH sensor, a temperature sensor, an ammonia nitrogen sensor, and a turbidity sensor installed at each monitoring node. Water quality parameters include dissolved oxygen concentration DO, temperature T, pH, ammonia nitrogen concentration NH, and turbidity NT.

[0070] The specific process of spatiotemporal alignment of image sequences is as follows:

[0071] S101, obtaining the polar coordinate position and acquisition timestamp of each underwater camera array, and mapping each image data and node coordinate to the corresponding position in the monitoring node topology network;

[0072] S102: For image data of different nodes at the same timestamp, spatial alignment of the image data is performed according to the node coordinates and preset alignment parameters;

[0073] S103. Based on underwater optical characteristics, an optical distortion model is established. Distortion compensation is performed on the graphics using the node coordinates and the optical distortion model. The formula is: m1sin(θ1) = m2sin(θ2), where m1 and m2 are the refractive indices of the incident and exiting media, respectively, and θ1 and θ2 are the incident and refraction angles, respectively.

[0074] S104: Perform a temporal consistency check on the aligned image sequence to ensure that the image data is continuous in time and has no obvious jumps;

[0075] The specific process of performing spatiotemporal calibration between nodes for real-time monitoring data is as follows:

[0076] S201, obtaining the polar coordinate position and acquisition timestamp of the sensor array consisting of the dissolved oxygen sensor, pH sensor, temperature sensor, ammonia nitrogen sensor, and turbidity sensor of each monitoring node, and mapping each real-time monitoring data and node coordinate to the corresponding position in the monitoring node topology network;

[0077] S202: Perform quality checks on real-time monitoring data, remove outliers, and clean the data;

[0078] S203. For the real-time monitoring data of different nodes at the same timestamp, perform time and space calibration of the real-time monitoring data according to the node coordinates and preset standard parameters. The formula is: Where e is the calibrated value, e1 and e2 are the original data points, t1 and t2 are the corresponding timestamps, and t is the target timestamp;

[0079] S204: Fusing the calibrated real-time monitoring data to generate a unified real-time monitoring data set, the formula of which is: in, is the fused value, k i is the original data point, w i is the corresponding weight;

[0080] The specific process of calculating the aggregation index S is as follows:

[0081] S301, pre-processing the fish school dynamic image sequence, detecting and segmenting the fish school target from the image according to image processing technology, and calculating the centroid coordinates (x1, y1), the formula is: Where x and y are the coordinates of the image pixel, and I(x,y) is the grayscale value of the pixel;

[0082] S302, using the centroid of each fish school target as a vertex, perform Delaunay triangulation, and calculate the average perimeter of all triangles to measure the aggregation index S of the fish school. The formula is: Among them, n is the number of Delaunay triangles in the current frame image, P i is the perimeter of the i-th triangle. The smaller the calculated aggregation index S value, the higher the aggregation degree of the fish school. The larger the calculated aggregation index S value, the lower the aggregation degree of the fish school.

[0083] The specific process of calculating the abnormal behavior frequency F is as follows:

[0084] S401, extracting the movement trajectory of the fish school from the image sequence by tracking the change of the center of mass position of the fish school target over time based on image processing technology, outputting the actual movement trajectory, and classifying the actual movement trajectory into a behavior pattern;

[0085] S402: Compare the actual motion trajectories of different categories with the corresponding preset normal behavior patterns. If a trajectory is significantly different from the normal behavior pattern, it is considered an abnormal behavior;

[0086] S403. Count the number of abnormal behaviors that occur during the monitoring period, which is regarded as the abnormal behavior frequency F;

[0087] The specific process of obtaining the water quality parameter deviation δ is as follows:

[0088] S501. Based on the real-time monitoring data and the coordinate information of each node, a spatiotemporal variation model of water quality parameters is established, the formula of which is: q = a + bs + cd + ∈;

[0089] Among them, q is a water quality parameter;

[0090] s is time, which is used to describe the changes of water quality parameters over time;

[0091] d is the spatial coordinate, which is used to describe the changes of water quality parameters at different spatial locations;

[0092] a is the intercept term of the regression model, which represents the predicted value of q when both s and d are 0;

[0093] b and c are the regression coefficients of time s and spatial coordinate d, respectively, which are used to quantify the rate of change of water quality parameters with time and space;

[0094] ∈ is the error term;

[0095] S502. Based on the spatiotemporal variation model, the water quality parameter deviation δ between the actual measured value of the water quality parameter at each monitoring node and the model predicted value is calculated according to the relative error method. The calculation formula is:

[0096]

[0097] S503, comprehensively assessing the water quality at each monitoring node based on the water quality parameter deviation δ. If the water quality parameter deviation δ exceeds a preset threshold, the abnormality detection unit is triggered to generate an abnormality alarm signal, and corresponding measures are taken through the corresponding feedback unit;

[0098] The specific process of determining the abnormality level λ∈{Ⅰ, Ⅱ, Ⅲ} is as follows:

[0099] S601. Based on historical data of marine fisheries, industry standards and expert experience, preset multi-level threshold conditions for joint judgment of fish school behavior characteristic parameters and water quality parameter deviation δ to determine the severity of the anomaly. For the aggregation index S, set the threshold S 低 、S 中 、S 高 ;

[0100] For the frequency of abnormal behavior F, set the threshold F 低 、F 中 、F 高 ;

[0101] For the water quality parameter deviation δ, set the threshold δ 低 , δ 中 , δ 高 ;

[0102] S602: Compare the received fish school behavior characteristic parameters and water quality parameter deviation δ with the preset multi-level threshold conditions, where if S>S 高 or F>F 高 or δ>δ 高 , then it is preliminarily judged as abnormal level III;

[0103] If S 中 <S≤S 高 or F 中 <F≤F 高 or δ 中 <δ≤δ 高 , then it is preliminarily judged as abnormal level II;

[0104] If S≤S 中 And F≤F 中 And δ≤δ 中 , but at least one parameter is close to its corresponding medium threshold, or there are multiple parameters that deviate slightly at the same time, then it is preliminarily judged as abnormal level I;

[0105] S603. Based on parameter comparison and judgment, the data of each monitoring node, as well as the correlation between the fish school behavior characteristic parameters and the water quality parameter deviation δ are comprehensively considered to finally determine the abnormality level λ. The formula is: λ = f(S, F, δ), where f is a function that determines the abnormality level based on the values of S, F, δ and their comparison with the threshold.

[0106] By integrating multi-source data from optical monitoring and water quality sensing and combining it with node coordinates to build a spatiotemporal change model, we not only break through the limitation that a single sonar or radar device cannot capture the details of fish behavior, but also solve the problem that single-point monitoring by water quality sensors cannot reflect the overall condition of the fishery, thus realizing multi-dimensional and high-precision monitoring of the fishery environment.

[0107] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technical personnel in this field for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0108] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by those skilled in the art according to actual conditions.

[0109] In the two embodiments provided in this application, it should be understood that the disclosed devices and systems can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the module division is merely a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not implemented. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interface, and the indirect coupling or communication connection of devices or modules may be electrical, mechanical or other forms.

[0110] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. The marine fishery monitoring system based on sensor network is characterized by: include: Data acquisition unit, data processing unit, behavior analysis unit, anomaly detection unit and response feedback unit; The data acquisition unit includes a node configuration module, an optical monitoring module, and a water quality monitoring module. The node configuration module is used to obtain a distribution map of the fishery monitoring system from a database, establish a polar coordinate system with the center of the fishery as the origin, convert the GPS latitude and longitude coordinates of each monitoring device into polar coordinates, and form a monitoring node topology network; The optical monitoring module is used to collect a dynamic image sequence of a school of fish to generate image data with node coordinate marks, and the water quality monitoring module is used to collect water quality parameters to generate real-time monitoring data with node coordinate marks; The data processing unit is used to acquire image data and real-time monitoring data, perform spatiotemporal alignment of image sequences at different times and perspectives based on timestamp synchronization and image registration technology, perform spatiotemporal calibration between nodes of the real-time monitoring data based on linear interpolation, and perform spatial fusion of the real-time monitoring data; The behavior analysis unit includes an image processing module and a water quality analysis module. The image processing module is used to pre-process the dynamic image sequence of the fish school, identify and track the fish school in the image sequence based on the target detection and tracking algorithm, extract the motion trajectory of the fish school at each node by combining image processing technology, and calculate the characteristic parameters of the fish school behavior, including the aggregation index S and the abnormal behavior frequency F; The water quality analysis module is used to establish a spatiotemporal variation model based on time series analysis and machine learning algorithms in combination with real-time monitoring data, and obtain the water quality parameter deviation δ by comparing the difference between the actual measurement value and the model prediction value; The anomaly detection unit is used to obtain the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ of each monitoring node, preset multi-level threshold conditions according to the configuration file, database and expert system, and combine the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ data of each monitoring node, as well as the correlation between the aggregation index S, abnormal behavior frequency F and water quality parameter deviation δ to determine the anomaly level λ∈{Ⅰ, Ⅱ, Ⅲ}.

2. The sensor network-based marine fishery monitoring system according to claim 1, characterized in that: The response feedback unit is used to obtain the abnormality level λ and trigger a differentiated response according to the abnormality level λ, including: Ⅰ: Mark abnormal nodes in the monitoring distribution map and generate logs; II: Push warning information containing abnormal node coordinates to the management terminal; III: Locate the abnormal node and control the feeding equipment within the radius of the node for targeted intervention.

3. The sensor network-based marine fishery monitoring system according to claim 2, characterized in that: The optical monitoring module includes an underwater camera array set at each monitoring node, and the water quality monitoring module includes a sensor array consisting of a dissolved oxygen sensor, a pH sensor, a temperature sensor, an ammonia nitrogen sensor and a turbidity sensor set at each monitoring node. The water quality parameters include dissolved oxygen concentration DO, temperature T, pH, ammonia nitrogen concentration NH and turbidity NT.

4. The sensor network-based marine fishery monitoring system according to claim 3, characterized in that: The specific process of spatiotemporal alignment of image sequences is as follows: S101, obtaining the polar coordinate position and acquisition timestamp of each underwater camera array, and mapping each image data and node coordinate to the corresponding position in the monitoring node topology network; S102: For image data of different nodes at the same timestamp, spatial alignment of the image data is performed according to the node coordinates and preset alignment parameters; S103. Based on underwater optical characteristics, an optical distortion model is established. Distortion compensation is performed on the graphics using the node coordinates and the optical distortion model. The formula is: m1sin(θ1) = m2sin(θ2), where m1 and m2 are the refractive indices of the incident and exiting media, respectively, and θ1 and θ2 are the incident and refraction angles, respectively. S104: Perform a temporal consistency check on the aligned image sequence to ensure that the image data is continuous in time and has no obvious jumps.

5. The sensor network-based marine fishery monitoring system according to claim 4, characterized in that: The specific process of performing spatiotemporal calibration between nodes for real-time monitoring data is as follows: S201, obtaining the polar coordinate position and acquisition timestamp of the sensor array consisting of the dissolved oxygen sensor, pH sensor, temperature sensor, ammonia nitrogen sensor, and turbidity sensor of each monitoring node, and mapping each real-time monitoring data and node coordinate to the corresponding position in the monitoring node topology network; S202: Perform quality checks on real-time monitoring data, remove outliers, and clean the data; S203. For the real-time monitoring data of different nodes at the same timestamp, perform time and space calibration of the real-time monitoring data according to the node coordinates and preset standard parameters. The formula is: Where e is the calibrated value, e1 and e2 are the original data points, t1 and t2 are the corresponding timestamps, and t is the target timestamp; S204: Fusing the calibrated real-time monitoring data to generate a unified real-time monitoring data set, the formula of which is: in, is the fused value, k i is the original data point, w i is the corresponding weight.

6. The sensor network-based marine fishery monitoring system according to claim 5, characterized in that: The specific process of calculating the aggregation index S is as follows: S301, pre-processing the fish school dynamic image sequence, detecting and segmenting the fish school target from the image according to image processing technology, and calculating the centroid coordinates (x1, y1), the formula is: Where x and y are the coordinates of the image pixel, and I(x,y) is the grayscale value of the pixel; S302, using the centroid of each fish school target as a vertex, perform Delaunay triangulation, and calculate the average perimeter of all triangles to measure the aggregation index S of the fish school. The formula is: Among them, n is the number of Delaunay triangles in the current frame image, P i is the perimeter of the i-th triangle. The smaller the calculated aggregation index S value, the higher the degree of aggregation of the fish school. The larger the calculated aggregation index S value, the lower the degree of aggregation of the fish school.

7. The sensor network-based marine fishery monitoring system according to claim 6, characterized in that: The specific process of calculating the abnormal behavior frequency F is as follows: S401, using image processing technology, extracting the movement trajectory of the fish school from the image sequence by tracking the change of the center of mass position of the fish school target over time and outputting it as the actual movement trajectory, and performing behavior pattern classification on the actual movement trajectory; S402: Compare the actual motion trajectories of different categories with the corresponding preset normal behavior patterns. If a trajectory is significantly different from the normal behavior pattern, it is considered an abnormal behavior; S403. Count the number of abnormal behaviors that occur during the monitoring period, and regard it as the abnormal behavior frequency F.

8. The sensor network-based marine fishery monitoring system according to claim 7, characterized in that: The specific process of obtaining the water quality parameter deviation δ is as follows: S501. Based on the real-time monitoring data and the coordinate information of each node, a spatiotemporal variation model of water quality parameters is established, the formula of which is: q = a + bs + cd + ∈; Among them, q is a water quality parameter; s is time, which is used to describe the changes of water quality parameters over time; d is the spatial coordinate, which is used to describe the changes of water quality parameters at different spatial locations; a is the intercept term of the regression model, which represents the predicted value of q when both s and d are 0; b and c are the regression coefficients of time s and spatial coordinate d, respectively, which are used to quantify the rate of change of water quality parameters with time and space; ∈ is the error term; S502. Based on the spatiotemporal variation model, the water quality parameter deviation δ between the actual measured value of the water quality parameter at each monitoring node and the model predicted value is calculated according to the relative error method. The calculation formula is: S503. Comprehensively evaluate the water quality status at each monitoring node based on the water quality parameter deviation δ. If the water quality parameter deviation δ exceeds a preset threshold, the abnormality detection unit is triggered to generate an abnormality alarm signal, and corresponding measures are taken through the corresponding feedback unit.

9. The sensor network-based marine fishery monitoring system according to claim 8, characterized in that: The specific process of determining the abnormality level λ∈{Ⅰ, Ⅱ, Ⅲ} is as follows: S601. Based on historical data of marine fisheries, industry standards and expert experience, preset multi-level threshold conditions for joint judgment of fish school behavior characteristic parameters and water quality parameter deviation δ to determine the severity of the anomaly. For the aggregation index S, set the threshold S 低 、S 中 、S 高 ; For the frequency of abnormal behavior F, set the threshold F 低 、F 中 、F 高 ; For the water quality parameter deviation δ, set the threshold δ 低 , δ 中 , δ 高 ; S602: Compare the received fish school behavior characteristic parameters and water quality parameter deviation δ with the preset multi-level threshold conditions, where if S>S 高 or F>F 高 or δ>δ 高 , then it is preliminarily judged as abnormal level III; If S 中 <S≤S 高 or F 中 <F≤F 高 or δ 中 <δ≤δ 高 , then it is preliminarily judged as abnormal level II; If S≤S 中 and F≤F 中 And δ≤δ 中 , but at least one parameter is close to its corresponding medium threshold, or there are multiple parameters that deviate slightly at the same time, then it is preliminarily judged as abnormal level I; S603. Based on parameter comparison and judgment, the data of each monitoring node, as well as the correlation between the fish school behavior characteristic parameters and the water quality parameter deviation δ are comprehensively considered to finally determine the abnormality level λ, which is formulated as follows: λ = f(S, F, δ), where f is a function that determines the abnormality level based on the values of S, F, and δ and the comparison results with the threshold.

Images