A smart fishery supervision method and system based on multi-source data fusion

Abstract

The present invention relates to the technical field of fishery management, and specifically to a smart fishery supervision method and system based on multi-source data fusion. The method and system comprise the following steps: obtaining multi-source trajectory data of a fishing vessel and performing data fusion to obtain precise trajectory data of the fishing vessel; performing trajectory feature analysis, trajectory behavior pattern analysis, and high-sensitivity point identification on the precise trajectory data to obtain trajectory distribution of the fishing vessel; obtaining multi-source fishery resource data and performing data fusion to obtain a fishery resource assessment vector; processing the fishery resource assessment vector using a fishery resource prediction model to obtain fishery resource distribution; performing spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis on the trajectory distribution of the fishing vessel and the fishery resource distribution to obtain analysis results; obtaining the analysis results and managing the data, and processing the data, and issuing an early warning if abnormal behavior is found.

Description

Technical Field

[0001] The present invention relates to the field of fishery management technology, and specifically to a smart fishery supervision method and system based on multi-source data fusion. Background Art

[0002] With increasing scarcity of fishery resources and intensifying ecological and environmental challenges, fishery supervision continues to intensify. Fisheries are not only crucial to food safety and ecological balance, but also a vital economic pillar for many coastal and inland regions. Traditional fishery supervision and management, primarily based on manual inspections, paper-based records, and periodic reporting, are inefficient and costly, making it difficult to promptly detect violations and ecological risks. Furthermore, their response speed and capacity to address issues such as illegal fishing, over-limit fishing, and water pollution are limited.

[0003] Against the backdrop of the rapid development of informatization, emerging technologies such as the Internet of Things, big data, cloud platforms, and artificial intelligence are providing new tools for fishery supervision and management. Smart fishery supervision and management aims to improve the comprehensiveness and timeliness of data collection by building a multi-layered perception and decision-making system covering waters, fishing vessels, aquaculture sites, and regulatory agencies, thereby achieving visual, quantifiable, and traceable management of the entire fishery production process.

[0004] The application of big data technology in smart fishery supervision encompasses the fusion and processing of multi-source, heterogeneous data, the identification of abnormal behavior, the construction of risk warning models, and compliance assessment and analysis. For example, by centrally processing and modeling data such as fishing vessel trajectories, water environment parameters, catch quantities, and operating times, it is possible to intelligently identify violations such as illegal fishing, overreaching operations, and unlicensed aquaculture. Furthermore, by combining video surveillance with dynamic aquaculture water quality data, an ecological risk warning mechanism can be established to assist supervisors in making real-time decisions.

[0005] Therefore, there is a need for an intelligent fishery supervision method and system based on multi-source data fusion, which can open up the whole chain process of data collection, processing, analysis and supervision decision-making, realize comprehensive supervision of fisheries, thereby improving supervision efficiency, reducing management costs, and promoting the sustainable utilization of fishery resources and the healthy development of the industry. Summary of the Invention

[0006] The present invention aims to provide a smart fishery supervision method and system based on multi-source data fusion to improve the efficiency of fishery resource supervision. The method obtains precise trajectory data of fishing vessels by acquiring and fusing multi-source trajectory data. Trajectory feature analysis, trajectory behavior pattern analysis, and high-sensitivity point identification are performed on the precise trajectory data to obtain the trajectory distribution of the fishing vessels. Multi-source fishery resource data is acquired and fused to obtain a fishery resource assessment vector. The fishery resource prediction model processes the fishery resource assessment vector to obtain the fishery resource distribution. Spatial, temporal, and intensity overlap analyses are performed on the trajectory distribution of the fishing vessels and the fishery resource distribution to obtain analysis results. The analysis results and management data are acquired and processed, and if abnormal behavior is detected, an early warning is issued.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A smart fishery supervision method based on multi-source data fusion, including:

[0009] Acquire multi-source trajectory data of fishing vessels and perform data fusion to obtain accurate trajectory data of fishing vessels;

[0010] Furthermore, the multi-source trajectory data includes at least automatic identification system data, fishing vessel monitoring system data, and fishing vessel trajectory information from satellite remote sensing;

[0011] The pre-processed multi-source trajectory data is segmented according to preset time intervals, and based on local motion characteristics and data sources, several behavior pattern hypotheses are generated for each trajectory segment, and an initial probability is assigned to each behavior pattern hypothesis;

[0012] The fishing vessel's trajectory segments are iteratively processed in chronological order. If the data density of a trajectory segment is less than a threshold, the trajectory range of the fishing vessel within the time period is predicted using the previous and next trajectory segment data, as well as the behavioral pattern hypothesis and its probability of this trajectory segment, and the trajectory segment data is fused.

[0013] If the data density of the trajectory segment is not less than the threshold, the Kalman filter is used to smooth and refine the trajectory;

[0014] After completing the preliminary reconstruction of all trajectory segments, a batch smoothing algorithm is used to perform global smoothing on the entire trajectory sequence to generate the final accurate trajectory data of the fishing vessel.

[0015] Conduct trajectory feature analysis, trajectory behavior pattern analysis, and high-sensitivity point identification on the precise trajectory data to obtain the trajectory distribution of the fishing vessel;

[0016] Further, performing trajectory feature analysis on the precise trajectory data, the trajectory features including at least speed, heading, position, and mooring / dwelling time; identifying the behavior pattern of the fishing vessel based on the trajectory feature analysis, the behavior pattern including at least normal navigation, trawling, purse seine, fixed-point operation, stopping, drifting, turning, acceleration or deceleration;

[0017] Based on trajectory characteristics and behavior patterns, trajectory points or trajectory segments that deviate from normal or legal behavior are identified as high-sensitivity points of fishing vessels; and a trajectory distribution is generated that represents the activity range, activity density, and location and characteristics of high-sensitivity points of fishing vessels.

[0018] Obtain fishery resource data from multiple sources and perform data fusion to obtain fishery resource assessment vectors;

[0019] Furthermore, multi-source fishery resource data include at least scientific survey data, fishery catch report data, environmental remote sensing data, and fishery biological data;

[0020] The pre-processed multi-source fishery resource data are uniformly projected onto a preset spatial grid;

[0021] For point data, match and associate it with other data in the corresponding space-time grid;

[0022] A hierarchical fusion model is constructed, including bottom-level fusion, middle-level fusion and high-level fusion. The bottom-level fusion includes proofreading, correcting and fusing the real data of scientific survey data and fishery catch report data; the middle-level fusion proofreads, corrects and fuses environmental remote sensing data and fishery biology data; the high-level fusion includes obtaining the results of bottom-level fusion and middle-level fusion, and fusing them to construct the fishery resource assessment vector corresponding to the spatial grid.

[0023] The fishery resource prediction model processes the fishery resource assessment vector to obtain the fishery resource distribution;

[0024] Furthermore, the fishery resource prediction model is a prediction model obtained by training based on historical multi-source fishery resource data, and the model includes but is not limited to a statistical model and a machine learning model;

[0025] The fishery resource prediction model receives the fishery resource assessment vector as input and outputs the distribution of fishery resources within the regulatory area, including the unit area abundance of different fishery resource species and total resources within the regulatory area;

[0026] The distribution of fishery resources is expressed in the form of a spatial grid and includes time dimension information.

[0027] The spatial overlap analysis, temporal overlap analysis and intensity overlap analysis of the trajectory distribution of fishing vessels and the distribution of fishery resources were carried out to obtain the analysis results;

[0028] Furthermore, spatial coincidence analysis involves spatially superimposing the trajectory distribution of fishing vessels with the distribution of fishery resources, and analyzing the spatial correlation of high-sensitivity points falling into fishery resource hotspots, including calculating spatial distances and falling proportions;

[0029] Temporal coincidence analysis includes determining the synchronization between the occurrence time of high-sensitive points and the resource high-value time window of fishery resource distribution;

[0030] Intensity overlap analysis involves correlating the sensitivity of highly sensitive points with the abundance or importance of resources in the area to determine whether highly sensitive behaviors target high-value or protected resources.

[0031] Obtain analysis results and management data and process them, and issue warnings if there are abnormal behaviors.

[0032] Furthermore, the management data includes at least: the operating permit of the fishing vessel, historical activity records, fishery regulations, and real-time meteorological information; the analysis results of spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis are comprehensively judged with the management data to construct an illegal fishing risk assessment rule set; based on the risk assessment rule set, it is assessed whether there are abnormal behaviors of fishing vessels within the time period and area, and early warning information is issued to relevant regulatory authorities, and the early warning information includes the fishing vessels suspected of abnormal behavior, time, location, behavioral characteristics, and risk assessment results.

[0033] The present invention also provides a smart fishery supervision system based on multi-source data fusion, comprising:

[0034] The fishing vessel data analysis module includes a trajectory processing unit and a trajectory analysis unit. The trajectory processing unit obtains multi-source trajectory data of the fishing vessel and performs data fusion to obtain the precise trajectory data of the fishing vessel. The trajectory analysis unit performs trajectory feature analysis, trajectory behavior pattern analysis and high-sensitivity point identification on the precise trajectory data to obtain the trajectory distribution of the fishing vessel.

[0035] The fishery data analysis module includes a fishery resource assessment unit and a fishery resource prediction unit. The fishery resource assessment unit obtains multi-source fishery resource data and performs data fusion to obtain a fishery resource assessment vector. The fishery resource prediction unit processes the fishery resource assessment vector according to the fishery resource prediction model to obtain the fishery resource distribution.

[0036] The fishery supervision module includes a comprehensive analysis unit and an early warning unit. The comprehensive analysis unit conducts spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis on the trajectory distribution of fishing vessels and the distribution of fishery resources to obtain analysis results; the early warning unit obtains and processes the analysis results and management data, and issues an early warning if there is any abnormal behavior.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. A method for generating precise trajectories based on multi-source data fusion. By intelligently processing the characteristics of different data sources and data missing patterns and incorporating behavioral patterns, it greatly improves the accuracy, continuity, and integrity of fishing vessel trajectories. By integrating sparse but critical observations such as satellite remote sensing data, it reconstructs fishing vessel movements that are closer to reality, especially revealing the activity trajectories of fishing vessels that attempt to evade supervision by turning off their signals. The generated precise trajectory data provides a solid and reliable foundation for subsequent high-sensitivity point identification, behavioral pattern analysis, and overlay analysis with resource distribution.

[0039] 2. By defining and identifying high-sensitivity points, potentially suspicious activity can be quickly screened from a large amount of normal trajectory data, significantly narrowing the attention span of supervisors and improving the efficiency of detecting abnormal behavior. The generated trajectory distribution clearly illustrates the key activity areas, main behavioral patterns, and suspicious locations and time periods requiring focused investigation, facilitating further integration and analysis of trajectory information with other data such as resource distribution.

[0040] 3. Bottom-level fusion improves the reliability of large-scale data such as catch reports; middle-level fusion uses environmental and biological data to correct indirect observations; and high-level fusion cleverly fuses multi-source data to form more comprehensive and robust resource assessment results. Especially in areas where scientific survey data is sparse or catch reports are unreliable, multi-layer fusion data processing can use environmental data to make reasonable inferences about resource status. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flowchart of a smart fishery supervision method based on multi-source data fusion provided by an embodiment of the present invention;

[0042] Figure 2 A flowchart for generating accurate trajectory data of a final fishing vessel provided by an embodiment of the present invention;

[0043] Figure 3 A schematic structural diagram of a smart fishery supervision system based on multi-source data fusion provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] 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.

[0045] Example 1:

[0046] In order to conduct more refined and efficient supervision and management of a certain fishery resource area, the present invention provides a smart fishery supervision method based on multi-source data fusion. The method flow is as follows: Figure 1 As shown, the specific implementation is as follows:

[0047] Acquire multi-source trajectory data of fishing vessels and perform data fusion to obtain accurate trajectory data of fishing vessels;

[0048] Furthermore, the multi-source trajectory data includes at least automatic identification system data, fishing vessel monitoring system data, and fishing vessel trajectory information from satellite remote sensing data;

[0049] The pre-processed multi-source trajectory data is segmented according to preset time intervals, and based on local motion characteristics and data sources, several behavior pattern hypotheses are generated for each trajectory segment, and an initial probability is assigned to each behavior pattern hypothesis;

[0050] Furthermore, the process of generating the final accurate trajectory data of the fishing vessel is as follows: Figure 2 As shown in the figure, for each trajectory segment, based on the trajectory point characteristics (such as point density, speed change, and heading stability) and data sources, several possible behavior pattern hypotheses are analyzed and generated. For example, the segment is most likely to be in normal navigation, low-speed trawling, fixed-point anchoring, signal interruption and possible operation, etc.; according to the degree of matching between the data characteristics and the known behavior patterns, an initial probability is assigned to each behavior pattern hypothesis.

[0051] The fishing vessel's trajectory segments are iteratively processed in chronological order. If the data density of a trajectory segment is less than a threshold, the trajectory range of the fishing vessel within the time period is predicted using the previous and next trajectory segment data, as well as the behavioral pattern hypothesis and its probability of this trajectory segment, and the trajectory segment data is fused.

[0052] Furthermore, for example, when there is a signal interruption or only a small number of satellite detection points, indicating significant data gaps or discontinuities, traditional filtering methods are ineffective in these situations. In this case, the behavioral pattern hypotheses and their corresponding probabilities for each trajectory segment are used. Based on the most probable behavioral pattern hypothesis, the possible trajectory range of the fishing vessel within that period is predicted using the motion model of that behavioral pattern, such as a tendency to move at low speed in a specific area in operating mode and in a straight line in navigation mode, as well as known trajectory points before and after the interruption. At the same time, all available sparse data points within the segment, particularly locations detected by satellite remote sensing, are considered observational evidence within the predicted range. Based on this observational evidence, the predicted trajectory range is further revised, and these sparse observations are integrated into the predicted trajectory range using methods such as particle filtering or optimization-based methods to reconstruct the most likely trajectory path for the fishing vessel within that period. This process dynamically adjusts the probability of the behavioral pattern hypothesis based on the location, number, and confidence of the observation points, converging to the trajectory that best explains all observation points.

[0053] If the data density of the trajectory segment is not less than the threshold, the Kalman filter is used to smooth and refine the trajectory;

[0054] After completing the initial reconstruction of all trajectory segments, a batch smoothing algorithm is used to perform global smoothing on the entire trajectory sequence to generate the final accurate trajectory data for the fishing vessel. Table 1 shows a sample of the accurate trajectory data for a fishing vessel after processing for some segments.

[0055] Table 1. Samples of accurate trajectory data for some segments

[0056]

[0057] This method for generating precise trajectories based on multi-source data fusion significantly improves the accuracy, continuity, and completeness of fishing vessel trajectories by intelligently processing the characteristics and data loss patterns of different data sources and incorporating behavioral patterns. Compared to single-source data or simple interpolation methods, it effectively fills in track gaps during periods of signal interruption. By integrating sparse but critical observations such as satellite remote sensing data, it reconstructs fishing vessel movements that are closer to reality, particularly revealing the activity trajectories of fishing vessels attempting to evade regulation by shutting down their signals. The resulting precise trajectory data provides a solid and reliable foundation for subsequent high-sensitivity point identification, behavioral pattern analysis, and overlay analysis with resource distribution.

[0058] Conduct trajectory feature analysis, trajectory behavior pattern analysis, and high-sensitivity point identification on the precise trajectory data to obtain the trajectory distribution of the fishing vessel;

[0059] Further, performing trajectory feature analysis on the precise trajectory data, the trajectory features including at least speed, heading, position, and mooring / dwelling time; identifying the behavior pattern of the fishing vessel based on the trajectory feature analysis, the behavior pattern including at least normal navigation, trawling, purse seine, fixed-point operation, stopping, drifting, turning, acceleration or deceleration;

[0060] Furthermore, behavioral pattern recognition can adopt rule-based methods, for example, if the speed changes continuously and evenly and the heading change rate is small, it is identified as trawling, or more complex machine learning algorithms such as hidden Markov models, support vector machines or deep learning models, which can automatically classify behavioral patterns by learning from large amounts of labeled data.

[0061] Based on trajectory characteristics and behavior patterns, trajectory points or trajectory segments that deviate from normal or legal behavior are identified as high-sensitivity points of fishing vessels; and a trajectory distribution is generated that represents the activity range, activity density, and location and characteristics of high-sensitivity points of fishing vessels.

[0062] Furthermore, high-sensitivity points are indicators of potential suspicious activity. The identification rules typically combine behavioral patterns with contextual information. For example, a fishing vessel's trajectory may show it moored for an extended period of time or accompanied by another vessel at a low speed in a non-port area; or it may show frequent turning, acceleration, or deceleration near the edge of known fishing grounds or sensitive areas. These high-sensitivity points warrant special attention and further analysis.

[0063] By defining and identifying high-sensitivity points, potentially suspicious activity can be quickly screened from a large amount of normal trajectory data, significantly narrowing the attention span of supervisors and improving the efficiency of detecting abnormal behavior. The generated trajectory distribution clearly illustrates the key activity areas, main behavioral patterns, and suspicious locations and time periods requiring focused investigation, facilitating further integration and analysis of trajectory information with other data such as resource distribution.

[0064] Obtain fishery resource data from multiple sources and perform data fusion to obtain fishery resource assessment vectors;

[0065] Furthermore, multi-source fishery resource data include at least scientific survey data, fishery catch report data, environmental remote sensing data, and fishery biological data;

[0066] The pre-processed multi-source fishery resource data are uniformly projected onto a preset spatial grid;

[0067] For point data, match and associate it with other data in the corresponding space-time grid;

[0068] A hierarchical fusion model is constructed, including bottom-level fusion, middle-level fusion and high-level fusion. The bottom-level fusion includes proofreading, correcting and fusing the real data of scientific survey data and fishery catch report data; the middle-level fusion proofreads, corrects and fuses environmental remote sensing data and fishery biology data; the high-level fusion includes obtaining the results of bottom-level fusion and middle-level fusion, and fusing them to construct the fishery resource assessment vector corresponding to the spatial grid.

[0069] Furthermore, the bottom-level fusion focuses on correcting and integrating observation data that directly reflect resource information; the middle-level fusion obtains environmental remote sensing data and fishery biology data to realize the description of the construction of fishery resource ecology; the high-level fusion integrates the bottom-level fusion and the middle-level fusion to generate the final fishery resource assessment vector.

[0070] Bottom-level fusion improves the reliability of large-scale data such as catch reports; mid-level fusion uses environmental and biological data to correct for indirect observations; and high-level fusion skillfully integrates multi-source data to produce more comprehensive and robust stock assessments. Particularly in areas where scientific survey data is sparse or catch reports are unreliable, the fusion of mid- and high-level data enables reasonable inferences about stock status using continuous environmental data.

[0071] The fishery resource prediction model processes the fishery resource assessment vector to obtain the fishery resource distribution;

[0072] Furthermore, the fishery resource prediction model is a prediction model obtained by training based on historical multi-source fishery resource data, and the model includes but is not limited to a statistical model and a machine learning model;

[0073] The fishery resource prediction model can be a fishery resource spatiotemporal prediction model based on the long short-term memory network (LSTM).

[0074] The fishery resource prediction model receives the fishery resource assessment vector as input and outputs the distribution of fishery resources within the regulatory area, including the unit area abundance of different fishery resource species and total resources within the regulatory area;

[0075] The fishery resource assessment vector includes the sequence of fishery resource abundance and environmental factors for the grid cell over a period of time (e.g., the past T time steps). Information about spatially neighboring grid cells is input as an additional feature to capture spatial correlation.

[0076] The distribution of fishery resources is expressed in the form of a spatial grid and includes time dimension information.

[0077] The training process first collects and processes historical fishery resource data from multiple sources, constructing time series samples for each spatial grid. The data is normalized to facilitate model training. The time series data is divided into training, validation, and test sets. Data from earlier time periods are used as the training and validation sets, while data from later time periods are used as the test set. The mean squared error (MSE) loss function is used, and the Adam optimizer is selected as the optimizer. Finally, the trained model is used to predict fishery resource distribution.

[0078] Furthermore, after obtaining the assessment vectors representing the status of fishery resources in a specific area and time period through hierarchical fusion of multi-source data, the next step is to use these assessment results as input to generate an intuitive and comprehensive spatial distribution map of fishery resources through a pre-trained prediction model, clearly showing the abundance of different resource types and total resources in the regulatory area.

[0079] Furthermore, the model calculates and outputs detailed distribution information of fishery resources for each spatial grid cell within the regulatory area during the target time period, including: the estimated abundance of different fishery resource species within that unit area, i.e., the fish biomass per square kilometer; the estimated abundance of the total resource within that unit area, i.e., the total biomass of all fishery resource species within that cell;

[0080] Furthermore, the model may also output the predicted uncertainty or confidence in these abundance estimates.

[0081] Furthermore, the output results are presented in the form of a grid map, and the value of each grid cell represents the estimated resource abundance in a specific time period within the cell, such as a week, month or quarter.

[0082] Through the processing of the fishery resource prediction model, comprehensive and continuous resource distribution information within the supervision area can be provided; the generated spatial grid resource distribution map can be directly used as input for overlay analysis and spatial correlation calculation with the high-sensitive points of the fishing vessel trajectory, greatly simplifying the subsequent illegal fishing risk assessment process.

[0083] The spatial overlap analysis, temporal overlap analysis and intensity overlap analysis of the trajectory distribution of fishing vessels and the distribution of fishery resources were carried out to obtain the analysis results;

[0084] Furthermore, spatial coincidence analysis involves spatially superimposing the trajectory distribution of fishing vessels with the distribution of fishery resources, and analyzing the spatial correlation of high-sensitivity points falling into fishery resource hotspots, including calculating spatial distances and falling proportions;

[0085] Temporal coincidence analysis includes determining the synchronization between the occurrence time of high-sensitive points and the resource high-value time window of fishery resource distribution;

[0086] Intensity overlap analysis involves correlating the sensitivity of highly sensitive points with the abundance or importance of resources in the area to determine whether highly sensitive behaviors target high-value or protected resources.

[0087] Furthermore, spatial distance calculation involves calculating the minimum distance from a high-sensitivity point to the nearest resource hotspot boundary or high-abundance grid cell. The distribution characteristics of the distance from high-sensitivity points to hotspots are analyzed, for example, to determine whether a large number of high-sensitivity points are clustered near the boundaries of no-fishing zones. Falling-in ratio analysis involves calculating the proportion of high-sensitivity points or their duration that fall within pre-defined fishery resource hotspots or within grid cells with resource abundance above a certain threshold, relative to the total number or total duration of all high-sensitivity points.

[0088] Furthermore, synchronization analysis involves calculating the proportion of high-sensitivity points occurring within predicted high-value time windows. This analysis also examines potential time lags or leads between the onset, duration, or end of high-sensitivity behavior and the resource high-value time windows. For example, it can be determined whether suspicious activity occurs during the season when a particular fish species is most likely to be found in the area or during the period when its abundance is predicted to be highest.

[0089] Furthermore, intensity overlap analysis aims to combine the "sensitivity" of high-sensitivity points with the "value or importance" of regional resources to assess the "risk intensity" of potential illegal activities. Specifically, it assigns a sensitivity score based on the type and characteristics of high-sensitivity points, such as the duration of AIS interruption, the severity of deviation from normal behavior, and the occurrence within a specific control area; assigns a resource value or importance score to the area based on factors such as the predicted abundance of fishery resources in the area, the type of resources, and whether the area is an important breeding ground, migratory channel or protected area; and correlates the sensitivity score of the high-sensitivity point with the resource value / importance score of the area where the point is located, such as multiplying them or mapping them through a risk matrix, to obtain a comprehensive "risk intensity" or "potential illegal fishing index", and analyzes whether highly sensitive behavior exists.

[0090] This series of comprehensive analysis steps solves the problem that illegal fishing cannot be directly determined by analyzing trajectories or resources alone. It creatively links the "behavior" of fishing vessels with the "goals" of fisheries, accurately identifies potential illegal fishing targets, provides quantitative evidence support, and focuses on key regulatory areas and time periods.

[0091] The analysis results and management data are obtained and processed, and if there is any abnormal behavior, an early warning is issued. Table 2 shows the management data of some fishing vessels.

[0092] Table 2. Management data of some fishing vessels

[0093]

[0094] Furthermore, the management data includes at least: the operating permit of the fishing vessel, historical activity records, fishery regulations, and real-time meteorological information; the analysis results of spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis are comprehensively judged with the management data to construct an illegal fishing risk assessment rule set; based on the risk assessment rule set, it is assessed whether there are abnormal behaviors of fishing vessels within the time period and area, and early warning information is issued to relevant regulatory authorities, and the early warning information includes the fishing vessels suspected of abnormal behavior, time, location, behavioral characteristics, and risk assessment results.

[0095] This final step closes the loop of the entire smart fisheries regulatory approach and is key to transforming data analysis capabilities into actual regulatory effectiveness. It combines complex analytical results with real-world regulatory rules, vessel background and environmental factors to automate risk assessment and achieve intelligent decision-making assistance.

[0096] The intelligent fishery supervision method based on multi-source data fusion provided by the present invention integrates the multiple information flows of fishing vessel trajectories and fishery resources, achieving higher-precision, more comprehensive, and more continuous supervision of fishery activities. Its core advantage lies in that it can not only generate accurate fishing vessel trajectories that overcome the limitations of a single data source and a fishery resource distribution map predicted based on multi-source data, but more importantly, it can actively associate and verify potential abnormal behaviors of fishing vessels with fishery resource hotspots through intelligent spatiotemporal and intensity overlap analysis, and make comprehensive judgments based on management data such as fishing vessel qualifications and regulations. It can accurately identify and filter out high-risk events that are likely to be suspected of illegal fishing, focusing supervision on specific ships, times, and locations where illegal activities are most likely to occur, greatly improving the efficiency and accuracy of fishery supervision.

[0097] Example 2:

[0098] The present invention also provides a smart fishery supervision system based on multi-source data fusion, the system structure is as follows Figure 3 As shown, the specific implementation is as follows:

[0099] The fishing vessel data analysis module includes a trajectory processing unit and a trajectory analysis unit. The trajectory processing unit obtains multi-source trajectory data of the fishing vessel and performs data fusion to obtain the precise trajectory data of the fishing vessel. The trajectory analysis unit performs trajectory feature analysis, trajectory behavior pattern analysis and high-sensitivity point identification on the precise trajectory data to obtain the trajectory distribution of the fishing vessel.

[0100] Furthermore, in the trajectory processing unit, the multi-source trajectory data includes at least automatic identification system data, fishing vessel monitoring system data, and fishing vessel trajectory information from satellite remote sensing data;

[0101] The pre-processed multi-source trajectory data is segmented according to preset time intervals, and based on local motion characteristics and data sources, several behavior pattern hypotheses are generated for each trajectory segment, and an initial probability is assigned to each behavior pattern hypothesis;

[0102] The fishing vessel's trajectory segments are iteratively processed in chronological order. If the data density of a trajectory segment is less than a threshold, the trajectory range of the fishing vessel within the time period is predicted using the previous and next trajectory segment data, as well as the behavioral pattern hypothesis and its probability of this trajectory segment, and the trajectory segment data is fused.

[0103] If the data density of the trajectory segment is not less than the threshold, the Kalman filter is used to smooth and refine the trajectory;

[0104] After completing the preliminary reconstruction of all trajectory segments, a batch smoothing algorithm is used to perform global smoothing on the entire trajectory sequence to generate the final accurate trajectory data of the fishing vessel.

[0105] Furthermore, in the trajectory analysis unit, trajectory feature analysis is performed on the precise trajectory data, where the trajectory features include at least speed, heading, position, and mooring / staying time; based on the trajectory feature analysis, a behavior pattern of the fishing vessel is identified, where the behavior pattern includes at least normal navigation, trawling, purse seine, fixed-point operation, stopping, drifting, turning, acceleration, or deceleration;

[0106] Based on trajectory characteristics and behavior patterns, trajectory points or trajectory segments that deviate from normal or legal behavior are identified as high-sensitivity points of fishing vessels; and a trajectory distribution is generated that represents the activity range, activity density, and location and characteristics of high-sensitivity points of fishing vessels.

[0107] The fishery data analysis module includes a fishery resource assessment unit and a fishery resource prediction unit. The fishery resource assessment unit obtains multi-source fishery resource data and performs data fusion to obtain a fishery resource assessment vector. The fishery resource prediction unit processes the fishery resource assessment vector according to the fishery resource prediction model to obtain the fishery resource distribution.

[0108] Furthermore, in the fishery resource assessment unit, the multi-source fishery resource data includes at least scientific survey data, fishery catch report data, environmental remote sensing data, and fishery biological data;

[0109] The pre-processed multi-source fishery resource data are uniformly projected onto a preset spatial grid;

[0110] For point data, match and associate it with other data in the corresponding space-time grid;

[0111] A hierarchical fusion model is constructed, including bottom-level fusion, middle-level fusion and high-level fusion. The bottom-level fusion includes proofreading, correcting and fusing the real data of scientific survey data and fishery catch report data; the middle-level fusion proofreads, corrects and fuses environmental remote sensing data and fishery biology data; the high-level fusion includes obtaining the results of bottom-level fusion and middle-level fusion, and fusing them to construct the fishery resource assessment vector corresponding to the spatial grid.

[0112] Furthermore, in the fishery resource prediction unit, the fishery resource prediction model is a prediction model obtained by training based on historical multi-source fishery resource data, and the model includes but is not limited to a statistical model and a machine learning model;

[0113] The fishery resource prediction model receives the fishery resource assessment vector as input and outputs the distribution of fishery resources within the regulatory area, including the unit area abundance of different fishery resource species and total resources within the regulatory area;

[0114] The distribution of fishery resources is expressed in the form of a spatial grid and includes time dimension information.

[0115] The fishery supervision module includes a comprehensive analysis unit and an early warning unit. The comprehensive analysis unit conducts spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis on the trajectory distribution of fishing vessels and the distribution of fishery resources to obtain analysis results; the early warning unit obtains and processes the analysis results and management data, and issues an early warning if there is any abnormal behavior.

[0116] Furthermore, in the comprehensive analysis unit, spatial coincidence analysis includes spatially superimposing the trajectory distribution of fishing vessels with the distribution of fishery resources, analyzing the spatial correlation of high-sensitivity points falling into the hotspot areas of fishery resources, including calculating the spatial distance and the proportion of falling into the hotspot areas;

[0117] Temporal coincidence analysis includes determining the synchronization between the occurrence time of high-sensitive points and the resource high-value time window of fishery resource distribution;

[0118] Intensity overlap analysis involves correlating the sensitivity of highly sensitive points with the abundance or importance of resources in the area to determine whether highly sensitive behaviors target high-value or protected resources.

[0119] Furthermore, in the early warning unit, the management data includes at least: the fishing vessel's operating permit, historical activity records, fishery regulations, and real-time meteorological information. The results of spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis are combined with the management data to construct an illegal fishing risk assessment rule set. Based on this risk assessment rule set, an assessment is made of whether a fishing vessel has engaged in abnormal behavior within a time period and region, and an early warning message is issued to the relevant regulatory authorities. The warning message includes the fishing vessel suspected of abnormal behavior, the time, location, behavioral characteristics, and the risk assessment results. Table 3 shows some of the monitored early warning data. The system provided by the present invention enables efficient supervision of fisheries in a region.

[0120] Table 3. Summary of early warning information

[0121]

[0122] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A smart fishery supervision method based on multi-source data fusion, characterized by: include: Acquire multi-source trajectory data of fishing vessels and perform data fusion to obtain accurate trajectory data of fishing vessels; Multi-source trajectory data includes at least automatic identification system data, fishing vessel monitoring system data, and fishing vessel trajectory information from satellite remote sensing; The pre-processed multi-source trajectory data is segmented according to preset time intervals, and based on local motion characteristics and data sources, several behavior pattern hypotheses are generated for each trajectory segment, and an initial probability is assigned to each behavior pattern hypothesis; The fishing vessel's trajectory segments are iteratively processed in chronological order. If the data density of a trajectory segment is less than a threshold, the trajectory range of the fishing vessel within the time period is predicted using the previous and next trajectory segment data, as well as the behavioral pattern hypothesis and its probability of this trajectory segment, and the trajectory segment data is fused. If the data density of the trajectory segment is not less than the threshold, the Kalman filter is used to smooth and refine the trajectory; After completing the preliminary reconstruction of all trajectory segments, a batch smoothing algorithm is used to perform global smoothing on the entire trajectory sequence to generate the final accurate trajectory data of the fishing vessel. Conduct trajectory feature analysis, trajectory behavior pattern analysis, and high-sensitivity point identification on the precise trajectory data to obtain the trajectory distribution of the fishing vessel; Obtain fishery resource data from multiple sources and perform data fusion to obtain fishery resource assessment vectors; Multi-source fishery resource data include at least scientific survey data, fishery catch report data, environmental remote sensing data, and fishery biology data; The pre-processed multi-source fishery resource data are uniformly projected onto a preset spatial grid; For point data, match and associate it with the data in the corresponding space-time grid; Construct a hierarchical fusion model, including bottom-level fusion, middle-level fusion, and high-level fusion. The bottom-level fusion involves proofreading, correcting, and fusing the real data of scientific survey data and fishery catch report data. The mid-level fusion process verifies, corrects and integrates environmental remote sensing data and fishery biological data; High-level fusion includes obtaining the results of bottom-level fusion and middle-level fusion, and fusing them to construct the fishery resource assessment vector corresponding to the spatial grid; The fishery resource prediction model processes the fishery resource assessment vector to obtain the fishery resource distribution; Conduct spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis on the trajectory distribution of fishing vessels and the distribution of fishery resources to obtain analysis results; obtain and manage the data and process them, and issue an early warning if there is any abnormal behavior.

2. The intelligent fishery supervision method based on multi-source data fusion according to claim 1 is characterized by: Performing trajectory feature analysis on the precise trajectory data, where the trajectory features include at least speed, heading, position, mooring and dwell time; identifying a behavior pattern of the fishing vessel based on the trajectory feature analysis, the behavior pattern including at least normal navigation, trawling, purse seine, fixed-point operation, stopping, drifting, turning, acceleration, and deceleration; Based on trajectory characteristics and behavior patterns, trajectory points and / or trajectory segments that deviate from normal and legal behavior are identified as high-sensitivity points of fishing vessels; and a trajectory distribution is generated that represents the activity range, activity density, and location and characteristics of high-sensitivity points of fishing vessels.

3. The intelligent fishery supervision method based on multi-source data fusion according to claim 1 is characterized in that: The fishery resource prediction model is a prediction model obtained by training based on historical multi-source fishery resource data, and the model includes but is not limited to a statistical model and a machine learning model; The fishery resource prediction model receives the fishery resource assessment vector and outputs the distribution of fishery resources within the regulatory area, including the unit area abundance of different fishery resource species and total resources within the regulatory area; The distribution of fishery resources is expressed in the form of a spatial grid and includes time dimension information.

4. The intelligent fishery supervision method based on multi-source data fusion according to claim 1 is characterized in that: Spatial coincidence analysis involves spatially overlaying the trajectory distribution of fishing vessels with the distribution of fishery resources, analyzing the spatial correlation of high-sensitivity points falling into fishery resource hotspots, including calculating the spatial distance and proportion of falling into hotspots. Temporal coincidence analysis includes determining the synchronization between the occurrence time of high-sensitive points and the resource high-value time window of fishery resource distribution; Intensity overlap analysis involves correlating the sensitivity of highly sensitive points with the abundance and importance of resources in the area to determine whether highly sensitive behaviors target high-value and protected resources.

5. The intelligent fishery supervision method based on multi-source data fusion according to claim 1 is characterized in that: The management data includes at least: the operating permit of the fishing vessel, historical activity records, fishery regulations and real-time meteorological information; the analysis results of spatial overlap analysis, temporal overlap analysis and intensity overlap analysis are comprehensively judged with the management data to construct an illegal fishing risk assessment rule set; based on the risk assessment rule set, an assessment is made as to whether there are abnormal behaviors of fishing vessels within a time period and area, and early warning information is issued to relevant regulatory authorities. The early warning information includes the fishing vessel suspected of abnormal behavior, time, location, behavioral characteristics and risk assessment results.

6. A smart fishery supervision system based on multi-source data fusion, which implements the smart fishery supervision method based on multi-source data fusion as claimed in claim 1, characterized in that: include: The fishing vessel data analysis module includes a trajectory processing unit and a trajectory analysis unit. The trajectory processing unit obtains multi-source trajectory data of the fishing vessel and performs data fusion to obtain the precise trajectory data of the fishing vessel. The trajectory analysis unit performs trajectory feature analysis, trajectory behavior pattern analysis and high-sensitivity point identification on the precise trajectory data to obtain the trajectory distribution of the fishing vessel. The fishery data analysis module includes a fishery resource assessment unit and a fishery resource prediction unit. The fishery resource assessment unit obtains multi-source fishery resource data and performs data fusion to obtain a fishery resource assessment vector. The fishery resource prediction unit processes the fishery resource assessment vector according to the fishery resource prediction model to obtain the fishery resource distribution. The fishery supervision module includes a comprehensive analysis unit and an early warning unit. The comprehensive analysis unit conducts spatial overlap analysis, temporal overlap analysis, and intensity overlap analysis on the trajectory distribution of fishing vessels and the distribution of fishery resources to obtain analysis results; the early warning unit obtains and processes the analysis results and management data, and issues an early warning if there is any abnormal behavior.

Images