Intelligent aquaculture whole life cycle management method

Abstract

The present application relates to the field of intelligent agriculture and aquaculture technology, and provides an intelligent aquaculture whole life cycle management method, comprising: dividing the aquaculture water area into several sub-regions according to the terrain slope, water flow potential line and light and shadow boundary; periodically collecting the growth state parameters of plants in each sub-region, comparing with the predicted value to determine the preliminary screening abnormal area; comparing the abnormal area with the adjacent area to determine whether the heterogeneous growth condition is met; if yes, waking up the dormant underwater robot to collect video and calculate the fish school activity parameters; determining whether stress should be generated based on the average swimming speed and spatial distribution density; determining the stress level based on the turning frequency and synchronism index; and performing the grading control of increasing aeration, adjusting water circulation flow and reducing feed, starting emergency oxygenation and alarming according to the mild, moderate and severe stress respectively. The present application realizes cross-domain perception, on-demand wake-up, and partitioned precise management, greatly reduces the monitoring energy consumption, and improves the accuracy of abnormal response.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture and aquaculture technology, specifically to a smart aquaculture full life cycle management method. Background Technology

[0002] Factory-style recirculating aquaculture and aquaponics systems are gradually developing towards high-density, intensive, and ecological practices. However, existing aquaculture management methods have the following shortcomings:

[0003] First, most systems rely solely on underwater water quality sensors (dissolved oxygen, pH, ammonia nitrogen, etc.) or single-point underwater cameras for fish observation, making it difficult to fully perceive the spatial heterogeneity of aquaculture waters. Within the same aquaculture pond, due to differences in terrain slope, water flow distribution, and lighting, the behavior of fish and water quality parameters in different areas may vary significantly. However, traditional methods typically employ uniform sampling or single-point sampling, resulting in monitoring blind spots.

[0004] Second, there is a lack of cross-domain collaborative early warning mechanisms: abnormal fish behavior is often an early signal of water quality deterioration or disease occurrence, but in the current technology, there is a lack of linkage analysis between the growth status of aquatic plants and the activities of fish underwater; in aquaponics systems, plant growth can directly reflect the nutritional status of the water, but there is no method to use plant growth as an indirect biosensor of the health status of fish.

[0005] Third, the problem of high energy consumption is prominent: the continuous operation of underwater high-definition camera equipment and multi-parameter water quality sensors consumes a lot of power, especially in large-scale aquaculture scenarios, and the cost of all-weather monitoring is high; existing technologies lack energy-saving strategies for on-demand wake-up.

[0006] Fourth, the control strategy is crude: once an anomaly is detected, uniform control of the entire pond is usually implemented (such as full pond oxygenation and full pond water exchange), lacking the ability to conduct precise control of the abnormal area in a graded and zoned manner, resulting in waste of resources and amplification of stress;

[0007] Therefore, there is an urgent need for a smart aquaculture full life cycle management method that can integrate the perception of aquatic plants and the perception of underwater fish, and has spatial zoning, on-demand activation, and hierarchical regulation. Summary of the Invention

[0008] To address these issues, this invention provides a smart aquaculture full lifecycle management method to overcome the problems in existing technologies, such as monitoring blind spots and misjudgments caused by ignoring spatial heterogeneity, lack of cross-domain collaborative sensing mechanisms between aquatic plants and underwater fish, high cost of continuous high-power monitoring, and extensive control strategies that cannot be differentiated and tiered for intervention.

[0009] To achieve the above objectives, this invention provides a smart aquaculture full life-cycle management method, comprising:

[0010] Step S1: Obtain the bottom topographic slope, water circulation equipotential lines, and light and shadow boundaries of the shading structure above the aquaculture area to divide the aquaculture area into several aquaculture sub-regions.

[0011] Step S2: Periodically collect the growth status parameters of plants in each of the cultivation sub-regions, and determine the region type of the cultivation sub-region based on the degree of deviation between the growth status parameters and the predicted status parameters.

[0012] The growth status parameters include plant height, leaf area index, and leaf color index, and the region types include pre-screening abnormal regions and pre-screening normal regions.

[0013] Step S3: Compare the growth status parameters of each of the initial screening abnormal regions with those of the adjacent culture sub-regions, and determine whether the initial screening abnormal region meets the heterogeneous growth conditions based on the comparison results.

[0014] Step S4: In response to the satisfaction of heterogeneous growth conditions, the dormant camera unit is awakened and underwater video is acquired of the abnormal area in the initial screening.

[0015] Step S5: Analyze the fish behavior of the collected underwater video and calculate the fish activity parameters;

[0016] Step S6: Determine whether the fish are in a state of stress and the stress level based on the fish activity parameters;

[0017] Step S7: Determine the control strategy for the aquaculture sub-region based on the stress level;

[0018] The control actions include increasing local aeration, adjusting water circulation flow, and initiating emergency oxygenation.

[0019] As a preferred technical solution for the intelligent aquaculture full life cycle management method, the process of dividing the aquaculture water area into several aquaculture sub-regions in step S1 includes:

[0020] Step S11: Divide the continuous areas where the slope difference of the pool bottom is less than the preset slope threshold into the same terrain unit;

[0021] Step S12: Divide the regions where the equipotential linear density difference of the water circulation is less than a preset density threshold into the same hydraulic unit;

[0022] Step S13: Divide the area where the difference in daily average solar radiation is less than a preset solar threshold into the same solar unit according to the light and shadow boundary of the upper shading structure.

[0023] Step S14: The spatial intersection of the terrain unit, hydraulic unit, and illumination unit is determined as the aquaculture sub-region.

[0024] As a preferred technical solution for the intelligent aquaculture full life cycle management method, the growth status parameters of plants in each of the aquaculture sub-areas and underwater videos are collected by the surface acquisition unit and the underwater acquisition unit, respectively.

[0025] The waterborne data acquisition unit is a multi-rotor UAV equipped with a hyperspectral camera, a multispectral camera, or an RGB camera, and is also equipped with a global positioning module and an inertial navigation module.

[0026] The underwater acquisition unit is an underwater robot comprising a body, a thruster, an underwater lighting lamp, a waterproof camera unit, a depth sensor, and a wireless communication module;

[0027] The underwater robot is normally docked at a charging station at the edge of the aquaculture area and is in a low-power sleep state.

[0028] As a preferred technical solution for the intelligent aquaculture full life cycle management method, the process of determining the region type of a single aquaculture sub-region in step S2 includes:

[0029] Step S21: Pre-construct plant growth prediction models for each of the aforementioned aquaculture sub-regions. The prediction models take historical growth data, ambient temperature, light intensity, and aquaculture density as inputs and the predicted state parameters at the current moment as outputs.

[0030] Step S22: Calculate the Euclidean distance between the measured growth state parameters and the predicted state parameters, and sum the weighted Euclidean distances to obtain the comprehensive deviation.

[0031] Step S23: Determine the region type of the corresponding aquaculture sub-region based on the comprehensive deviation, wherein:

[0032] Step S24: In response to the overall deviation being greater than the deviation threshold, the aquaculture sub-region is determined to be an abnormal region in the initial screening.

[0033] Conversely, if the condition is not normal, it is identified as a normal area in the initial screening.

[0034] As a preferred technical solution for the intelligent aquaculture full life cycle management method, in step S3, the growth status parameters of each of the initially screened abnormal regions are compared with those of adjacent aquaculture sub-regions. Based on the comparison results, it is determined whether the initially screened abnormal region meets the heterogeneous growth conditions, including:

[0035] Step S31: Determine all aquaculture sub-regions that are geographically adjacent to the abnormal area in the initial screening, and calculate the proportion of the normal area in the initial screening.

[0036] Step S32: Calculate the comprehensive deviation difference between the abnormal area in the initial screening and each adjacent normal area in the initial screening;

[0037] Step S33: Calculate the average value of each of the comprehensive deviation differences and record it as the average deviation difference;

[0038] Step S34: Determine whether the heterogeneous growth conditions are met based on the stated proportion and the average deviation difference, wherein:

[0039] If the proportion is greater than a preset proportion and the average deviation difference is greater than a preset difference, then the heterogeneous growth condition is determined to be met.

[0040] Conversely, if the conditions for heterogeneous growth are not met, it is determined that the conditions are not met.

[0041] As a preferred technical solution for the intelligent aquaculture full life cycle management method, step S5 involves analyzing the fish behavior of the collected underwater video and calculating the fish activity parameters, including:

[0042] Step S51: Detect fish bodies and track targets in the underwater video frames, extracting the position coordinates, orientation angle, and body length information of each fish frame by frame;

[0043] Step S52: Calculate the instantaneous swimming speed and average swimming speed of each fish based on the position coordinates of multiple consecutive frames;

[0044] Step S53: Calculate the spatial distribution density of the fish school, the aggregation center point, and the distance distribution between each fish and the aggregation center based on the position coordinates of all fish at the same time.

[0045] Step S54: Calculate the fish school turning frequency and synchronization index based on the time series of the orientation angle;

[0046] The fish activity parameters include average swimming speed, spatial distribution density, turning frequency, and synchronization index.

[0047] As a preferred technical solution for the intelligent aquaculture full life cycle management method, in step S6, determining whether the aquaculture is in a stress state based on the average swimming speed and spatial distribution density includes:

[0048] If the average swimming speed is not within the speed reference range and / or the spatial distribution density is within the density reference range, then the fish population in the corresponding aquaculture sub-area is determined to be in a state of stress.

[0049] As a preferred technical solution for the intelligent aquaculture full life cycle management method, in step S6, in response to the fish population in the aquaculture sub-area being in a state of stress, the stress level is determined based on the turning frequency and the synchronization index;

[0050] Calculate the percentage deviation of the frequency and the percentage deviation of the synchronization index for the steering frequency and the synchronization index, respectively.

[0051] The stress level is determined based on the percentage of frequency deviation and the percentage of synchronization deviation, wherein:

[0052] If both the frequency deviation percentage and the synchronization deviation percentage are greater than the corresponding deviation threshold, the stress level is determined to be severe stress.

[0053] If the percentage deviation of the frequency or the percentage deviation of the synchronization is greater than the corresponding deviation threshold, it is determined to be moderate stress.

[0054] If both the frequency deviation percentage and the synchronization deviation percentage are not greater than the corresponding deviation threshold, then it is determined to be mild stress.

[0055] As a preferred technical solution for the intelligent aquaculture full life cycle management method, in step S7, the control strategy for the aquaculture sub-region is determined according to the stress level, including:

[0056] In response to mild stress, the air supply to the microporous aeration discs at the bottom of the breeding area was increased;

[0057] In response to moderate stress, the opening of the inlet valve of the aquaculture sub-area was increased while aeration was increased;

[0058] In response to severe stress, the emergency oxygenation system was activated to replenish pure oxygen to the breeding sub-area and send an alarm message to the management terminal.

[0059] As a preferred technical solution for the intelligent aquaculture full life cycle management method, in step S4, in response to the normal area of ​​the initial screening where the fish are not in a state of stress, the sampling frequency of the underwater sampling unit is increased.

[0060] Compared with existing technologies, the beneficial effects of this invention are as follows: The intelligent aquaculture full life cycle management method provided by this invention jointly senses the growth status of aquatic plants and the behavior of fish underwater. Through spatial sub-region division, it fully considers the heterogeneity of topography, hydraulics, and light within the aquaculture area, making subsequent anomaly detection more accurate and avoiding false alarms or missed alarms across the entire pond due to a uniform threshold. The two-level plant detection strategy, which employs initial screening for anomalies and comparison with adjacent plants, effectively eliminates false positives caused by global environmental fluctuations (such as pond-wide cooling or shading), only detecting false positives when the plant growth in a local area differs significantly from the surrounding area. Underwater vision is triggered only at specific times, thereby reducing the underwater robot's duty cycle and saving energy. Based on this, the average swimming speed and spatial distribution density of fish activity parameters are used for rapid stress detection, and then a three-level classification is performed using turning frequency and synchronization indicators, achieving refined diagnosis of stress phenomena. Finally, differentiated control actions are executed for different stress levels, avoiding excessive intervention that could cause secondary stress to the fish while ensuring rapid response in critical situations. In summary, this invention significantly improves the intelligence level of aquaculture management, resource utilization efficiency, and biosecurity capabilities.

[0061] In particular, this invention divides the aquaculture area into several sub-regions with uniform internal conditions based on the pond bottom slope, water flow equipotential lines, and light and shadow boundaries. The topographic slope controls the deposition location of fish feces and uneaten feed, the water flow equipotential lines determine the transport efficiency of nutrients and dissolved oxygen, and light and shadow directly affect plant photosynthesis. By dividing the sub-regions through the spatial intersection of these three factors, it ensures that fish behavior and water quality parameters within each sub-region have high autocorrelation, while identifiable differences exist between different sub-regions. This provides a reasonable reference system for comparing deviations between subsequent plant growth parameters and predicted values; that is, plant deviations within the same sub-region are more likely to stem from abnormal underwater fish activity rather than differences in topography or light itself. Simultaneously, by using comparisons between adjacent sub-regions to determine heterogeneous growth conditions, spatial autocorrelation analysis is introduced into the field of agricultural early warning. Only when the plant growth in a sub-region is significantly worse than in surrounding sub-regions is it identified as a true abnormal hotspot, thereby significantly reducing the false trigger rate caused by synchronous environmental changes throughout the pond (such as cold waves or rain).

[0062] In particular, the aquatic plant monitoring uses drones for periodic patrols and data collection. Compared to installing multispectral sensors in each sub-area, the drone solution has lower equipment costs and more flexible coverage. Compared to continuously operating underwater cameras, the power consumption of drone patrols is only comparable to that of tilling the land. In addition, the underwater robot is normally docked at the charging station and is in a low-power sleep state, only being awakened to collect video when it determines that heterogeneous growth conditions are met. This plant sentinel mechanism takes advantage of the slow physiological response speed (several hours to several days) but extremely low cost of plants, serving as an early warning information for acute anomalies in fish schools (several minutes to several hours). At the same time, since plant monitoring does not require high frame rate video, drone patrols 2 to 3 times a week are sufficient to meet the early warning needs, reducing the overall system energy consumption. Thus, while ensuring timely monitoring, the aquaculture IoT system achieves green and low-carbon operation.

[0063] In particular, this invention extracts four fish activity parameters, among which average swimming speed and spatial distribution density are most sensitive to acute stress and are used to quickly determine whether the fish are under stress. Once stress is determined, turning frequency and synchronicity indicators are used to further distinguish the stress level. Turning frequency reflects the agitation of the fish (chronic stress leads to frequent changes in direction), and synchronicity indicators reflect whether the fish are exhibiting abnormal clustering or chaotic swimming (neurobehavioral abnormalities caused by disease or toxins). This hierarchical determination avoids underreporting due to overly broad threshold settings for a single indicator or false alarms due to overly narrow threshold settings. At the control level, the hierarchical control strategy reduces oxygenation energy consumption and growth losses caused by unnecessary feed reduction compared to the traditional uniform control strategy. Attached Figure Description

[0064] Figure 1This is a flowchart of the intelligent aquaculture full life cycle management method according to an embodiment of the present invention. Detailed Implementation

[0065] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0066] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0067] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0068] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0069] It should be understood that in an aquaponics system, the activities of fish and the growth of vegetables constitute a dynamically balanced micro-ecosystem: fish excrement is transformed into plant nutrients by microorganisms, and the plants absorb nutrients while purifying the water and nourishing the fish's living environment; the two are interdependent, and any abnormality in one will be reflected in the other.

[0070] This invention constructs a cross-domain perception method that uses vegetables as a whistle and fish as a diagnostic tool through this inherent coupling relationship. The growth status of plants (height, leaf area, leaf color) changes relatively slowly and can be collected by drones at low cost and over a large area, thus serving as an indirect indicator of fish health. When the plant growth in a sub-region is significantly worse than that in the surrounding areas, it is highly likely that the fish below are exhibiting abnormal activity or water quality deterioration. At this time, the underwater robot is then activated to perform precise video collection and graded diagnosis of fish behavior in the abnormal area.

[0071] That is, low-cost plant monitoring filters out most normal periods, and the high-energy-consuming underwater vision system is only activated during high-risk periods. Compared with the traditional solution where the underwater camera works continuously for 24 hours, this invention compresses the actual working time of the underwater robot, significantly reducing the overall energy consumption and operation and maintenance costs of the system, while ensuring the timeliness and accuracy of monitoring, and truly realizing the green, intelligent and full life cycle management of the aquaponics system.

[0072] Please see Figure 1 As shown, it is a flowchart of the intelligent aquaculture full life cycle management method according to an embodiment of the present invention.

[0073] This invention provides a method for intelligent aquaculture lifecycle management, including:

[0074] Step S1: Obtain the bottom topographic slope, water circulation equipotential lines, and light and shadow boundaries of the shading structure above the aquaculture area to divide the aquaculture area into several aquaculture sub-regions.

[0075] Specifically, including:

[0076] Step S11: Divide continuous areas with a slope difference of less than a preset slope threshold into the same topographic unit. It is understood that the pond bottom slope determines the settling and accumulation patterns of solid particles such as fish feces and uneaten feed. In areas with a steeper slope, particles migrate to lower areas under the influence of water flow and are less likely to remain. In areas with a gentler slope, particles are more likely to settle, leading to increased local organic matter and ammonia nitrogen concentrations. If areas with large slope differences are grouped into the same sub-region, the nutrient load within that sub-region will be extremely uneven, making the average plant growth unrepresentative. Therefore, continuous areas with similar slopes are grouped into the same topographic unit to ensure consistent sedimentation conditions within the sub-region. In practice, the pond bottom slope is usually expressed as a percentage (elevation difference / horizontal distance × 100%), with a preset slope threshold of 1%–3%. It is understood that the critical slope is estimated based on the settling velocity of fish feces particles and the water flow velocity. Generally, when the slope difference exceeds 2%, the lateral migration distance of particles exceeds 10 meters (at typical flow velocities). (Below), sufficient to create significant sedimentary differences at the sub-regional scale;

[0077] Step S12: Divide the regions where the equipotential line density difference in water circulation is less than a preset density threshold into the same hydraulic unit. It is understood that the inlet, outlet, and aeration device together form a specific flow field. Water circulation equipotential lines refer to lines with equal flow velocities or stream functions. Regions with dense equipotential lines have large hydraulic gradients, fast water exchange, and are less prone to pollutant accumulation. Regions with sparse equipotential lines (such as dead zones) have long hydraulic retention times, making it easier for nutrients and metabolic waste to accumulate. Dividing regions with similar equipotential line densities into the same hydraulic unit ensures relatively uniform self-purification capacity and nutrient supply rate within the same sub-region. It is understood that the water circulation equipotential line density (the number of equipotential lines per unit length) reflects the strength of the hydraulic gradient. In implementation, computational fluid dynamics software (such as Fluent) is used to simulate the flow field in the aquaculture pond, outputting the flow velocity and hydraulic residence time at each point. Using a hydraulic residence time variation coefficient CV ≤ 20% as a standard, the maximum allowable equipotential line density difference is calculated.

[0078] Step S13: Based on the light and shadow boundaries of the upper shading structure, areas where the difference in average daily light radiation (the absolute difference in photosynthetically active radiation PAR) is less than a preset light threshold are divided into the same light unit. It is understood that the upper shading structure (such as photovoltaic panels, shading nets, or greenhouse frames) casts a fixed shadow on the water surface, and the difference in photosynthetically active radiation received by different sub-regions can be several times. The baseline of plant growth directly depends on light intensity. If areas with excessively large light differences are mixed together, fluctuations in plant growth will be mainly caused by changes in light intensity, rather than feedback from fish activity. Therefore, dividing sub-regions based on average daily light radiation can effectively isolate light interference. In implementation, the preset light threshold is... Through actual measurement and calibration, that is, using a quantum meter to measure the photosynthetically active radiation at various points during multiple time periods of the day (sunny days and cloudy days), calculate the daily average photosynthetically active radiation, statistically analyze the PAR difference on both sides of the same shadow boundary under natural conditions, and determine whether to divide the two sides of the shadow boundary into different illumination units.

[0079] Step S14: The spatial intersection of the topographic unit, hydraulic unit, and illumination unit is determined as the aquaculture sub-region. In practice, using the spatial intersection of the topographic unit, hydraulic unit, and illumination unit as the final aquaculture sub-region is equivalent to requiring each sub-region to simultaneously satisfy similar slope, similar hydraulic conditions, and similar illumination. Only in such sub-regions can the plant growth conditions be truly comparable, and the deviation between the subsequent measured values ​​and the predicted values ​​can be reliably attributed to the abnormal activity of the underwater fish. If only the union is taken or only a single factor is relied upon, false positives caused by the spatial variability of topography, water flow, or illumination itself cannot be ruled out.

[0080] Step S2: Periodically collect the growth status parameters of plants in each of the cultivation sub-regions, and determine the region type of the cultivation sub-region based on the degree of deviation between the growth status parameters and the predicted status parameters.

[0081] The growth status parameters include plant height, leaf area index, and leaf color index, and the region types include pre-screening abnormal regions and pre-screening normal regions.

[0082] Understandably, in aquaponics systems, plant growth is a comprehensive reflection of the aquatic nutrient status (derived from fish excrement) and environmental conditions (temperature, light). Establishing a personalized expected growth baseline for each aquaculture sub-region and then comparing the measured growth status with this baseline can determine if any anomalies have occurred in that region. If the plants in a sub-region grow significantly worse (or better) than expected, it indicates a possible deviation in the underwater environment of that region, requiring further monitoring. Specifically, this includes:

[0083] Step S21: A plant growth prediction model for each of the aforementioned aquaculture sub-regions is pre-constructed. This model takes historical growth data, ambient temperature, light intensity, and aquaculture density as input, and outputs the predicted state parameters at the current moment. In practice, the plant growth prediction model is constructed using a random forest regression algorithm. Taking a specific aquaculture sub-region as an example, historical data from 30 consecutive days of stable system operation and healthy fish populations is collected as the training set. Each sample contains an input vector and an output vector. The input vector includes the previous day's plant height, leaf area index, and leaf color index (3-dimensional), the current day's average ambient temperature (1-dimensional), the current day's cumulative light radiation (1-dimensional), and the previous week's average aquaculture density (1-dimensional). The output vector is the actual measured plant growth parameters for the current day. Height, leaf area index, and leaf color index (3-dimensional); 500 decision trees were used for model training, with 4 features randomly selected for node splitting in each tree, and the minimum number of leaf node samples set to 5; after training, the model can accept the current day's ambient temperature, light intensity, stocking density, and the previous day's growth status as input, and output the predicted values ​​of each growth parameter at the current moment, along with a 95% prediction interval; after deployment, newly collected 7 days of data are automatically added to the training set every week, and incremental learning or full retraining is performed every weekend to adapt to local seasonal changes and the drift of plant variety growth rhythms; for newly built systems (initially without historical data), a pre-trained model migrated from similar farms is used for the first two weeks, and then switched to a local adaptive model after two weeks;

[0084] Step S22: Calculate the Euclidean distance between the measured growth state parameters and the predicted state parameters, and sum the weighted Euclidean distances to obtain the comprehensive deviation. It is understood that the weighted summed Euclidean distance needs to be normalized to its absolute value. In addition, leaf color (such as NDVI or chlorophyll content) is most sensitive to the nitrogen level in the water. After ammonia nitrogen produced by fish excrement is converted into nitrate, the first manifestation of nitrogen deficiency in plants is a lighter leaf color. The leaf area index reflects the canopy structure and photosynthetic capacity of plants. Its comprehensive response time to nutrients and water is between that of leaf color and height. When the water body is chronically nutrient deficient or toxins accumulate, the LAI will decrease significantly. Height changes relatively slowly and is easily affected by the inherent growth rhythm of the species, and has low sensitivity to short-term water quality fluctuations. Therefore, the weight of leaf color index ≥ the weight of leaf area index > the weight of plant height. Typically, the weight of leaf color index is 0.4 to 0.6, the weight of leaf area index is 0.3 to 0.4, and the weight of plant height is 0.1 to 0.1.

[0085] Step S23: Determine the region type of the corresponding aquaculture sub-region based on the comprehensive deviation, wherein:

[0086] Step S24: In response to the overall deviation being greater than the deviation threshold, the aquaculture sub-region is determined to be an abnormal region in the initial screening.

[0087] Conversely, it is determined to be a normal area in the initial screening;

[0088] In practice, the overall deviation threshold is typically set between 0.15 and 0.30. A threshold that is too low will lead to numerous false alarms of normal fluctuations, while a threshold that is too high will miss genuine early anomalies. Understandably, it is also possible to calculate the overall deviation of each sub-region within a 30-day period assuming stable system operation and healthy fish populations, obtain a set of normal values, and then calculate their respective deviations. Set the initial threshold to .

[0089] Step S3: Compare the growth status parameters of each of the initial screening abnormal regions with those of the adjacent culture sub-regions, and determine whether the initial screening abnormal region meets the heterogeneous growth conditions based on the comparison results.

[0090] It is understandable that the deviation of plant growth in a sub-region from expectations may stem from two completely different reasons: (1) abnormal fish activity, water quality deterioration, sediment accumulation, etc. below the sub-region, which are true anomalies requiring intervention; (2) the entire aquaculture area has encountered environmental changes (such as cold waves, rain, and uniform adjustment of shade nets), causing all sub-regions of plants to deviate from expectations synchronously, but this deviation is a normal environmental response and does not require intervention for specific areas; therefore, step S3 uses spatial comparison to distinguish between local true anomalies and global environmental fluctuations. If most sub-regions around a pre-screened abnormal area are normal and the deviation of this area is significantly greater than that of the surrounding normal areas, then it is a true heterogeneous growth and needs to be confirmed by waking up the underwater robot; conversely, if the surrounding areas also deviate synchronously, it indicates that it is caused by global environmental factors and should not trigger subsequent high-cost underwater monitoring; step S3 specifically includes:

[0091] Step S31: Determine all aquaculture sub-regions that are geographically adjacent to the abnormal area in the initial screening, and calculate the proportion of the normal areas in the initial screening. It can be understood that all sub-regions adjacent to the abnormal area are identified, and the proportion of those marked as normal in the initial screening is counted. If the proportion is high, it means that the abnormality is local and only that area has a problem. If the proportion is low, it may be a global fluctuation.

[0092] Step S32: Calculate the comprehensive deviation difference between the abnormal area in the initial screening and each adjacent normal area in the initial screening;

[0093] Step S33: Calculate the average value of each of the comprehensive deviation differences and record it as the average deviation difference;

[0094] Step S34: Determine whether the heterogeneous growth conditions are met based on the stated proportion and the average deviation difference, wherein:

[0095] If the proportion is greater than a preset proportion and the average deviation difference is greater than a preset difference, then the heterogeneous growth condition is determined to be met.

[0096] Conversely, if the conditions for heterogeneous growth are not met, it is determined that the conditions are not met.

[0097] In practice, the preset ratio is usually between 0.5 and 0.8. If the preset ratio is too low, it is easy to misjudge the situation where a small number of normal areas are mixed in with the global fluctuations as local anomalies. If the preset ratio is too high, it may miss some core hot areas that are also partially abnormal but are actually more serious. The preset difference is generally between 0.10 and 0.25.

[0098] Step S4: In response to the satisfaction of heterogeneous growth conditions, the dormant camera unit is awakened and underwater video is acquired of the abnormal area in the initial screening.

[0099] Step S5: Analyze the fish behavior of the collected underwater video and calculate the fish activity parameters;

[0100] Step S51 involves fish detection and target tracking on underwater video frames, extracting the position coordinates, orientation angle, and body length information of each fish frame by frame. In practice, current mainstream fish detection methods (including the YOLO series and its improved models) are used to achieve an accuracy of over 95% in detecting farmed fish populations, while maintaining a detection speed of 30 frames per second. Then, target tracking is used to associate the bounding boxes of fish obtained in different frames during the detection phase, assigning a unique identifier (ID) to each fish, thereby obtaining the continuous motion trajectory of each fish on the timeline. Existing technologies include DeepSORT, DeepSORT, and FishTracker. (A multi-scale hollow attention module is introduced into the neck of YOLOv8 to improve detection accuracy. An adaptive Kalman filter is used to dynamically adjust motion prediction parameters during the tracking stage to cope with the nonlinear motion and adhesion problems of fish in high-density aquaculture environments. The algorithm achieves a MOTA of 93.22% and 87.24% under bright and dark lighting conditions, respectively, and a MOTA of 76.70% in real high-density aquaculture scenarios.) and AOMO-FishTracker (an orientation-aware multi-target fish swarm tracking framework that introduces a rotated bounding box to replace the traditional horizontal bounding box. A unique forward orientation is defined for each fish through geometric derivation. A rotation is used during the tracking stage.) The matching strategy and angle difference filtering mechanism effectively solve the mismatch problem caused by opposite orientations under high-density occlusion.

[0101] Step S52: Calculate the instantaneous swimming speed and average swimming speed of each fish based on the position coordinates of multiple consecutive frames; in practice, the average swimming speed is first calculated by taking the arithmetic mean of all instantaneous swimming speeds of the same fish within a certain time window (e.g., 10 seconds to 30 seconds), and then by taking the group average for all fish.

[0102] Step S53: Calculate the spatial distribution density of the fish school, the aggregation center point, and the distance distribution between each fish and the aggregation center based on the position coordinates of all fish at the same time. It can be understood that the spatial distribution parameters reflect the collective arrangement of the fish school in the sub-region and are the core indicators for determining whether the fish school has abnormal aggregation or dispersion.

[0103] Spatial distribution density is the ratio of the number of fish detected at the current moment within a given sub-region area to the normal stocking density (based on historical health statistics) within that sub-region. A spatial distribution density > 1 indicates that the fish are congregating in that sub-region, while a spatial distribution density < 1 indicates that they are dispersing or leaving. In one implementation, when ammonia nitrogen levels are too high, largemouth bass will change from a dispersed state to concentrated activity, resulting in a significant increase in spatial distribution density. Conversely, a decrease in dissolved oxygen concentration may force the fish to gather in the corners of the pond bottom.

[0104] The aggregation center point is calculated by taking the coordinates of all fish as a two-dimensional point set and calculating its centroid. The change in the position of the aggregation center point can reflect the collective migration trend of the fish school. When the fish school is frightened or stressed by water quality, its aggregation center point will move towards the water inlet (where dissolved oxygen is high) or the water surface (where fish float when oxygen is low).

[0105] Calculate the Euclidean distance between each fish and the aggregation center to determine the distance distribution of all fish. The dispersion of the distance distribution (expressed as standard deviation) reflects the cohesion of the fish school: a high-density healthy fish school usually maintains a moderate degree of aggregation, with a relatively concentrated but non-overlapping distance distribution; excessive aggregation (the distance distribution narrows sharply) often means that the fish school is under stress and huddled together, while excessive dispersion (the distance distribution widens significantly) may mean that the fish school is in a feeding state (uniform foraging) or the fish school density is too low.

[0106] Step S54: Calculate the fish school turning frequency and synchronization index based on the time series of the orientation angle;

[0107] In practice, the turning frequency (times / minute) is calculated based on the time series of the fish's facing angle, assuming the facing angle is... (Value range 0~360°, based on the direction the fish's head is pointing), the number of times the angle of change exceeds 30° per unit time (per minute); the turning frequency can reflect the agitation level of the fish: healthy fish turn less when swimming in the water and their trajectory is relatively straight; fish in a state of stress will frequently change direction.

[0108] In implementation, in AOMO-FishTracker, by introducing a rotated bounding box (OBB) during the detection phase, a unique forward orientation is defined for each fish. This method adds a lightweight angle regression branch to the YOLO detection head, and the orientation angle can be obtained by geometric derivation to achieve accurate modeling of the fish's orientation.

[0109] Understandably, the formula for calculating the directional consistency index and the swimming speed correlation coefficient is as follows:

[0110] ,

[0111] In the formula: Let be the velocity vector of the i-th fish. The average speed direction of the fish group (weighted average) is represented by the value of Ssync, which ranges from [-1, 1]. 1 indicates complete synchronization (all fish move in the same direction), 0 indicates complete randomness, and -1 indicates completely opposite movement. Therefore, a synchronicity index that is too high or too low may indicate a state of stress.

[0112] The fish activity parameters include average swimming speed, spatial distribution density, turning frequency, and synchronization index.

[0113] Step S6: Determine whether the fish school is under stress and the stress level based on the fish school activity parameters, wherein:

[0114] Step S61: If the average swimming speed is not within the speed reference range and / or the spatial distribution density is within the density reference range, then the fish population in the corresponding aquaculture sub-area is determined to be in a state of stress; otherwise, the fish population in the corresponding aquaculture sub-area is determined not to be in a state of stress.

[0115] It is understandable that average swimming speed and spatial distribution density are the most intuitive measures of the macroscopic movement state of fish groups: (1) Both respond fastest to environmental fluctuations (dissolved oxygen decrease, ammonia nitrogen increase, temperature change), and can usually be detected within minutes to tens of minutes when water quality parameters deviate from the normal range; (2) Macroscopic aggregation density and average speed of the group are less affected by occasional occlusion or small detection errors, and are suitable as stable trigger signals; (3) Fish swimming activities are directly related to metabolic oxygen consumption rate, swimming speed deviating from the baseline range usually reflects changes in the body's metabolic state, and spatial distribution density is directly related to the fish group's perception and feedback to the local water quality environment;

[0116] In practice, there are significant differences in the suitable swimming speed baseline for different fish species. Therefore, the speed baseline range should be adjusted according to the fish species and their body length. When raising fish fry, there should be a definite overall distribution density, so that the density baseline range corresponding to each farming sub-area can be calculated.

[0117] Step S62: In response to the fish population in the aquaculture sub-area being under stress, the stress level is determined based on the turning frequency and the synchronization index;

[0118] Step S63: Calculate the frequency deviation percentage and synchronization deviation percentage of the steering frequency and synchronization index, respectively; in practice, the frequency deviation percentage = (measured steering frequency - reference frequency) ÷ reference frequency × 100%, and the synchronization deviation percentage = (measured synchronization index - reference synchronization index) ÷ reference synchronization index × 100%;

[0119] Step S64: Determine the stress level based on the frequency deviation percentage and the synchronization deviation percentage, wherein:

[0120] If the frequency deviation percentage is greater than the frequency deviation threshold and the synchronization deviation percentage is greater than the synchronization deviation threshold, then the stress level is determined to be severe stress. It is understood that at this stage, the fish school simultaneously exhibits agitation (high turning frequency) and a collapse in group coordination (synchronization index is far from the benchmark), indicating that the body is in a state of high energy consumption stress and large-scale emergency interventions such as emergency oxygenation and water sample testing are required immediately.

[0121] If the percentage of frequency deviation is greater than the frequency deviation threshold or the percentage of synchronization deviation is greater than the synchronization deviation threshold, it is determined to be moderate stress; it is understood that at this stage, the fish's agitated behavior (increased turning frequency) or decreased group coordination (change in synchronization) begins to appear;

[0122] If the percentage of frequency deviation is not greater than the frequency deviation threshold or the percentage of synchronization deviation is not greater than the synchronization deviation threshold, it is determined to be mild stress. It can be understood that if the fish population meets the stress criteria but the deviation of fine motor skills has not exceeded the threshold, it means that the environmental stress factor has reached the level of triggering basic behavioral changes, but the impact on the group structure is still within a slightly controllable range.

[0123] In practice, both the frequency deviation threshold and the synchronization deviation threshold are set to 20% to 40%.

[0124] Step S7, determining the control strategy for the aquaculture sub-region based on the stress level, including:

[0125] In response to mild stress, the air supply of the microporous aeration discs at the bottom of the aquaculture sub-area was increased, raising the dissolved oxygen concentration in the sub-area to 1.1 to 1.2 times the upper limit of the baseline dissolved oxygen value for that sub-area. It is understood that mild stress is usually caused by a slight decrease in dissolved oxygen or local water flow obstruction. At this time, although the fish behavior shows initial abnormalities (such as a slight decrease in swimming speed and slight aggregation), it has not yet caused substantial damage to the body.

[0126] In response to moderate stress, while increasing aeration, the opening of the inlet valve of the aquaculture sub-area was increased, and the water circulation flow rate of the sub-area was increased by 20% to 30% to accelerate the dilution of ammonia nitrogen or nitrite accumulated in the area. It is understood that moderate stress means that the behavior of the fish has deviated beyond the range that can be explained by a single dissolved oxygen factor (such as a significant increase in turning frequency), which usually indicates that ammonia nitrogen, nitrite or organic matter has begun to accumulate in the water. At this time, simply increasing dissolved oxygen is not enough to eliminate the stress source, and water circulation must be accelerated to dilute and remove toxic metabolites.

[0127] In response to severe stress, the emergency oxygenation system was activated to supplement pure oxygen to the aquaculture sub-area and send an alarm message to the management terminal. It is understandable that severe stress (significant deviation in both turning frequency and synchronicity) indicates that the fish population is in an endangered state, most likely caused by severe hypoxia, acute ammonia / nitrite poisoning, or disease outbreak. At this time, local micropore aeration and water circulation regulation can no longer quickly reverse the situation, and extreme intervention must be taken.

[0128] Specifically, the surface acquisition unit and the underwater acquisition unit respectively acquire the growth status parameters of the plants in each of the aforementioned aquaculture sub-areas and underwater videos;

[0129] The above-water data acquisition unit is a multi-rotor UAV equipped with a hyperspectral camera, a multispectral camera, or an RGB camera, and is equipped with a global positioning module and an inertial navigation module; in practice, the data acquisition frequency of the above-water data acquisition unit is usually 1 to 2 times per day;

[0130] Understandably, based on the coordinates of each aquaculture sub-region divided in step S1, drone cruise waypoints are pre-planned to ensure that there is a hovering collection point directly above each sub-region; the drone automatically takes off according to a preset cycle, flies along the flight path to each waypoint, hovers, and triggers the camera to collect multispectral / RGB images of the plants in that sub-region; the plant canopy in each sub-region is extracted through image segmentation algorithms, and the average plant height (based on the difference between the drone and the ground), leaf area index (LAI, retrieved through vegetation index inversion), and leaf color index (such as NDVI, GCC, etc.) are calculated; after completing the collection of all sub-regions, the drone automatically returns to the landing pad, connects to the charging device, and recharges for the next mission;

[0131] The underwater acquisition unit is an underwater robot comprising a body, a thruster, an underwater light, a waterproof camera unit, a depth sensor, and a wireless communication module; in practice, the acquisition frequency of the underwater acquisition unit is typically 1 to 3 times per week.

[0132] In practice, when the underwater robot is not on a mission, it docks at the charging station, and the main control chip enters a low-power mode (power consumption < 1W). Only the wireless receiver that receives the wake-up command remains in standby mode, and the charging station floats and charges its battery to ensure it can depart at any time. When the heterogeneous growth condition is triggered in step S4, the surface base station sends a wake-up command to the underwater robot corresponding to the sub-area. The robot starts its thrusters, leaves the dock, and autonomously navigates to the target area using the preset coordinates of the sub-area and its own depth sensor and inertial navigation. After reaching the target area, it cruises along a predetermined trajectory, while turning on the lights and camera unit to continuously collect underwater video for 10 minutes at a frame rate of more than 1 frame / second. After the collection is completed, it autonomously returns to the charging station and uploads the video data to the data processing center via wireless communication or the dock's wired interface. Then it re-enters the low-power sleep state and begins charging.

[0133] Specifically, in response to the normal initial screening area where the fish population is not under stress, the sampling frequency of the underwater sampling unit is increased.

[0134] The underwater robot is normally docked at a charging station at the edge of the aquaculture area and is in a low-power sleep state.

[0135] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0136] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart aquaculture full life-cycle management method, characterized in that, include: Step S1: Obtain the bottom topographic slope, water circulation equipotential lines, and light and shadow boundaries of the shading structure above the aquaculture area to divide the aquaculture area into several aquaculture sub-regions. Step S2: Periodically collect the growth status parameters of plants in each of the cultivation sub-regions, and determine the region type of the cultivation sub-region based on the degree of deviation between the growth status parameters and the predicted status parameters. The growth status parameters include plant height, leaf area index, and leaf color index, and the region types include pre-screening abnormal regions and pre-screening normal regions. Step S3: Compare the growth status parameters of each of the initial screening abnormal regions with those of the adjacent culture sub-regions, and determine whether the initial screening abnormal region meets the heterogeneous growth conditions based on the comparison results. Step S4: In response to the satisfaction of heterogeneous growth conditions, the dormant camera unit is awakened and underwater video is acquired of the abnormal area in the initial screening. Step S5: Analyze the fish behavior of the collected underwater video and calculate the fish activity parameters; Step S6: Determine whether the fish are in a state of stress and the stress level based on the fish activity parameters; Step S7: Determine the control strategy for the aquaculture sub-region based on the stress level; The control actions include increasing local aeration, adjusting water circulation flow, and initiating emergency oxygenation.

2. The intelligent aquaculture full life cycle management method according to claim 1, characterized in that, In step S1, the process of dividing the aquaculture area into several aquaculture sub-areas includes: Step S11: Divide the continuous areas where the slope difference of the pool bottom is less than the preset slope threshold into the same terrain unit; Step S12: Divide the regions where the equipotential linear density difference of the water circulation is less than a preset density threshold into the same hydraulic unit; Step S13: Divide the area where the difference in daily average solar radiation is less than a preset solar threshold into the same solar unit according to the light and shadow boundary of the upper shading structure. Step S14: The spatial intersection of the terrain unit, hydraulic unit, and illumination unit is determined as the aquaculture sub-region.

3. The intelligent aquaculture full life cycle management method according to claim 1, characterized in that, The growth status parameters of plants in each of the aforementioned aquaculture sub-regions and underwater videos were collected using both surface and underwater acquisition units. The waterborne data acquisition unit is a multi-rotor UAV equipped with a hyperspectral camera, a multispectral camera, or an RGB camera, and is also equipped with a global positioning module and an inertial navigation module. The underwater acquisition unit is an underwater robot comprising a body, a thruster, an underwater lighting lamp, a waterproof camera unit, a depth sensor, and a wireless communication module; The underwater robot is normally docked at a charging station at the edge of the aquaculture area and is in a low-power sleep state.

4. The intelligent aquaculture full life cycle management method according to claim 1, characterized in that, In step S2, the process of determining the region type of a single aquaculture sub-region includes: Step S21: Pre-construct plant growth prediction models for each of the aforementioned aquaculture sub-regions. The prediction models take historical growth data, ambient temperature, light intensity, and aquaculture density as inputs and the predicted state parameters at the current moment as outputs. Step S22: Calculate the Euclidean distance between the measured growth state parameters and the predicted state parameters, and sum the weighted Euclidean distances to obtain the comprehensive deviation. Step S23: Determine the region type of the corresponding aquaculture sub-region based on the comprehensive deviation, wherein: Step S24: In response to the overall deviation being greater than the deviation threshold, the aquaculture sub-region is determined to be an abnormal region in the initial screening. Conversely, if the condition is not normal, it is identified as a normal area in the initial screening.

5. The intelligent aquaculture full life cycle management method according to claim 1, characterized in that, In step S3, the growth status parameters of each of the initially screened abnormal regions are compared with those of adjacent culture sub-regions. Based on the comparison results, it is determined whether the initially screened abnormal region meets the heterogeneous growth conditions, including: Step S31: Determine all aquaculture sub-regions that are geographically adjacent to the abnormal area in the initial screening, and calculate the proportion of the normal area in the initial screening. Step S32: Calculate the comprehensive deviation difference between the abnormal area in the initial screening and each adjacent normal area in the initial screening; Step S33: Calculate the average value of each of the comprehensive deviation differences and record it as the average deviation difference; Step S34: Determine whether the heterogeneous growth conditions are met based on the stated proportion and the average deviation difference, wherein: If the proportion is greater than a preset proportion and the average deviation difference is greater than a preset difference, then the heterogeneous growth condition is determined to be met. Conversely, if the conditions for heterogeneous growth are not met, it is determined that the conditions are not met.

6. The intelligent aquaculture full life cycle management method according to claim 1, characterized in that, In step S5, the process of analyzing fish behavior and calculating fish activity parameters from the acquired underwater video includes: Step S51: Detect fish bodies and track targets in the underwater video frames, extracting the position coordinates, orientation angle, and body length information of each fish frame by frame; Step S52: Calculate the instantaneous swimming speed and average swimming speed of each fish based on the position coordinates of multiple consecutive frames; Step S53: Calculate the spatial distribution density of the fish school, the aggregation center point, and the distance distribution between each fish and the aggregation center based on the position coordinates of all fish at the same time. Step S54: Calculate the fish school turning frequency and synchronization index based on the time series of the orientation angle; The fish activity parameters include average swimming speed, spatial distribution density, turning frequency, and synchronization index.

7. The intelligent aquaculture full life cycle management method according to claim 6, characterized in that, In step S6, determining whether the swimmer is in a state of stress based on the average swimming speed and spatial distribution density includes: If the average swimming speed is not within the speed reference range and / or the spatial distribution density is within the density reference range, then the fish population in the corresponding aquaculture sub-area is determined to be in a state of stress.

8. The intelligent aquaculture full life cycle management method according to claim 7, characterized in that, In step S6, in response to the fish population in the aquaculture sub-area being under stress, the stress level is determined based on the turning frequency and the synchronization index; Calculate the percentage deviation of the frequency and the percentage deviation of the synchronization index for the steering frequency and the synchronization index, respectively. The stress level is determined based on the percentage of frequency deviation and the percentage of synchronization deviation, wherein: If both the frequency deviation percentage and the synchronization deviation percentage are greater than the corresponding deviation threshold, the stress level is determined to be severe stress. If the percentage deviation of frequency or the percentage deviation of synchronization is greater than the corresponding deviation threshold, it is determined to be moderate stress. If both the frequency deviation percentage and the synchronization deviation percentage are not greater than the corresponding deviation threshold, then it is determined to be mild stress.

9. The intelligent aquaculture full life cycle management method according to claim 1, characterized in that, In step S7, the control strategy for the aquaculture sub-region is determined based on the stress level, including: In response to mild stress, the air supply to the microporous aeration discs at the bottom of the breeding area was increased; In response to moderate stress, the opening of the inlet valve of the aquaculture sub-area was increased while aeration was increased; In response to severe stress, the emergency oxygenation system was activated to replenish pure oxygen to the breeding sub-area and send an alarm message to the management terminal.

10. The intelligent aquaculture full life cycle management method according to claim 7, characterized in that, In step S4, in response to the initial screening of normal areas where the fish population is not under stress, the sampling frequency of the underwater sampling unit is increased.

Images