Fish supervision and decision-making system based on multi-source data fusion and risk level evaluation

Abstract

The application discloses a fish supervision decision system based on multi-source data fusion and risk level evaluation, relates to the technical field of multi-source data fusion, and obtains water environment data, fish school behavior data, breeding process data and historical supervision data of a target fish supervision object, fuses to obtain a comprehensive risk feature set and determines an initial risk level; according to supervision measure execution information and intervention influence features, supervision action-observation response coupling parameters and cross-modal risk recovery inertia mismatch parameters are obtained, and then a human intervention misjudgment index is determined and the fusion weight of the intervened features is adjusted to obtain an intervention corrected risk value; according to the corrected risk level and the risk source, the execution intensity, the duration and the review frequency of the supervision measures are dynamically adjusted, so that the interference of the transient data change caused by the supervision measures on the risk judgment is reduced, and the accuracy and stability of the fish supervision decision are improved.

Description

Technical Field

[0001] This invention relates to the field of multi-source data fusion technology, and more specifically to a fish regulatory decision-making system based on multi-source data fusion and risk level assessment. Background Technology

[0002] Current fish management methods typically involve risk assessment based on data collection of water quality, fish behavior, aquaculture processes, and historical monitoring data. Based on the risk level, appropriate management measures such as aeration, water exchange, disinfection, or fish relocation are then implemented. However, these measures directly alter dissolved oxygen, water flow, turbidity, and fish movement, causing significant intervention-related changes in sensor data. Existing technologies struggle to determine whether these changes stem from a genuine shift in fish risk status or are merely transient disturbances caused by the management measures. This can easily lead to misinterpreting data improvements or deteriorations resulting from human intervention as genuine risk changes, resulting in erroneous risk level adjustments, repeated implementation of management measures, or premature revocation of measures, ultimately reducing the accuracy and stability of fish management decisions. Summary of the Invention

[0003] The purpose of this invention is to provide a fish regulatory decision-making system based on multi-source data fusion and risk level assessment to address the shortcomings of the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a fish regulatory decision-making system based on multi-source data fusion and risk level assessment, comprising: The data acquisition module acquires water environment data, fish behavior data, aquaculture process data, and historical regulatory data of the target fish species. After time-series alignment and fusion, a comprehensive risk feature set is obtained, and an initial risk level is determined. Regulatory measures are matched according to the initial risk level, the execution time and intensity of the regulatory measures are obtained, and the change features in the comprehensive risk feature set before and after the execution are extracted to obtain the intervention impact feature set. The parameter calculation module obtains the regulatory action-observation response coupling parameters based on the execution time, execution intensity, and the change time, change direction, and response sequence of each feature in the intervention impact feature set; and divides the intervention impact features into fast response risk features and slow response risk features, and obtains the cross-modal risk recovery inertia mismatch parameters based on the difference in recovery rate between the two types of risk features before and after the implementation of regulatory measures. The weight adjustment module obtains the human intervention misjudgment index based on the regulatory action-observation response coupling parameters and cross-modal risk recovery inertia mismatch parameters, adjusts the fusion weights corresponding to the intervention impact features based on the human intervention misjudgment index, and recalculates the intervention correction risk value based on the adjusted fusion weights. The regulatory decision adjustment module redetermines the risk level based on the intervention correction risk value, and dynamically adjusts the intensity, duration, and review frequency of regulatory measures based on the redetermined risk level and risk source.

[0005] Preferably, obtaining the intervention impact feature set includes: using the execution time of the regulatory measures as a time reference, obtaining comprehensive risk feature sets before and after the execution of the regulatory measures, comparing the same risk features before and after execution to determine the risk features that have undergone state changes; determining the change time and change direction corresponding to each change feature, and sorting the change time of each change feature according to the chronological order to determine the response order of each change feature; and associating and aggregating each change feature and its corresponding change time, change direction, and response order to obtain the intervention impact feature set.

[0006] Preferably, obtaining the regulatory action-observation response coupling parameters includes: acquiring a reference response relationship corresponding to the current regulatory measure, wherein the reference response relationship is determined based on historical execution information of the same type of regulatory measures and corresponding risk characteristic change information; Based on the intensity of the current regulatory measures, a reference response relationship corresponding to or similar to the current intensity of implementation is determined from the reference response relationship. The reference response relationship includes the reference response time difference, reference change direction, and reference response order corresponding to each change feature, which serves as a comparison benchmark for the response correlation analysis between the current intervention impact features and regulatory measures.

[0007] Preferably, obtaining the regulatory action-observation response coupling parameters further includes: determining the actual response time difference of each change feature relative to the regulatory action based on the execution time of the regulatory measure and the change time of each change feature, and comparing the actual response time difference with the corresponding reference response time difference to obtain the time response coupling result; The actual change direction of each change feature is compared with the corresponding reference change direction to obtain the directional response coupling result; The actual response sequence is formed based on the change time of each change characteristic, and the actual response sequence is compared with the reference response sequence to obtain the sequential response coupling result.

[0008] Preferably, the time response coupling results, directional response coupling results, and sequential response coupling results are combined according to their respective fusion weights to obtain the regulatory action-observation response coupling parameters.

[0009] Preferably, obtaining the cross-modal risk recovery inertia mismatch parameters includes: dividing the intervention impact features into fast-response risk features and slow-response risk features based on the response characteristics of each risk feature in the intervention impact features to the regulatory measures; and determining the corresponding fast-response risk status and slow-response risk status based on the fast-response risk features and slow-response risk features before and after the implementation of the regulatory measures.

[0010] Preferably, the fast response risk recovery rate is determined based on the fast response risk status before and after the implementation of regulatory measures, and the slow response risk recovery rate is determined based on the slow response risk status before and after the implementation of regulatory measures. The difference between the fast response risk recovery rate and the slow response risk recovery rate is compared to obtain the cross-modal risk recovery inertia mismatch parameter.

[0011] Preferably, the human intervention misjudgment index is obtained by: integrating the regulatory action-observation response coupling parameters and the cross-modal risk recovery inertia mismatch parameters, and further combining the joint relationship between the two to obtain the human intervention misjudgment index.

[0012] Preferably, the risk characteristics participating in the risk assessment are re-fused according to the adjusted fusion weights to obtain the intervention-corrected risk value, and the intervention-corrected risk value is input into the same risk level determination relationship as when the initial risk level was determined, so as to redetermine the risk level of the target fish species under supervision. Based on the correspondence between each risk characteristic and risk source, the adjusted fusion weights and risk characterization values ​​corresponding to risk characteristics belonging to the same risk source are aggregated to determine the degree of risk contribution of each risk source to the current intervention to correct the risk state, and the current main risk source is determined based on the degree of risk contribution.

[0013] Preferably, the redefined risk level and the main source of risk are used together as the basis for regulatory decision-making adjustments. Based on the pre-established correspondence between risk level, risk source and regulatory measure execution parameters, the intensity, duration and review frequency of regulatory measures that match the current risk status are determined. Supervision continues to be implemented based on the adjusted intensity, duration and review frequency of regulatory measures.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention differs from existing technologies that update risk levels solely based on changes in sensor values. Instead, it constructs a regulatory action-observation response coupling relationship by recording the execution time and intensity of regulatory measures, and combining this with the timing, direction, and sequence of changes in intervention impact characteristics. Furthermore, it categorizes risk characteristics into fast-response and slow-response risk characteristics, comparing the recovery differences between the two types of risks before and after regulatory measures are implemented. This allows for the identification of intervention scenarios where sensor data has changed rapidly, but the fish's condition or persistent risk has not yet recovered synchronously. It avoids directly interpreting transient observational changes caused by regulatory measures such as oxygenation and water changes as genuine risk improvement or deterioration, thus enhancing the accuracy and stability of risk status identification.

[0015] Furthermore, this invention does not simply shield relevant data after identifying human intervention. Instead, it dynamically adjusts the fusion weights of the intervened risk characteristics based on the degree of coupling between regulatory action and observation response and the degree of mismatch in cross-modal risk recovery inertia. Based on this, it recalculates the intervention-corrected risk value and then adjusts the intensity, duration, and review frequency of regulatory measures in conjunction with the corrected risk level and the main sources of risk. This invention enables the reverse impact of regulatory measures on sensor data to be continuously weakened and corrected through feedback, avoiding erroneous increases or decreases in risk levels and excessive or premature implementation of regulatory measures. This forms a closed-loop regulatory process that seamlessly integrates risk identification, intervention judgment, risk correction, and regulatory decision adjustment. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of the system modules of the present invention.

[0018] Figure 2 This is a flowchart of the feature extraction process for the intervention effects of this invention.

[0019] Figure 3 This is a flowchart illustrating the determination of coupling parameters between regulatory actions and observation responses in this invention.

[0020] Figure 4 This is a flowchart illustrating the determination of inertial mismatch parameters for cross-modal risk recovery in this invention.

[0021] Figure 5 This is a flowchart illustrating the contribution of risk sources to the main risk sources in this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1, please refer to Figure 1 As shown in this embodiment, the fish regulatory decision-making system based on multi-source data fusion and risk level assessment includes: The data acquisition module acquires aquatic environment data, fish behavior data, aquaculture process data, and historical regulatory data of the target fish species. After time-series alignment and fusion, a comprehensive risk feature set is obtained, and an initial risk level is determined. Regulatory measures are matched according to the initial risk level, and the execution time and intensity of the regulatory measures are obtained. The changes in the comprehensive risk feature set before and after the execution are extracted to obtain the intervention impact feature set.

[0024] The process involves acquiring aquatic environment data, fish behavior data, aquaculture process data, and historical regulatory data for the target fish species. This data is then time-series aligned and fused to obtain a comprehensive risk feature set, and an initial risk level is determined. The specific steps include: Specifically, the fish farmed species subject to regulatory decisions are designated as the target fish species for regulatory oversight. Multi-source data related to the current farming status and past regulatory activities of these target fish species are acquired. Specifically, the aquatic environment data characterizes the environmental state of the water body in which the target fish are located; the fish behavior data characterizes the activity status and behavioral changes of the fish population; the farming process data characterizes the production management information generated during the farming process of the target fish; and the historical regulatory data characterizes the regulatory information generated during the past regulatory processes of the target fish species. These data from different sources are then correlated according to the target fish species to form a corresponding set of original regulatory data.

[0025] Considering the differences in collection time and update cycle of regulatory data from different sources, the original regulatory data set undergoes time-series alignment processing. This time-series alignment process establishes a temporal correspondence between water environment data, fish behavior data, aquaculture process data, and historical regulatory data belonging to the same analysis period, enabling data from different sources to collectively characterize the status of the target fish species within the corresponding analysis period. After time-series alignment, information from various data types that characterizes the risk status of the target fish species is fused, forming a unified data expression based on corresponding risk characteristics, resulting in a comprehensive risk feature set corresponding to the corresponding analysis period. Each comprehensive risk feature set for each analysis period is characterized using corresponding risk feature dimensions, allowing for comparison of the same risk feature over time. This comprehensive risk feature set comprehensively characterizes the risk status of the target fish species within the corresponding analysis period, retaining risk features formed from regulatory data from different sources as the data basis for subsequent risk level determination and analysis of risk feature changes before and after the implementation of regulatory measures.

[0026] Furthermore, a risk assessment is conducted on the target fish species based on the comprehensive risk feature set. The risk level corresponding to each comprehensive risk feature is determined according to a preset risk level determination relationship, and an initial risk level for the target fish species is determined based on this risk level. This initial risk level serves as the initial determination of the risk status of the target fish species before the implementation of regulatory measures and is used for matching subsequent regulatory measures.

[0027] Specifically, regulatory measures are matched based on the initial risk level to obtain the execution time and intensity of the regulatory measures, and the changes in the comprehensive risk feature set before and after execution are extracted to obtain the intervention impact feature set. Please refer to [link to relevant documentation]. Figure 2 As shown, it specifically includes: Specifically, based on the initial risk level of the target fish species determined in the aforementioned steps, a regulatory measure corresponding to the initial risk level is determined from a pre-defined regulatory measure correspondence, and corresponding execution information is obtained after the regulatory measure is implemented. The execution information includes at least the execution time and execution intensity. The execution time characterizes the time at which the regulatory measure begins to intervene in the target fish species; the execution intensity characterizes the degree of effectiveness of the current regulatory measure and is determined based on the execution records of the corresponding type of regulatory measure. The execution time and execution intensity serve as the regulatory action-side input for subsequent analysis of the correlation between the regulatory measures and observed changes in risk characteristics.

[0028] Furthermore, using the execution time of the aforementioned regulatory measures as a time reference, comprehensive risk feature sets are obtained before and after the implementation of the regulatory measures. Since the comprehensive risk feature sets corresponding to each of the aforementioned analysis periods are expressed using corresponding risk feature dimensions, the same risk feature before and after the implementation of the regulatory measures is compared to identify risk features whose status has changed before and after the implementation of the regulatory measures.

[0029] For each identified change characteristic, the corresponding change time and change direction are determined. The change time characterizes the temporal position of the corresponding risk characteristic's state change before and after the implementation of regulatory measures; the change direction characterizes the relationship of increase, decrease, or stabilization of the corresponding risk characteristic relative to its state before the implementation of regulatory measures.

[0030] Furthermore, based on the time sequence of each change characteristic, the response sequence of different change characteristics after the implementation of regulatory measures is determined, and the response order corresponding to each change characteristic is obtained accordingly.

[0031] Therefore, for the first in the set of intervention impact characteristics Each change feature can generate corresponding change information: ;in, Indicates the first One characteristic of change, Indicates the time of change of the aforementioned change characteristic. Indicates the direction of change of the aforementioned characteristic. This indicates the response sequence determined based on the timing of changes in each characteristic.

[0032] By compiling the various change characteristics and their corresponding change information, we obtain: ;in, This represents the set of intervention impact characteristics corresponding to the current regulatory measures. This indicates the number of extracted variation features.

[0033] It should be noted that the intervention impact feature set is used to represent a set of candidate risk features that have changed before and after the implementation of regulatory measures and may be affected by the regulatory measures. It does not directly assume that each changed feature is caused by the regulatory measures in this step. The actual correlation between each changed feature and the regulatory measures will be further determined in subsequent steps based on the implementation time and intensity of the regulatory measures, as well as the change time, direction of change, and response order of each feature in the intervention impact feature set.

[0034] At the same time, information on the implementation of regulatory measures will be provided: ; and the intervention impact feature set Establish a corresponding relationship, where, Indicates the timing of the implementation of regulatory measures. This indicates the intensity of regulatory measures in effect, thereby generating information on regulatory actions and observed changes in response to current regulatory events.

[0035] The parameter calculation module obtains the regulatory action-observation response coupling parameters based on the execution time, execution intensity, and the change time, change direction, and response sequence of each feature in the intervention impact feature set. It also divides the intervention impact features into fast response risk features and slow response risk features. Based on the difference in recovery rate between the two types of risk features before and after the implementation of regulatory measures, it obtains the cross-modal risk recovery inertia mismatch parameters.

[0036] Specifically, please refer to Figure 3 As shown, the intervention impact feature set obtained in the previous step and the execution information corresponding to the current regulatory measures are used as inputs. The regulatory measure execution information includes the execution time of the regulatory measures. and execution intensity For the first in the intervention impact feature set Each change characteristic determines the timing of its change. , change direction and response order Among them, the time of change This indicates the moment when the change characteristic shows a significant change in state relative to the implementation process of regulatory measures; the direction of the change. This indicates that the change characteristic, relative to the state before the implementation of regulatory measures, shows an increasing, decreasing, or stabilizing relationship; response order. The results are obtained by sorting the changes of each characteristic in chronological order.

[0037] To provide a basis for comparison of the above-mentioned actual observation results, a reference response relationship corresponding to the current regulatory measures is obtained. This reference response relationship is determined based on historical implementation information of similar types of regulatory measures and their corresponding changes in risk characteristics, combined with the enforcement intensity of the current regulatory measures. A reference response relationship corresponding to or similar to the current execution intensity is selected. This reference response relationship includes at least the reference response time difference, reference change direction, and reference response order corresponding to each relevant feature, expressed as: ; in, For the first Reference response time lag of each change characteristic under the corresponding regulatory measures For reference, the direction of change, This serves as a reference response order. Therefore, the intensity of regulatory measures is not an independently suspended parameter, but rather a reference response relationship used to determine the appropriateness of the current regulatory action, providing a corresponding benchmark for the observed responses under different intensity levels.

[0038] Furthermore, based on the current timing of regulatory measures implementation With the The moment of change of each characteristic Calculate the actual response time difference: ; The actual response time difference Time difference with reference response By comparing the time response of each change feature, the consistency of the response over time is determined, and the time response results of each change feature are combined to obtain the time response coupling result. For example, it can be represented as: ; in, Indicates the first The weight of each change feature in time response analysis This indicates the permissible deviation of the response time difference for a corresponding changing characteristic. The closer the actual response time difference is to the corresponding reference response time difference, the higher the consistency of the time response of the corresponding characteristic, thus... The greater the deviation between the actual response time difference and the reference response time difference, the better; conversely, the greater the deviation, the better. The corresponding reduction.

[0039] Furthermore, the actual direction of change of each change feature in the intervention impact feature set will be determined. Its reference direction of change By comparison, the directional response coupling results are obtained. , represented as: ; in, This is a direction consistency judgment function. It obtains a consistent result when the actual direction of change matches the reference direction of change; otherwise, it obtains an inconsistent result. Therefore, This is used to characterize the overall consistency between the actual direction of change of multiple change features and the expected direction of the current regulatory measures, avoiding the conclusion that a change is driven by regulatory measures simply because a single risk feature has changed.

[0040] At the same time, the actual response sequence is formed according to the timing of the changes in each characteristic from first to last: ; And a reference response sequence is formed based on the aforementioned reference response relationship: ; Compare actual response sequences With reference response sequence Determine the sequence of changes in the response characteristics to identify differences in response order. And obtain the sequential response coupling results: ; in, This indicates the degree of difference between the actual response order and the reference response order. This indicates the degree of difference between the actual response order and the reference response order. The smaller, the corresponding The larger.

[0041] Therefore, the current regulatory measures result in an intermediate coupling of time-based, directional, and sequential responses: ; in, This indicates whether each characteristic changes at a time and location consistent with regulatory actions. This indicates whether each characteristic changes in a direction consistent with regulatory actions. This characterizes whether multiple features respond in a sequence consistent with regulatory actions. The three results describe the correlation between the same regulatory action and observed changes from different dimensions.

[0042] In obtaining the time response coupling result Directional response coupling results Coupling results with sequential response Then, the three are fused to obtain the regulatory action-observation response coupling parameters. The evaluation expression can be: ;in, These are the fusion weights corresponding to the time response, directional response, and sequential response, respectively, and satisfy the following conditions: The aforementioned This is used to quantify the overall response coupling between observed changes in the current intervention impact feature set and regulatory measures. The larger the value, the more the multiple change characteristics conform to the reference response relationship corresponding to the current regulatory measures in terms of the timing of change, direction of change, and sequence of response. In other words, the higher the degree to which the change of the current risk characteristics is affected by the regulatory action. The smaller the value, the more likely the change in the current risk characteristics cannot be fully explained by the process of the regulatory measure's action.

[0043] For example, when the target fish species under surveillance are in a high-risk state, oxygenation measures are implemented, with the start time of the oxygenation measures as the starting point. And determine the corresponding reference response relationship based on its implementation intensity. After the implementation of regulatory measures, the intervention impact characteristics are concentrated in the changes of aquatic environmental characteristics and fish behavior characteristics. If the actual change time of each characteristic is relative to the time of the intervention, the reference response relationship is determined accordingly. If the actual response time difference is close to the corresponding reference response time difference, the actual increase or decrease direction of each feature is consistent with the reference change direction, and the order of the actual changes of multiple features is basically consistent with the order of the reference response, then higher values ​​are obtained. , and This leads to higher regulatory action-observation response coupling parameters. If some changes occur after the implementation of regulatory measures, but their response time deviates significantly from the reference response time, or the direction of change and response sequence are inconsistent with the reference response, then the corresponding coupling result will decrease, thereby reducing the final outcome. .

[0044] The intervention impact characteristics are divided into fast-response risk characteristics and slow-response risk characteristics. Based on the difference in recovery rates before and after the implementation of regulatory measures for these two types of risk characteristics, a cross-modal risk recovery inertia mismatch parameter is obtained. (See [link to relevant documentation]). Figure 4 As shown, it specifically includes: Specifically, this step uses the intervention impact feature set As the object of processing. Specifically, based on the response characteristics of each risk characteristic in the intervention impact characteristics to regulatory measures, they are divided into fast-response risk characteristics and slow-response risk characteristics. The fast-response risk characteristics refer to risk characteristics that can produce significant changes within a relatively short period after the implementation of regulatory measures, and whose current state is easily directly affected by the regulatory measures, forming a set of fast-response risk characteristics: The slow response risk characteristic refers to a risk characteristic whose state change has a certain persistence and will not recover synchronously with the fast response characteristic after the implementation of regulatory measures, forming a slow response risk characteristic set: Among them, fast-response risk features are mainly used to characterize the immediate state that regulatory measures can directly change, while slow-response risk features are used to characterize the continuous changes in the risk state of individual fish, fish schools, or accumulated risks. This transforms the same set of intervention impact features into two subsets of features with different recovery inertia.

[0045] Furthermore, the rapid response risk status is calculated based on the rapid response risk characteristics before and after the implementation of regulatory measures. and And calculate the slow response risk status based on the slow response risk characteristics before and after the implementation of regulatory measures. and .in, This indicates that before the regulatory measures are implemented, This indicates the period following the implementation of regulatory measures. Based on the aforementioned risk status, the rapid response risk recovery rate is determined accordingly. and slow response risk recovery rate , can be represented as: ; in, This is a correction amount used to avoid the denominator being zero; This indicates the degree of recovery from the pre-implementation state in response to risks associated with the rapid response to regulatory measures. This indicates the degree of recovery of slow response risk relative to the pre-execution state. Both are calculated using the same risk change direction to ensure comparability of recovery levels.

[0046] In obtaining and Then, the cross-modal risk recovery inertial mismatch parameters are calculated. ;in, Used to characterize the degree to which the speed of risk recovery in a fast response is ahead of the speed of risk recovery in a slow response. When and When they approach each other, it indicates that the immediate risk status and the persistent risk status change essentially in sync after the implementation of regulatory measures. Smaller; when Significantly greater than This indicates that the risks associated with rapid response have shown significant improvement, while the risks associated with slow response have not yet recovered accordingly. The corresponding increase. Therefore, the aforementioned It does not reflect the level of risk alone, but is used to describe the differences in recovery inertia between different risk modes.

[0047] For example, if the aquatic environment-related rapid response risk decreases rapidly from a high level after oxygenation measures are implemented for the target fish species, then... The risk of slow response corresponding to the continuous abnormal state of the fish population only decreased slightly, making ,but: ; This indicates that the recovery of fast-response risks significantly outpaces that of slow-response risks, and the currently observed risk improvements exhibit a noticeable asynchronous recovery across modalities. Conversely, if the recovery rates of the two types of risks are similar, it suggests a high degree of consistency between apparent risk improvement and persistent risk improvement.

[0048] This application, by acquiring regulatory action-observation response coupling parameters and cross-modal risk recovery inertia mismatch parameters, can characterize post-intervention risk changes from two different dimensions: "the degree of correlation between regulatory measures and observed changes" and "whether the recovery of fast and slow risk states is synchronized." This can identify whether changes in sensors and risk characteristics may be directly caused by regulatory measures, and determine whether the apparent risk improvement is consistent with the actual recovery of the fish body or persistent risk state. This avoids judging the true risk state based solely on a single sensor change or a single risk value, providing a more reliable parameter basis for subsequently identifying risk misjudgments caused by human intervention, correcting risk levels, and dynamically adjusting regulatory measures.

[0049] For the weight adjustment module, please refer to [link / reference]. Figure 5As shown, the human intervention misjudgment index is obtained based on the regulatory action-observation response coupling parameters and the cross-modal risk recovery inertia mismatch parameters. The fusion weights corresponding to the intervention impact features are adjusted based on the human intervention misjudgment index, and the intervention correction risk value is recalculated based on the adjusted fusion weights.

[0050] Specifically, firstly, regarding and By combining the results, we obtain the human intervention misjudgment index. , can be represented as: ;in, These represent the weights of the coupling degree between regulatory action and observation response, the degree of mismatch in cross-modal risk recovery inertia, and the joint relationship between the two in the misjudgment analysis, respectively, and satisfy the following conditions: .in, This is used to characterize the joint intervention features produced when both parameters are at high levels simultaneously. A higher level indicates a strong correlation between observed changes and regulatory actions; when When both are high, it indicates that the risk of rapid response has changed significantly, while the risk of slow response has not changed synchronously. Therefore, when both increase simultaneously, the degree to which the current risk change is caused by regulatory measures and may be mistakenly identified as a true risk change in the risk assessment process increases accordingly. This leads to the resulting human intervention misjudgment index. This is a key intermediate result in this step.

[0051] Furthermore, based on the aforementioned human intervention misjudgment index The original fusion weights of each risk feature in the intervention impact feature set are adjusted. For the first... The adjusted fusion weights of the intervention impact features can be expressed as follows: ;in, For the first The fusion weight of each intervention impact feature in the original risk fusion process The fusion weights are those that have been modified by human intervention. This indicates the degree to which the impact of the intervention is affected by current regulatory measures. With As the risk profile increases, the fusion weight corresponding to the risk characteristics that are more significantly affected by regulatory measures decreases accordingly; when When the values ​​are relatively small, the corresponding adjustment range of the fusion weights is also relatively small. This results in a weight correction result that includes the characteristics of each intervention impact and their adjusted fusion weights, thereby suppressing the contribution of observational changes that may be affected by human intervention in the subsequent comprehensive risk calculation.

[0052] After obtaining the adjusted fusion weights, each risk feature is re-fused according to the risk fusion method corresponding to the aforementioned comprehensive risk feature set to obtain the intervention-corrected risk value. , can be represented as: ;in, Indicates the first The risk characterization value corresponding to each risk feature This indicates its adjusted fusion weight. This indicates the number of risk features involved in risk fusion. It is used to characterize the overall risk status of target fish species after reducing the impact of human regulatory intervention.

[0053] For example, when regulatory measures are implemented, a higher level of achievement is obtained. Meanwhile, the risks associated with a fast response have recovered significantly, while those associated with a slow response have changed little, making... When both are at a high level, a higher value is obtained from both. At this point, the fusion weight of relevant intervention impact characteristics that are significantly affected by regulatory measures is reduced, and the comprehensive risk is recalculated using the adjusted weights, thereby avoiding the excessive dominance of transient changes in relevant characteristics in the risk assessment results.

[0054] The regulatory decision adjustment module redetermines the risk level based on the intervention correction risk value, and dynamically adjusts the intensity, duration, and review frequency of regulatory measures based on the redetermined risk level and risk source.

[0055] Each risk feature maintains a correspondence with its risk source when forming a comprehensive risk feature set, where the risk source characterizes the risk category to which the corresponding risk feature belongs. Thus, the intervention-corrected risk value is obtained. Subsequently, it is possible to simultaneously obtain the risk contribution of each risk source in the intervention-corrected risk value without having to acquire new risk data.

[0056] First, adjust the intervention risk value. Input the same risk level determination relationship used when determining the initial risk level to redetermine the risk level of the target fish species under supervision. By adopting a judgment relationship consistent with the initial risk level, the risk levels before and after the intervention correction have a unified judgment benchmark, thereby enabling the redefined risk level to reflect the risk status after the impact of human intervention has been weakened.

[0057] Furthermore, based on the adjusted fusion weights and corresponding risk characterization values ​​for each risk characteristic, the contribution of different risk sources to the current risk state is determined. For the first... Each risk source, and its risk contribution value, can be expressed as: ;in, Indicates belonging to the first A set of risk characteristics from each risk source This indicates that the weight adjustment module determined the first... Adjusted fusion weights for each risk characteristic, This represents the risk characterization value corresponding to the risk characteristics. Indicates the first The degree to which each risk source contributes to the current intervention to correct the risk status.

[0058] Comparison of different sources of risk The sources of risk that contribute the most to the risk assessment are identified as the primary sources of risk at present. This leads to the current regulatory decision-making status corresponding to the regulated entity: Among them, the redefined risk level Characterizes the overall severity of the current risk, and the main sources of risk. This characterizes the primary risk direction corresponding to the current risk. Both serve as inputs for adjusting regulatory measures, thus avoiding adjustments based solely on a single comprehensive risk value while ignoring the specific sources leading to the risk state.

[0059] Furthermore, for implemented regulatory measures, a pre-established correspondence between risk levels, risk sources, and regulatory measure implementation parameters is established. These implementation parameters include at least implementation intensity, duration, and review frequency. The current regulatory decision-making status is then considered. Matching with the aforementioned correspondence yields: ;in, This indicates the pre-defined correspondence between risk levels, risk sources, and regulatory measure implementation parameters; This indicates the intensity of enforcement of the adjusted regulatory measures. Indicates the adjusted duration. This indicates the adjusted review frequency. The parameter adjustment is based on current regulatory measures and refers to the redefined target parameters. Respectively related to the current enforcement intensity of regulatory measures Duration and review frequency The corresponding risk levels are then used to determine the subsequent implementation status of regulatory measures based on the differences between the two. The redefined risk level then defines the overall level of regulation required for the measures, while the risk sources are used to align the relevant implementation parameters with the current primary risk status.

[0060] For example, after regulatory measures are implemented, some intervention impact characteristics manifest as a rapid reduction in risk. However, after weighting and adjusting for the aforementioned regulatory action-observation response coupling parameters, cross-modal risk recovery inertia mismatch parameters, and human intervention misjudgment index, the resulting intervention-corrected risk value becomes... The risk level remains within a relatively high range. Simultaneously, the primary risk sources are determined based on their risk contribution values. At this point, the redefined risk level and primary risk sources are input into the regulatory measure parameter correspondence to obtain the execution intensity, duration, and review frequency that match the current risk status. This ensures that regulatory measures are not prematurely weakened simply because some sensor data recovers quickly after intervention. Conversely, if the risk level corresponding to the intervention-corrected risk value has indeed decreased, subsequent regulatory parameters are adjusted accordingly based on the redefined risk level and risk sources.

[0061] Finally, this step outputs the adjusted regulatory measure parameters: The regulatory measures will continue to be implemented based on the adjusted regulatory parameters. The corresponding execution results will be incorporated into the data acquisition process of the next regulatory cycle, and used together with subsequently acquired water environment data, fish behavior data, aquaculture process data, and historical regulatory data to reconstruct the comprehensive risk characteristics.

[0062] This invention addresses the problem that regulatory measures such as oxygenation, water exchange, and disinfection during fish supervision can adversely alter sensor data and risk characteristics, leading to risk models misinterpreting transient changes caused by human intervention as real risk changes. It establishes an initial risk state by integrating aquatic environment data, fish behavior data, aquaculture process data, and historical regulatory data. Furthermore, it constructs feature parameters from two directions: the "coupling degree between regulatory actions and observed changes" and the "asynchrony in recovery between fast-response and slow-response risks." This further quantifies the degree of misjudgment regarding human intervention. Based on this, the fusion weights of the intervened risk characteristics are dynamically adjusted, the intervention-corrected risk value is recalculated, and the risk level is determined. Finally, based on the corrected risk level and the main sources of risk, the intensity, duration, and review frequency of regulatory measures are dynamically adjusted, thus forming a closed-loop regulatory mechanism of risk identification, intervention impact judgment, risk correction, and regulatory decision adjustment.

[0063] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A fish regulatory decision-making system based on multi-source data fusion and risk level assessment, characterized in that, include: The data acquisition module acquires water environment data, fish behavior data, aquaculture process data, and historical regulatory data of the target fish species. After time-series alignment and fusion, a comprehensive risk feature set is obtained, and an initial risk level is determined. Regulatory measures are matched according to the initial risk level, the execution time and intensity of the regulatory measures are obtained, and the change features in the comprehensive risk feature set before and after the execution are extracted to obtain the intervention impact feature set. The parameter calculation module obtains the regulatory action-observation response coupling parameters based on the execution time, execution intensity, and the change time, change direction, and response sequence of each feature in the intervention impact feature set; and divides the intervention impact features into fast response risk features and slow response risk features, and obtains the cross-modal risk recovery inertia mismatch parameters based on the difference in recovery rate between the two types of risk features before and after the implementation of regulatory measures. The weight adjustment module obtains the human intervention misjudgment index based on the regulatory action-observation response coupling parameters and cross-modal risk recovery inertia mismatch parameters, adjusts the fusion weights corresponding to the intervention impact features based on the human intervention misjudgment index, and recalculates the intervention correction risk value based on the adjusted fusion weights. The regulatory decision adjustment module redetermines the risk level based on the intervention correction risk value, and dynamically adjusts the intensity, duration, and review frequency of regulatory measures based on the redetermined risk level and risk source.

2. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 1, characterized in that, Obtaining the intervention impact feature set includes: using the execution time of the regulatory measures as a time reference, obtaining comprehensive risk feature sets before and after the execution of the regulatory measures; comparing the same risk features before and after execution to determine the risk features that have undergone state changes; determining the change time and direction of each change feature, and sorting them according to the chronological order of the change time of each change feature to determine the response order of each change feature; and associating and aggregating each change feature and its corresponding change time, change direction, and response order to obtain the intervention impact feature set.

3. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 1, characterized in that, Obtaining regulatory action-observation response coupling parameters includes: acquiring a reference response relationship corresponding to the current regulatory measure, wherein the reference response relationship is determined based on historical execution information of the same type of regulatory measures and corresponding risk characteristic change information; Based on the intensity of the current regulatory measures, a reference response relationship corresponding to or similar to the current intensity of implementation is determined from the reference response relationship. The reference response relationship includes the reference response time difference, reference change direction, and reference response order corresponding to each change feature, which serves as a comparison benchmark for the response correlation analysis between the current intervention impact features and regulatory measures.

4. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 3, characterized in that, The method of obtaining the regulatory action-observation response coupling parameters further includes: determining the actual response time difference of each change feature to the regulatory action based on the execution time of the regulatory measures and the change time of each change feature, and comparing the actual response time difference with the corresponding reference response time difference to obtain the time response coupling result; The actual change direction of each change feature is compared with the corresponding reference change direction to obtain the directional response coupling result; The actual response sequence is formed based on the change time of each change characteristic, and the actual response sequence is compared with the reference response sequence to obtain the sequential response coupling result.

5. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 4, characterized in that, The time response coupling results, directional response coupling results, and sequential response coupling results are combined according to their respective fusion weights to obtain the regulatory action-observation response coupling parameters.

6. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 5, characterized in that, Obtaining cross-modal risk recovery inertia mismatch parameters includes: dividing the intervention impact features into fast-response risk features and slow-response risk features based on the response characteristics of each risk feature in the intervention impact features to regulatory measures; and determining the corresponding fast-response risk status and slow-response risk status based on the fast-response risk features and slow-response risk features before and after the implementation of regulatory measures.

7. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 6, characterized in that, The fast response risk recovery rate is determined based on the fast response risk status before and after the implementation of regulatory measures, and the slow response risk recovery rate is determined based on the slow response risk status before and after the implementation of regulatory measures. The difference between the fast response risk recovery rate and the slow response risk recovery rate is compared to obtain the cross-modal risk recovery inertia mismatch parameter.

8. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 7, characterized in that, The human intervention misjudgment index is obtained by integrating the regulatory action-observation response coupling parameters and the cross-modal risk recovery inertia mismatch parameters, and further combining the joint relationship between the two.

9. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 8, characterized in that, The risk characteristics involved in the risk assessment are re-integrated according to the adjusted fusion weights to obtain the intervention-corrected risk value. The intervention-corrected risk value is then input into the same risk level determination relationship as when the initial risk level was determined to redetermine the risk level of the target fish species under supervision. Based on the correspondence between each risk characteristic and risk source, the adjusted fusion weights and risk characterization values ​​corresponding to risk characteristics belonging to the same risk source are aggregated to determine the degree of risk contribution of each risk source to the current intervention to correct the risk state, and the current main risk source is determined based on the degree of risk contribution.

10. The fish regulatory decision-making system based on multi-source data fusion and risk level assessment according to claim 9, characterized in that, The redefined risk level and the main source of risk will be used as the basis for regulatory decision-making adjustments. Based on the pre-established correspondence between risk level, risk source and regulatory measure execution parameters, the intensity, duration and review frequency of regulatory measures that match the current risk status will be determined. Supervision will continue to be implemented based on the adjusted intensity, duration and review frequency of regulatory measures.

Images