Breeding operation state modeling system and method driven by bionic fish sensing data

Abstract

The invention relates to the technical field of aquaculture operation state sensing and management, and discloses an aquaculture operation state modeling system and method driven by bionic fish sensing data. Comprising the following steps: continuously sensing and collecting water body disturbance caused by fish school activities, a fish body space migration track and local environment response influenced by behavior state change to form a sensing data set; constructing a behavior change description sequence of time recursion along a fish body motion path direction, and generating a multi-scale behavior evolution feature expression; extracting candidate state segments by identifying structural inflection points of transition from stability to stress of a behavior mode; mapping to a corresponding breeding operation environment unit, and generating a local operation state association structure; and performing adaptive adjustment on the acquisition density and the feature weight of the bionic fish sensing data in different water areas and time periods to form a state characterization result of the current culture operation situation. The method has the advantage of improving the timeliness and accuracy of breeding operation state recognition.

Description

Technical Field

[0001] This invention relates to the field of aquaculture operation status perception and management technology, specifically to a biomimetic fish perception data-driven aquaculture operation status modeling system and method. Background Technology

[0002] Currently, in the operation and management of aquaculture, fixed water quality sensors are typically deployed in the aquaculture water body to continuously monitor environmental parameters such as water temperature, dissolved oxygen, and pH value. The aquaculture operation status is then determined based on preset thresholds. However, in actual aquaculture scenarios, especially in outdoor ponds or factory-style recirculating aquaculture environments, short-term, localized microscale environmental anomalies often occur due to factors such as uneven water flow, concentrated feeding, and increased local biological oxygen consumption. These anomalies often do not cause the overall water quality parameters to exceed the alarm threshold, but they have already caused obvious physiological stress and behavioral changes in the fish, such as abnormal swimming speed, gathering and surfacing, or avoiding specific water areas. Existing technologies mainly rely on environmental parameters at fixed points for status determination, making it difficult to detect the above-mentioned operational anomalies led by fish responses in a timely manner. This can easily lead to a lag in the identification of aquaculture operation status, resulting in missed opportunities for optimal intervention, increasing the risk of fish stress and aquaculture losses. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a biomimetic fish perception data-driven aquaculture operation status modeling system and method, which has the advantages of improving the timeliness and accuracy of aquaculture operation status identification and solves the problems mentioned in the background technology.

[0004] To achieve the aforementioned goal of improving the timeliness and accuracy of aquaculture operation status identification, this invention provides the following technical solution: a biomimetic fish perception data-driven aquaculture operation status modeling method, comprising the following steps: Continuous sensing and data collection are performed on water disturbances caused by fish activity, fish spatial migration trajectories, and local environmental responses affected by changes in behavioral state. Based on the fish activity range and the spatial continuity characteristics of the water body, the collected raw sensing dataset is spatially reorganized to form a sensing data set that can be expanded along the fish movement path. Based on the sensory data set, a time-recursive behavioral change description sequence is constructed along the fish movement path. The historical activity baseline of the fish group and the behavioral offset within the current collection period are introduced to generate a multi-scale behavioral evolution feature expression. Based on the multi-scale behavioral evolution characteristics, by identifying the structural inflection point of the transition from stable to stress behavior patterns, candidate state fragments reflecting the emergence of abnormal aquaculture operation status are extracted. Candidate state fragments are mapped to corresponding aquaculture operation environment units, and the correspondence between candidate state fragments and aquaculture operation environment is established by combining water circulation mode, aquaculture density distribution and recent operation records, thus generating a local operation state association structure. Based on the local operational status association structure, the collection density and feature weights of biomimetic fish perception data in different water areas and time periods are adaptively adjusted to form the current state representation of aquaculture operation.

[0005] Preferably, the process of continuously sensing and collecting data on water disturbances caused by fish activity, fish spatial migration trajectories, and local environmental responses affected by changes in behavioral states is as follows: A biomimetic fish sensing unit with autonomous swimming ability is deployed to simultaneously acquire water flow velocity perturbation signals, fish relative displacement information and corresponding environmental response parameters during its cruising movement in different water depths and spatial areas. The data collected by each biomimetic fish sensing unit is timestamped and cached locally, and the time consistency correction of the sensing data of multiple units is completed based on the wireless synchronization mechanism. The time-corrected sensing data is uploaded to the aquaculture operation management node to form a raw sensing dataset containing water disturbance, fish migration, and environmental response.

[0006] Preferably, the process of forming a set of sensory data that can be deployed along the fish's movement path is as follows: Based on the spatial positioning information of the bionic fish sensing unit, the spatial location of the water body corresponding to each sensing data and the associated activity range are determined. Based on the spatial connectivity of aquaculture water bodies, path association mapping and regional merging are performed on the original sensing data collected at different times and spatial locations. The merged sensory dataset is sequentially aligned and expanded according to the actual movement path of the bionic fish, generating a sensory data set that can be expanded along the movement path of the fish.

[0007] The preferred process for constructing a time-recursive behavioral change description sequence along the fish's movement path is as follows: Using the continuous spatial nodes on the fish's movement path corresponding to the sensing data set as indexes, the corresponding data are arranged in chronological order. Between adjacent time slices, based on the data differences between adjacent nodes in the sensing dataset, the changes in fish movement state, water disturbance intensity, and environmental response parameters are calculated. By concatenating the changes in a time-recursive manner, the spatially continuous data in the dataset is transformed into a temporal evolution description, forming a sequence of behavioral changes that characterize the fish school's behavior over time.

[0008] The preferred process for generating multi-scale behavioral evolution feature representations is as follows: Extract the normal behavioral characteristics of fish populations under water and seasonal conditions corresponding to the behavioral change description sequences from historical aquaculture operation data, and construct a baseline of historical fish population activity. The current behavioral change description sequence is compared with the historical activity baseline time by time, and the offset of the behavioral change description sequence relative to the normal state in each time period is calculated to form a behavioral offset sequence arranged by time index. Based on the behavioral offset sequence, aggregation and expansion processes are performed at different time scales to generate a multi-scale behavioral evolution feature expression that reflects the behavioral evolution characteristics of fish groups.

[0009] Preferably, the process of extracting candidate state fragments reflecting the abnormal germination of aquaculture operation status is as follows: Continuous time window analysis was performed on the multi-scale behavioral evolution feature representation to obtain the stability index of behavioral features changing over time at different time scales; Based on stability indicators, we identify abrupt changes in behavioral characteristics over time, and combine these with behavioral shifts within the corresponding time period to determine whether the abrupt changes conform to the characteristic pattern of transition from a stable state to a stress state. The time periods corresponding to mutation points that are determined to conform to the transitional feature pattern are extracted from the multi-scale behavioral evolution feature expression to form candidate state fragments reflecting the abnormal emergence of aquaculture operation status.

[0010] Preferably, the process of mapping candidate state fragments to corresponding aquaculture operating environment units is as follows: Based on the path-expanded perception data corresponding to the candidate state fragments in the perception data set, the spatial movement path of the bionic fish during the occurrence of abnormal behavior and its continuous distribution characteristics in the time dimension are extracted to determine the main spatial coverage of abnormal behavior in the aquaculture water. Based on the degree of concentration of the main spatial coverage area in the water space and the spatial adjacency relationship between the aquaculture operation structure, the main spatial coverage area is matched with the functional zones, equipment layout areas or water management units in the aquaculture water body. From the matching results, select aquaculture operation environment units that are consistent with the spatial characteristics of abnormal behavior, and establish a spatial correspondence between candidate state segments and corresponding aquaculture operation environment units.

[0011] Preferably, the process of generating the local running state association structure is as follows: Acquire water circulation parameters, stocking density information, and recent operation records corresponding to the aquaculture operation environment unit to form an operation information set characterizing the current operating conditions; The candidate state fragments in the set of operational information and spatial correspondence are associated and matched in the time and spatial dimensions to analyze the influence of different operating conditions on the generation and evolution of abnormal behavior fragments. Based on the association matching results, a local operational state association structure is constructed to describe the influence relationship between candidate state segments and aquaculture operation environment units.

[0012] The preferred process for forming the current state characterization results of aquaculture operations is as follows: Based on the influence relationship between candidate state segments represented in the local operational state association structure and aquaculture operational environment units, the water area where abnormal behavior is concentrated and the corresponding risk level are identified. Based on the identified abnormal concentrated water areas and risk levels, the patrol frequency and sensing data collection density of the bionic fish sensing units in the corresponding water areas are adjusted. Based on the risk weights of candidate state segments in the local operational state association structure, differentiated feature weights are assigned to the sensing data collected in different time periods. By integrating the collected sensing data after adjusting for density and feature weights, a status representation result reflecting the current aquaculture operation status is generated.

[0013] A biomimetic fish perception data-driven aquaculture operation status modeling system, comprising: Continuous sensing module: continuously collects water disturbances caused by fish activity, fish migration trajectories and local environmental responses, and performs spatial assignment reshaping on the raw sensing data to form a sensing data set that can be expanded along the fish movement path. Behavior modeling module: Based on the sensory data set, construct a time-recursive behavioral change description sequence along the fish's movement path, and generate multi-scale behavioral evolution feature representations; Anomaly extraction module: Identifies structural inflection points in behavior transitioning from stability to stress from multi-scale behavioral evolution feature representation, and extracts corresponding candidate state fragments; Association Construction Module: Maps candidate state fragments to aquaculture operation environment units and generates a local operation state association structure describing the correspondence between the two. Situation characterization module: Based on the local operational status association structure, the collection density and feature weights of the sensing data are adaptively adjusted to form the current state characterization result of the aquaculture operation.

[0014] Compared with existing technologies, this invention provides a biomimetic fish perception data-driven aquaculture operation status modeling system and method, which has the following beneficial effects: This invention introduces a biomimetic fish-like sensing unit with autonomous movement capabilities to continuously sense and collect data on water disturbances caused by fish activity, fish spatial migration trajectories, and local environmental responses. The discrete sensing data is then path-reorganized and time-recursively modeled along the actual movement path of the fish, thus more realistically reflecting the dynamic characteristics of fish behavior evolving over time and space. Furthermore, by combining the historical activity baseline of the fish population with the current behavioral offset, a multi-scale behavioral evolution feature expression is constructed. By identifying structural inflection points in the transition from a stable state to a stress state, early detection of abnormalities in aquaculture operation is achieved. Simultaneously, spatial and operational condition correlations are established between abnormal behavioral segments and specific aquaculture operation environment units, elevating operational status analysis from simple behavioral judgment to "behavior-environment-operational" correlation modeling. Based on this, the sensing collection density and feature weights are adaptively adjusted, thereby improving the timeliness, accuracy, and relevance of aquaculture operation status representation. This overcomes the problems of lagging status recognition, coarse spatial positioning, and difficulty in correlating operational influencing factors in existing technologies. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method of the present invention; Figure 2 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0016] 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, and 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.

[0017] Example 1: Please refer to Figure 1 As shown in the figure, the biomimetic fish perception data-driven aquaculture operation status modeling method of this invention includes the following steps: S1: Continuously sense and collect data on water disturbances caused by fish activity, fish spatial migration trajectories, and local environmental responses affected by changes in behavioral states. Based on the fish activity range and the spatial continuity characteristics of the water body, the collected raw sensing dataset is spatially reorganized to form a sensing data set that can be expanded along the fish movement path.

[0018] The process of continuously sensing and collecting data in S1 on water disturbances caused by fish activity, fish spatial migration trajectories, and local environmental responses affected by changes in behavioral states is as follows: A biomimetic fish sensing unit with autonomous swimming ability is deployed to simultaneously acquire water flow velocity perturbation signals, fish relative displacement information and corresponding environmental response parameters during its cruising movement in different water depths and spatial areas. Bionic fish sensing units with autonomous swimming capabilities are deployed in the aquaculture water. These units can move along preset or adaptive trajectories in different water depths and spatial areas, while continuously collecting water flow velocity perturbation signals, relative displacement information of the fish, and local environmental response parameters, including dissolved oxygen, temperature, turbidity, and other indicators. Each bionic fish unit records its own spatial position and time information in real time, ensuring that the collected data corresponds one-to-one with the fish's movement state, thereby fully reflecting the impact of fish activity on water disturbance and the local environment.

[0019] The data collected by each biomimetic fish sensing unit is timestamped and cached locally, and the time consistency correction of the sensing data of multiple units is completed based on the wireless synchronization mechanism. The data collected by each bionic fish sensing unit is first timestamped locally and cached in the internal memory. At the same time, the collection period and sampling interval information are recorded. To ensure the consistency of data collected by multiple units, a wireless synchronization mechanism is adopted. The time of each unit is corrected through an underwater or surface wireless communication network, so that the signals collected by each bionic fish are aligned under the same time reference, eliminating data deviations caused by collection delays or communication jitter. This ensures that data from different units can be directly compared and fused during the processing.

[0020] The time-corrected sensing data is uploaded to the aquaculture operation and management node to form a raw sensing dataset containing water disturbance, fish migration and environmental response. After time consistency correction, the multi-unit sensing data is uploaded to the aquaculture operation management node via wireless or wired network. This node is responsible for integrating water disturbance data, fish migration trajectory information, and environmental response parameters into a structured raw sensing dataset. The dataset is stored according to time and space dimensions, recording the collection time, location, and corresponding sensing value of each data point, providing complete and traceable basic data for path merging, behavior sequence generation, and aquaculture operation status modeling.

[0021] The process of forming a set of sensory data that can be expanded along the fish's movement path in S1 is as follows: Based on the spatial positioning information of the bionic fish sensing unit, the spatial location of the water body corresponding to each sensing data and the associated activity range are determined. During the bionic fish sensing unit's navigation and data collection process, each data collection node records its real-time spatial positioning information in the water, including three-dimensional coordinates, depth, and the water zone it is located in. Based on the positioning information, each sensing data is mapped to its corresponding spatial location in the water. Combined with the bionic fish's swimming range and speed, the boundary of the activity area corresponding to each data is determined. These boundaries are used for subsequent path association and spatial merging to ensure that each data can accurately reflect its spatial coverage in the water, while providing a basic index for data to be expanded along the path.

[0022] Based on the spatial connectivity of aquaculture water bodies, path association mapping and regional merging are performed on the original sensing data collected at different times and spatial locations. Based on the spatial structure and connectivity of the aquaculture water body, the original sensing data collected at different times and spatial locations are path-correlated, and adjacent or overlapping spatial activity areas are identified as the same continuous path segment. Using water body grid division or three-dimensional spatial adjacency analysis, nodes that are close to each other are merged into the same area, and redundant and isolated data points are eliminated to generate a continuous spatial data set and ensure that the data is continuously traceable in space, providing a complete data chain for unfolding along the actual movement path of the biomimetic fish.

[0023] The merged sensory dataset is sequentially aligned and expanded according to the actual movement path of the bionic fish, generating a sensory data set that can be expanded along the movement path of the fish. The merged spatial dataset is arranged according to the actual movement path and timestamp information of the bionic fish. This aligns the data along the direction of the fish's movement, correcting deviations caused by acquisition delays or multi-unit synchronization errors. This ensures that the time and spatial location of each data point are accurately matched, and the resulting sensory dataset can be fully unfolded along the fish's movement path. This preserves spatial continuity while also taking into account the temporal progression, providing a directly usable data foundation for behavioral change analysis and state modeling.

[0024] S2: Based on the sensory data set, a time-recursive behavioral change description sequence is constructed along the fish's movement path. The historical activity baseline of the fish group and the behavioral offset within the current collection period are introduced to generate a multi-scale behavioral evolution feature expression.

[0025] The process of constructing a time-recursive behavioral change description sequence along the fish's movement path in S2 is as follows: Using the continuous spatial nodes on the fish's movement path corresponding to the sensing data set as indexes, the corresponding data are arranged in chronological order. Based on the sensory data set unfolding along the fish's movement path, the three-dimensional position coordinates of each spatial node and the corresponding fish movement state, water disturbance, and environmental response parameters are extracted. These continuous spatial nodes are used as indexes to establish a mapping relationship. The node information is sorted according to the data acquisition time of each node to ensure that a time series is formed from the earliest acquired node to the latest acquired node. Timestamps are used to correct the time deviation between each biomimetic fish sensory unit to ensure the temporal consistency of multi-unit data. After sorting, each spatial node not only retains its spatial position in the water, but also contains the movement state, disturbance changes, and environmental response information at the corresponding time point, thus forming a node sequence arranged in a time-recursive manner, which can be directly used for calculating behavioral changes and describing time evolution.

[0026] Between adjacent time slices, based on the data differences between adjacent nodes in the sensing dataset, the changes in fish movement state, water disturbance intensity, and environmental response parameters are calculated. In the time-recursive node sequence, two adjacent spatial nodes are selected, and the fish motion state, water disturbance data, and environmental response parameters corresponding to each node are extracted, including changes in fish position, velocity, and acceleration, water flow velocity perturbation, and environmental parameters such as local temperature, dissolved oxygen, and pH. The difference between the data of each pair of adjacent nodes is calculated to obtain the displacement increment, velocity change, and acceleration change of the fish motion state. At the same time, the change in water disturbance intensity, such as the change in the amplitude of flow velocity perturbation, and the change amplitude of environmental response parameters within the time slice are calculated. A unified time interval normalization method is used to compensate for the sampling frequency differences of data collected by different sensing units to ensure that the changes are comparable at the same time scale. The changes obtained by calculating the changes for each pair of nodes form a complete time-recursive change sequence, which can be directly used for the generation of multi-scale behavioral evolution features.

[0027] By connecting the changes in a time-recursive manner, the spatially continuous data in the dataset is transformed into a temporally evolving description, forming a sequence of behavioral changes that characterize the fish school's behavior over time. The continuous spatial nodes in the sensor data set unfolding along the fish's movement path are used as time indices. For each node, the three-dimensional position, instantaneous velocity, acceleration, water disturbance intensity, and environmental response parameters of the fish are recorded, including local flow velocity, flow direction, dissolved oxygen, temperature, and water transparency. The node data are arranged in chronological order of acquisition. The changes in fish movement state, water disturbance, and environmental response are calculated between adjacent time nodes. Sensor noise is removed using sliding window filtering and weighted difference methods to ensure that the changes accurately reflect the fish's behavior and environmental dynamics. The changes at each time node are recursively concatenated in chronological order, so that the original spatial continuous data forms a continuous temporal evolution description. Each node not only retains its spatial position but also contains the corresponding movement and environmental change characteristics, thereby generating a complete behavioral change description sequence. In this sequence, the change amplitude can be marked according to the historical normal baseline. Movement or disturbance nodes exceeding the normal threshold are marked as potential stress behaviors, thus ensuring that the sequence can clearly reflect the behavioral evolution of the fish over time.

[0028] The process of generating multi-scale behavioral evolution feature representations in S2 is as follows: Extract the normal behavioral characteristics of fish populations under water and seasonal conditions corresponding to the behavioral change description sequences from historical aquaculture operation data, and construct a baseline of historical fish population activity. Fish activity data under specified water conditions and seasons are extracted from historical aquaculture operation databases. The raw data types include fish location coordinates, swimming speed, direction vector, depth changes, and aquatic environmental parameters such as water temperature, dissolved oxygen, and turbidity. These data are organized into time series on a daily or hourly basis. These time series are then statistically calculated according to behavioral indicators such as average speed, turning frequency, and group density to obtain the normal behavioral characteristics of each time period, including mean, standard deviation, and higher-order statistics such as skewness and kurtosis. These data are then summarized to form a baseline matrix of historical fish activity, where rows represent time periods and columns represent behavioral indicators for comparison with current observed behavior.

[0029] The current behavioral change description sequence is compared with the historical activity baseline time by time, and the offset of the behavioral change description sequence relative to the normal state in each time period is calculated to form a behavioral offset sequence arranged by time index. For each behavioral indicator, the difference between the current observation value and the corresponding historical baseline mean is calculated, and the difference is standardized into an offset indicator. For example, z-score standardization is used, and the offset is defined as the current value minus the historical mean and then divided by the historical standard deviation, thereby eliminating the influence of different indicator dimensions. By performing the above calculation on all indicators in each time period, a behavioral offset sequence arranged in chronological order is generated, where each time period corresponds to a set of offset vectors, which is convenient for capturing the degree and trend of changes in fish behavior relative to the normal state.

[0030] Based on the behavioral offset sequence, aggregation and expansion processes are performed at different time scales to generate a multi-scale behavioral evolution feature expression that reflects the behavioral evolution characteristics of fish groups. Based on the behavioral offset sequence, the mean, variance, and rate of change of the offset within the sliding window are calculated at short time scales, such as minutes or hours, to obtain local behavioral evolution characteristics. Then, at medium time scales, such as days or weeks, the short time scale characteristics are accumulated and statistically analyzed, for example, by calculating the average, maximum, minimum, and trend slope, to obtain medium-scale evolution characteristics. Finally, at long time scales, such as months or quarters, the medium-scale characteristics are modeled using time series, for example, by calculating the overall evolution trend and fluctuation range through autoregressive models or moving average models, to generate global behavioral evolution characteristics.

[0031] S3: Based on the multi-scale behavioral evolution characteristics, by identifying the structural inflection point of the transition from stable to stress behavior patterns, candidate state fragments reflecting the abnormal emergence of aquaculture operation status are extracted.

[0032] The process of extracting candidate state fragments reflecting the abnormal emergence of aquaculture operation status in S3 is as follows: Continuous time window analysis was performed on the multi-scale behavioral evolution feature representation to obtain the stability index of behavioral features changing over time at different time scales; The generated multi-scale behavioral evolution features are divided into preset continuous time windows. The length of the time window can be determined according to the rate of change of fish behavior. For example, a short time window can be 10 minutes, a medium time window can be 1 hour, and a long time window can be 1 day. Each window contains the behavioral offset and multi-scale statistical indicators for the corresponding time period. Stability indicators are calculated for the behavioral features within each time window, including the variance, skewness, and moving average rate of change of the indicators. These statistics are used to measure the fluctuation range and consistency of the behavioral features in that time period, thereby obtaining the stability curve matrix of the behavioral features over time at each time scale, providing basic data for the determination of anomaly emergence.

[0033] Based on stability indicators, we identify abrupt changes in behavioral characteristics over time, and combine these with behavioral shifts within the corresponding time period to determine whether the abrupt changes conform to the characteristic pattern of transition from a stable state to a stress state. Change point detection algorithms, such as CumulativeSum control charts, sliding window mean change detection, or Bayesian online change point detection, are applied to stability indicators to determine the abrupt changes in behavioral characteristics over time. The behavioral offset corresponding to each abrupt change point is compared with the historical activity baseline, and it is determined whether the change pattern conforms to the characteristic pattern of transition from a stable state to a stress state. For example, a situation where the offset exceeds two standard deviations of the historical mean and the fluctuation direction is consistent is identified as a stress germination signal, thereby screening out potential abnormal germination events.

[0034] The time periods corresponding to the mutation points that are determined to conform to the transition feature pattern are extracted from the multi-scale behavioral evolution feature expression to form candidate state segments that reflect the abnormal emergence of the aquaculture operation status. After determining that a mutation point conforms to the characteristic pattern of transition from a stable state to a stress state, the start and end times corresponding to the mutation point are obtained. The corresponding segments are located in the multi-scale behavioral evolution feature expression matrix in units of time periods. The data included includes the behavioral offset vector at each time point and statistical indicators such as mean, variance, and rate of change calculated at different time scales. These data are continuously extracted according to the time index to form candidate state segments. The start and end times, behavioral indicators, and statistics at each scale of the segments are completely preserved to ensure traceability. To ensure data verification, the extraction operation is performed using a clear index range and matrix slicing method. The output candidate state segments are stored in the form of matrices or tensors, with each row corresponding to a time point and each column corresponding to behavioral indicators and statistical features, thus forming candidate state segments that can reflect the abnormal emergence of aquaculture operation status.

[0035] S4: Map candidate state fragments to corresponding aquaculture operating environment units, and combine water circulation mode, aquaculture density distribution and recent operation records to establish the correspondence between candidate state fragments and aquaculture operating environment, and generate local operating state association structure.

[0036] The process of mapping candidate state fragments to corresponding aquaculture operating environment units in S4 is as follows: Based on the path-expanded perception data corresponding to the candidate state fragments in the perception data set, the spatial movement path of the bionic fish during the occurrence of abnormal behavior and its continuous distribution characteristics in the time dimension are extracted to determine the main spatial coverage of abnormal behavior in the aquaculture water. By deploying multi-source sensors in the aquaculture water, such as underwater cameras, acoustic sensors, acceleration or position sensors, the movement trajectory and behavior data of the bionic fish are collected. The data is then time-synchronized and cleaned to remove sensor abnormalities or signal loss. The abnormal behavior time periods corresponding to candidate state segments are aligned with the sensing data set to extract the spatial coordinate sequence of the bionic fish within that time period. The trajectory density estimation method is used to calculate the residence time and movement frequency distribution in various areas of the water body, thereby forming a continuous spatial distribution feature map of abnormal behavior. By setting thresholds or density clustering methods, the main spatial coverage of abnormal behavior can be identified, that is, the water body areas where the bionic fish behavior is highly concentrated or abnormally frequent.

[0037] Based on the degree of concentration of the main spatial coverage area in the water space and the spatial adjacency relationship between the aquaculture operation structure, the main spatial coverage area is matched with the functional zones, equipment layout areas or water management units in the aquaculture water body. The aquaculture water body is spatially divided to establish a water body management unit map, including functional zones such as feeding areas and aeration areas, equipment deployment areas such as areas where water pumps, aeration equipment, and monitoring sensors are located, and other management units such as water flow control areas and feeding paths. The main spatial coverage area of ​​abnormal behavior is mapped to the management unit map, and the spatial adjacency or overlap of the coverage area with each unit is calculated. For example, the minimum envelope rectangle, the overlap area ratio, or the weighted proximity matrix is ​​used for quantitative matching. In this way, the area where abnormal behavior occurs can be correlated with the actual aquaculture structural unit, and the potentially associated management or equipment areas can be identified, providing a spatial basis for the analysis of the cause of the anomaly.

[0038] From the matching results, select aquaculture operation environment units that are consistent with the spatial characteristics of abnormal behavior, and establish a spatial correspondence between candidate state segments and corresponding aquaculture operation environment units; Based on the matching results, the aquaculture operating environment units with the highest overlap or closest proximity to the main spatial coverage of abnormal behavior are extracted as candidate association units. For each candidate unit, feature consistency indicators such as spatial overlap rate, behavioral density similarity, or time synchronization deviation are further calculated. Units with high consistency are retained to form the final screening list. The candidate state fragments are mapped one-to-one with the screened aquaculture operating environment units to establish a clear spatial correspondence, so that they can be directly called in abnormal behavior analysis, intelligent regulation, or automated management to achieve precise location of abnormal behavior of biomimetic fish and environmental response control.

[0039] The process of generating the local runtime state association structure in S4 is as follows: Acquire water circulation parameters, stocking density information, and recent operation records corresponding to the aquaculture operation environment unit to form an operation information set characterizing the current operating conditions; Monitoring devices, such as flow velocity sensors, water level sensors, dissolved oxygen and water temperature sensors, as well as fish density monitoring systems, are deployed in the aquaculture water. Data is automatically recorded in real time through the acquisition and storage devices. Operational records include information such as feeding time, aeration equipment start-up and shutdown records, and water flow regulation operations. This data is categorized and processed in time according to aquaculture operating environment units to generate an operational information set for each unit, including water circulation status, stocking density, equipment operation history, and environmental change information. This information can be stored in a database table or time-series vector format, providing detailed environmental context for behavioral analysis.

[0040] The candidate state fragments in the set of operational information and spatial correspondence are associated and matched in the time and spatial dimensions to analyze the influence of different operating conditions on the generation and evolution of abnormal behavior fragments. The set of operational information and candidate state segments are mapped spatially using aquaculture unit IDs and aligned temporally using timestamps. Data fusion methods, such as time series cross-correlation analysis, dynamic time warping, or weighted index-based matching algorithms, are used to quantify the impact of changes in operational parameters on the occurrence and duration of abnormal behavior. For example, it can be analyzed whether the initiation of a certain aeration operation corresponds to the start of abnormal behavior, or whether a specific water flow circulation pattern causes a shift in the concentrated area of ​​fish behavior. Through this correlation and matching, the response characteristics of candidate state segments under different operational conditions can be formed, revealing the causal or correlational relationship between abnormal behavior and environmental factors.

[0041] Based on the association matching results, a local operational state association structure is constructed to describe the influence relationship between candidate state segments and aquaculture operation environment units; Based on the response characteristics of candidate state segments under different operating conditions, candidate state segments are used as nodes, and aquaculture operating environment units are used as another type of node. A dual-node network or correlation matrix is ​​constructed to represent the local operating state. The weight of the edges can be determined by the intensity of abnormal behavior response, the influence of water parameters, or time synchronization measures. It can be implemented in the form of graph data structures, table matrices, or knowledge graphs. Furthermore, it can support the generation of anomaly prediction and control strategies based on rules or machine learning. For example, when a certain water circulation unit is highly correlated with a specific behavioral anomaly, it can be marked as a key control point in the system to provide a basis for automatic regulation. The resulting local operating state correlation structure can intuitively and quantitatively describe the influence relationship between abnormal behavior and environmental factors, realizing targeted monitoring and intelligent intervention.

[0042] S5: Based on the local operational status association structure, the collection density and feature weights of bionic fish perception data in different water areas and time periods are adaptively adjusted to form the current state representation of aquaculture operation.

[0043] The process of forming the current state characterization results of aquaculture operation in S5 is as follows: Based on the influence relationship between candidate state segments represented in the local operational state association structure and aquaculture operational environment units, the water area where abnormal behavior is concentrated and the corresponding risk level are identified. The association weights and abnormal response characteristics of each candidate state segment with the aquaculture operating environment unit are extracted from the local operational state association structure. Spatial density clustering or heat map analysis methods are used to identify water areas with high frequency and long duration of abnormal behavior, forming anomalous concentrated water areas. Combining the association weights, abnormal behavior intensity and the importance indicators of water management units, each anomalous concentrated water area is assigned a risk level, such as low, medium and high, and a water risk distribution map is generated, which can serve as the basis for dynamic monitoring and operational decision-making.

[0044] Based on the identified abnormal concentrated water areas and risk levels, the patrol frequency and sensing data collection density of the bionic fish sensing units in the corresponding water areas are adjusted. Based on the risk level of the water area, the patrol strategy of biomimetic fish sensing units, such as underwater cameras, acoustic sensors, or position sensors, is dynamically adjusted. For high-risk areas, the patrol frequency of biomimetic fish can be increased to improve data acquisition density and timeliness. For medium-risk areas, the patrol frequency can be appropriately increased to ensure data coverage. For low-risk areas, the basic patrol frequency is maintained to save energy consumption and computing resources. Biomimetic fish movement path optimization algorithms, grid-based acquisition scheduling strategies, or dynamic sampling control algorithms can be adopted to ensure that sensing resources are rationally allocated in the water body according to the risk level.

[0045] Based on the risk weights of candidate state segments in the local operational state association structure, differentiated feature weights are assigned to the sensing data collected in different time periods. The perception data collected in each time period are weighted by combining the risk weights corresponding to the candidate state segments. Data from high-risk segments are given higher feature weights in analysis and modeling to ensure that the response characteristics of abnormal behavior are fully reflected in the situation assessment. Data from low-risk segments are given lower weights to reduce interference with the overall situation. The weighting method can adopt linear weighting, normalized weighting, or exponential weighting strategies, and can also combine time windows to perform sliding weighting on historical data to reflect the evolution trend of abnormal behavior.

[0046] By integrating the collected sensing data after adjusting for density and feature weights, a status representation result reflecting the current aquaculture operation status is generated. By fusing dynamically collected data and weighted features, a multidimensional state representation matrix is ​​formed, including the spatial location of the water area, the intensity of abnormal behavior, the risk level, and the temporal evolution characteristics. Through aggregation, standardization, or dimensionality reduction processing, such as PCA, feature selection, or deep feature extraction methods, state representation results that can be directly used for operational analysis and anomaly prediction are generated. These results can be used for visualization, such as water area situation heat maps, operational decision support, or automatic control strategy generation, to achieve real-time dynamic reflection of the behavior of biomimetic fish and the environmental status in aquaculture water bodies.

[0047] Example 2: Please refer to Figure 2As shown, a biomimetic fish perception data-driven aquaculture operation status modeling system includes: Continuous sensing module: continuously collects water disturbances caused by fish activity, fish migration trajectories and local environmental responses, and performs spatial assignment reshaping on the raw sensing data to form a sensing data set that can be expanded along the fish movement path. Behavior modeling module: Based on the sensory data set, construct a time-recursive behavioral change description sequence along the fish's movement path, and generate multi-scale behavioral evolution feature representations; Anomaly extraction module: Identifies structural inflection points in behavior transitioning from stability to stress from multi-scale behavioral evolution feature representation, and extracts corresponding candidate state fragments; Association Construction Module: Maps candidate state fragments to aquaculture operation environment units and generates a local operation state association structure describing the correspondence between the two. Situation characterization module: Based on the local operational status association structure, the collection density and feature weights of the sensing data are adaptively adjusted to form the current state characterization result of the aquaculture operation.

[0048] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0049] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A biomimetic fish perception data-driven method for modeling aquaculture operation status, characterized in that... This includes the following steps: Continuous sensing data is collected on water disturbances caused by fish activity, fish spatial migration trajectories, and local environmental responses affected by changes in behavioral states. Based on the fish activity range and the continuity of water space, the collected raw sensing dataset is spatially reorganized to form a sensing data set that can be expanded along the fish movement path. Based on the sensory data set, a time-recursive behavioral change description sequence is constructed along the fish's movement path. The behavioral offset between the historical activity baseline of the fish school and the current collection period is introduced to generate a multi-scale behavioral evolution feature expression. Based on the multi-scale behavioral evolution characteristics, by identifying the structural inflection points in the transition of behavioral patterns from stable to stress, candidate state fragments reflecting the emergence of abnormalities in aquaculture operation are extracted; Candidate state fragments are mapped to corresponding aquaculture operating environment units, and the correspondence between candidate state fragments and aquaculture operating environment is established by combining water circulation mode, aquaculture density distribution and recent operation records, thus generating a local operating state association structure; Based on the local operational status association structure, the collection density and feature weights of biomimetic fish perception data in different water areas and time periods are adaptively adjusted to form the current state representation of aquaculture operation.

2. The biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 1, characterized in that... The process of continuously sensing and collecting data on water disturbances caused by fish activity, fish spatial migration trajectories, and local environmental responses affected by changes in behavioral states is as follows: A biomimetic fish sensing unit with autonomous swimming capability is deployed to simultaneously acquire water flow velocity perturbation signals, fish relative displacement information, and corresponding environmental response parameters during its cruising movement in different water depths and spatial areas. The data collected by each biomimetic fish sensing unit is timestamped and locally cached, and time consistency correction of the sensing data of multiple units is completed based on a wireless synchronization mechanism; The time-corrected sensing data is uploaded to the aquaculture operation management node to form a raw sensing dataset containing water disturbance, fish migration, and environmental response.

3. The biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 2, characterized in that... The process of forming a set of sensory data that can be deployed along the fish's movement path is as follows: Based on the spatial positioning information of the bionic fish sensing unit, the spatial location of each sensing data point in the water body and its associated activity range are determined. Based on the spatial connectivity of aquaculture water bodies, path association mapping and regional merging are performed on raw sensing data collected at different times and spatial locations. The merged sensory dataset is sequentially aligned and expanded according to the actual movement path of the bionic fish, generating a sensory data set that can be expanded along the movement path of the fish.

4. The biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 3, characterized in that... The process of constructing a time-recursive behavioral change description sequence along the fish's movement path is as follows: Using the continuous spatial nodes along the fish's movement path corresponding to the sensing data set as indices, the corresponding data are arranged in chronological order; Between adjacent time slices, based on the data differences between adjacent nodes in the sensing dataset, the changes in fish movement state, water disturbance intensity, and environmental response parameters are calculated; By concatenating the changes in a time-recursive manner, the spatially continuous data in the dataset is transformed into a temporal evolution description, forming a sequence of behavioral changes that characterize the fish school's behavior over time.

5. The biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 4, characterized in that... The process of generating multi-scale behavioral evolution feature representations is as follows: Extracting the normal behavioral characteristics of fish populations under water and seasonal conditions corresponding to the behavioral change description sequences from historical aquaculture operation data, and constructing a baseline of historical fish population activity; The current behavioral change description sequence is compared with the historical activity baseline time by time, and the offset of the behavioral change description sequence relative to the normal state in each time period is calculated to form a behavioral offset sequence arranged by time index; Based on the behavioral offset sequence, aggregation and expansion processes are performed at different time scales to generate a multi-scale behavioral evolution feature expression that reflects the behavioral evolution characteristics of fish groups.

6. The biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 5, characterized in that... The process of extracting candidate state fragments reflecting the abnormal emergence of aquaculture operation status is as follows: Continuous time window analysis was performed on the multi-scale behavioral evolution characteristics to obtain stability indices of behavioral characteristics over time at different time scales; Based on stability indicators, abrupt changes in behavioral characteristics over time are identified, and combined with behavioral shifts within the corresponding time periods, it is determined whether the abrupt changes conform to the characteristic pattern of transition from a stable state to a stress state. The time periods corresponding to mutation points that are determined to conform to the transitional feature pattern are extracted from the multi-scale behavioral evolution feature expression to form candidate state fragments reflecting the abnormal emergence of aquaculture operation status.

7. The biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 6, characterized in that... The process of mapping candidate state fragments to corresponding aquaculture operating environment units is as follows: Based on the path-expanded sensing data corresponding to candidate state fragments in the sensing dataset, the spatial movement path of the biomimetic fish during the occurrence of abnormal behavior and its continuous distribution characteristics in the time dimension are extracted to determine the main spatial coverage of the abnormal behavior in the aquaculture water body; Based on the concentration of the main spatial coverage area in the water body and the spatial adjacency relationship between the main spatial coverage area and the aquaculture operation structure, the main spatial coverage area is matched with the functional zones, equipment layout areas, or water body management units in the aquaculture water body; From the matching results, select aquaculture operation environment units that are consistent with the spatial characteristics of abnormal behavior, and establish a spatial correspondence between candidate state segments and corresponding aquaculture operation environment units.

8. The biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 7, characterized in that... The process of generating the local runtime state association structure is as follows: Acquire water circulation parameters, stocking density information, and recent operation records corresponding to the aquaculture operation environment unit to form an operational information set characterizing the current operating conditions; The candidate state fragments in the set of operational information and their spatial correspondence are correlated and matched in the temporal and spatial dimensions to analyze the influence of different operational conditions on the generation and evolution of abnormal behavior fragments; Based on the association matching results, a local operational state association structure is constructed to describe the influence relationship between candidate state segments and aquaculture operation environment units.

9. A biomimetic fish perception data-driven aquaculture operation status modeling method according to claim 8, characterized in that... The process of forming the current state characterization results of aquaculture operation is as follows: Based on the influence relationship between candidate state segments represented in the local operational state association structure and aquaculture operational environment units, the water areas where abnormal behaviors are concentrated and the corresponding risk levels are identified. Based on the identified areas of concentrated abnormality and the degree of risk, the patrol frequency and data collection density of the bionic fish sensing units in the corresponding areas are adjusted. Based on the risk weights of candidate state segments in the local operational state association structure, differentiated feature weights are assigned to the sensing data collected in different time periods; By integrating the collected sensing data after adjusting for density and feature weights, a status representation result reflecting the current aquaculture operation status is generated.

10. A biomimetic fish perception data-driven aquaculture operation status modeling system, applied to the method described in any one of claims 1-9, characterized in that... ,include: Continuous sensing module: continuously collects data on water disturbances caused by fish activity, fish migration trajectories, and local environmental responses, and performs spatial reassignment reshaping on the raw sensing data to form a sensing data set that can be expanded along the fish movement path; Behavioral modeling module: Based on the sensory data set, constructs a time-recursive behavioral change description sequence along the fish's movement path, and generates multi-scale behavioral evolution feature representations; Anomaly extraction module: Identifies structural inflection points in behavior transitioning from stable to stress states from multi-scale behavioral evolution feature representations, and extracts corresponding candidate state fragments; Association Construction Module: Maps candidate state fragments to aquaculture operating environment units and generates a local operating state association structure describing the correspondence between the two; Situation characterization module: Based on the local operational status association structure, the collection density and feature weights of the sensing data are adaptively adjusted to form the current state characterization result of the aquaculture operation.

Images