A method of increasing survival of juvenile lobsters in the planktonic stage

Abstract

The present application relates to the technical field of corrugated lobster breeding, and discloses a method for improving survival rate of planktonic larvae of corrugated lobster. The method divides the larvae density monitoring area according to the physical structure parameters of the breeding pond, continuously collects the initial water environment parameters to obtain the time series data of the water environment, and based on the time series data, adopts the dynamic time axis generation technology to mark the key water quality fluctuation nodes to obtain the water quality change graph with time mark. Combined with the spatial distribution information of the larvae density monitoring area and the water quality change graph, the multi-dimensional environmental factor mapping processing is carried out to generate the larvae survival environment stress matrix. The adaptive learning mechanism is introduced to carry out the pattern recognition processing on the dissolved oxygen fluctuation section and the ammonia nitrogen accumulation point in the stress matrix, and the environmental stress response rule library is established. The rule library is used for dynamically analyzing the breeding equipment operation instructions to generate the feeding path optimization strategy and the oxygenation intensity adjustment strategy.

Description

Technical Field

[0001] This invention relates to the field of wavy lobster farming technology, specifically a method for improving the survival rate of wavy lobster larvae during their planktonic stage. Background Technology

[0002] As a marine crustacean with high economic value, the artificial breeding industry of the striped lobster has gradually developed in recent years. In the lobster's breeding cycle, the planktonic larval stage is a critical period for its growth and development. During this stage, the larvae are extremely sensitive to changes in their living environment; even slight fluctuations in environmental factors can significantly affect their survival. Currently, in the breeding of striped lobster planktonic larvae, breeders typically manage the breeding environment based on experience, lacking precise monitoring and classification of larval density in different areas of the breeding pond. Due to differences in the physical structure of the breeding ponds, such as pond shape, water depth distribution, and water flow direction, the distribution of aquatic environmental parameters in different areas of the pond is uneven. Monitoring only the overall environmental parameters cannot accurately reflect the actual survival status of the larvae in each area.

[0003] In monitoring aquatic environmental parameters, existing technologies mostly employ intermittent sampling, making it difficult to obtain continuous time-series data of the aquatic environment and consequently, to capture dynamic changes in water quality in a timely manner. When water quality fluctuates, aquaculture operators often cannot accurately identify key fluctuation points or predict water quality trends in advance, only taking countermeasures after larvae exhibit obvious abnormal reactions, by which time their survival has already been adversely affected. Furthermore, when analyzing the correlation between environmental factors and larval survival, existing methods typically consider only one or a few environmental factors, failing to effectively combine the spatial distribution information of larval density with multi-dimensional environmental factors. This results in an inability to comprehensively construct the stress status of the larval survival environment, leading to an unscientific and unreasonable assessment of the larval survival environment.

[0004] Current aquaculture management lacks effective pattern recognition and response mechanisms to changes in key environmental factors such as dissolved oxygen and ammonia nitrogen. Fluctuations in dissolved oxygen and accumulation of ammonia nitrogen are significant factors affecting the survival of planktonic larvae of the wavy crayfish. Different dissolved oxygen fluctuation ranges and ammonia nitrogen accumulation points exert varying stress on larvae. However, there is currently no systematic method to identify the changing patterns of these key factors, nor has a corresponding environmental stress response rule base been established, resulting in a lack of targeted and timely operational instructions for aquaculture equipment. For example, the control of aeration equipment often employs a fixed-intensity aeration method, failing to dynamically adjust according to real-time fluctuations in dissolved oxygen. In feeding operations, the planning of feeding paths does not consider larval density distribution and water quality conditions, potentially leading to overfeeding in some areas causing water quality deterioration, or underfeeding in others affecting larval growth. These problems all restrict the improvement of the survival rate of planktonic larvae of the wavy crayfish. Summary of the Invention

[0005] The purpose of this invention is to provide a method for improving the survival rate of planktonic larvae of wavy lobsters, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a method for improving the survival rate of planktonic larvae of the wavy lobster, the method comprising:

[0007] Based on the physical structure parameters of the aquaculture pond, the larval density monitoring area was divided, and the initial water environment parameters were continuously collected and processed to obtain water environment time series data.

[0008] Based on the water environment time series data, key water quality fluctuation nodes are marked using dynamic time axis generation technology to obtain a water quality change map with time markers.

[0009] Based on the spatial distribution information of the larval density monitoring area, and combined with the water quality change map, a multi-dimensional environmental factor mapping process is performed to generate a larval survival environment pressure matrix.

[0010] An adaptive learning mechanism is introduced to perform pattern recognition processing on dissolved oxygen fluctuation segments and ammonia nitrogen accumulation points in the environmental stress matrix of the larvae's survival environment, and to establish an environmental stress response rule base.

[0011] Using the environmental stress response rule base, the operation instructions of the aquaculture equipment are dynamically parsed and processed to generate feeding path optimization strategies and oxygenation intensity adjustment strategies.

[0012] Preferably, based on the water environment time series data, key water quality fluctuation nodes are marked using dynamic time axis generation technology to obtain a water quality change map with time stamps, including:

[0013] Extract dissolved oxygen concentration curves and temperature change curves from time-series data of aquatic environment;

[0014] Based on the abrupt gradient and duration threshold of the dissolved oxygen concentration curve, the first type of critical water quality fluctuation nodes are marked;

[0015] By combining the periodic characteristics of the temperature change curve and the timestamps of the first type of key water quality fluctuation nodes, the nodes of composite water quality anomaly events are marked.

[0016] The composite water quality anomaly event nodes are time-series aligned with the data of the larval development stage to generate the water quality change map.

[0017] Preferably, based on the spatial distribution information of the larval density monitoring area, and combined with the water quality change map, a multi-dimensional environmental factor mapping process is performed to generate a larval survival environment pressure matrix, including:

[0018] Obtain the geometric center coordinates and boundary topology of the larval density monitoring area;

[0019] Historical dissolved oxygen fluctuation data at the corresponding location in the water quality change map are matched based on the geometric center coordinates.

[0020] Based on the boundary topology, the gradient difference of environmental parameters between adjacent monitoring areas is calculated;

[0021] By integrating the historical dissolved oxygen fluctuation data and the gradient difference of environmental parameters, a three-dimensional environmental pressure distribution model is constructed.

[0022] Preferably, an adaptive learning mechanism is introduced to perform pattern recognition processing on the dissolved oxygen fluctuation range and ammonia nitrogen accumulation points in the larval survival environment stress matrix, and to establish an environmental stress response rule base, including:

[0023] Extract continuous spatiotemporal blocks with dissolved oxygen concentrations below a critical threshold from the environmental pressure matrix of larvae's survival environment;

[0024] Identify spatial clusters of ammonia nitrogen concentrations exceeding safe levels and their duration characteristics;

[0025] Based on the spatiotemporal correlation of the continuous spatiotemporal blocks and spatial aggregation points, environmental stress coupling event tags are generated;

[0026] The decision weight parameters in the environmental stress response rule base are updated using an incremental learning algorithm.

[0027] Preferably, the environmental stress response rule base is used to dynamically parse and process the operation instructions of the aquaculture equipment to generate feeding path optimization strategies and oxygenation intensity adjustment strategies, including:

[0028] Input the real-time collected larval distribution density data into the environmental stress response rule base;

[0029] The decision weight parameters in the environmental stress response rule base are called to calculate the allocation scheme for the duration of the feeding device in the larval density monitoring area;

[0030] Based on the spatial distribution heat map of ammonia nitrogen accumulation points, a multi-level intensity control command sequence for the oxygenation equipment is generated.

[0031] Preferably, after generating the feeding path optimization strategy and the oxygenation intensity adjustment strategy, the method further includes:

[0032] The feeding path optimization strategy drives the movement trajectory of the feeding device, while acquiring image data of larval aggregation behavior.

[0033] The dwell time allocation scheme is adjusted in real time based on the changes in motion vectors in the juvenile grouping behavior image data.

[0034] Preferably, the dwell time allocation scheme is adjusted in real time based on the changes in motion vectors in the juvenile grouping behavior image data, including:

[0035] Extract motion direction consistency indicators from juvenile group aggregation behavior image data;

[0036] When the consistency index of movement direction exceeds the aggregation threshold, the feeding delay mechanism in high-density areas is triggered.

[0037] By combining the multi-level intensity control command sequence of the oxygenation equipment, the dissolved oxygen compensation coefficient under the feeding delay mechanism is calculated.

[0038] Preferably, the method further includes:

[0039] Based on the nodes of complex water quality anomalies in the water quality change map, an emergency water exchange time window is pre-generated;

[0040] During the emergency water exchange time window, the feeding path optimization strategy is simultaneously disabled and the water replacement protocol is initiated.

[0041] Preferably, after initiating the water replacement protocol, the method further includes:

[0042] Monitor the deviation between the water environment parameters after replacement and the pressure matrix of the larvae's living environment;

[0043] When the deviation value continues to exceed the fault tolerance threshold, the learning parameters of the environmental stress response rule base are reinitialized.

[0044] Preferably, after reinitializing the learning parameters of the environmental stress response rule base, the method further includes:

[0045] Based on the latest results of the larval density monitoring area division, the spatial resolution of the larval survival environment pressure matrix is ​​iteratively updated.

[0046] The updated larval survival environment stress matrix is ​​input into the retrained environmental stress response rule base to generate an optimized version of the oxygenation intensity adjustment strategy.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This method for improving the survival rate of planktonic larvae in wavy lobsters firstly divides larval density monitoring areas according to the physical structure parameters of the culture pond, and then continuously collects and processes initial aquatic environmental parameters to obtain time-series data of the aquatic environment. This changes the traditional situation of extensive larval density monitoring and discontinuous environmental parameter collection in aquaculture. Dividing monitoring areas according to the physical structure of the culture pond can accurately pinpoint the larval distribution in different areas, avoiding monitoring blind spots caused by differences in pond structure, and allowing aquaculture personnel to clearly understand the actual number and distribution of larvae in each area. Furthermore, the continuous time-series data of the aquatic environment can completely present the dynamic changes in water quality parameters, providing comprehensive data support for subsequent accurate analysis of water quality conditions, making the monitoring of the larval survival environment more targeted and accurate.

[0049] Based on time-series data of the aquatic environment, key water quality fluctuation nodes are marked using dynamic timeline generation technology to obtain time-stamped water quality change maps. This transforms abstract water quality data into intuitive visual maps. Aquaculture personnel can quickly identify specific time points where significant water quality changes occur through these maps, clearly understanding the trends and patterns of water quality changes. This eliminates the need to rely on experience or wait until water quality problems become apparent, allowing for earlier detection of potential water quality risks and providing more time for timely intervention, thus reducing the adverse effects of water quality fluctuations on larvae.

[0050] Based on the spatial distribution information of larval density monitoring areas, and combined with water quality change maps, multi-dimensional environmental factor mapping was performed to generate a larval survival environment stress matrix, enabling a comprehensive assessment of the larval survival environment. This process correlates spatial differences in larval density with multi-dimensional environmental factors such as dissolved oxygen, ammonia nitrogen, temperature, and salinity, moving beyond the consideration of single environmental factors and comprehensively reflecting the magnitude of environmental pressure and major stressors faced by larvae in different areas. Aquaculture personnel can use the stress matrix to identify areas requiring focused attention and improvement, avoiding the problem of blindly adjusting environmental parameters in traditional management and making environmental control more scientific.

[0051] An adaptive learning mechanism was introduced to perform pattern recognition processing on dissolved oxygen fluctuation segments and ammonia nitrogen accumulation points in the environmental stress matrix of larvae, and an environmental stress response rule base was established, providing a strong guarantee for intelligent management of the aquaculture environment. Pattern recognition can accurately distinguish dissolved oxygen fluctuations and ammonia nitrogen accumulation with different characteristics, clarifying the degree of environmental stress on larvae under different patterns. The establishment of the environmental stress response rule base combines these recognition results with corresponding coping strategies, forming a systematic and reliable set of operational guidelines. This replaces the previous fragmented coping methods that relied on manual experience, ensuring a clear direction for handling different environmental stress situations.

[0052] By utilizing an environmental stress response rule base to dynamically analyze and process the operation commands of aquaculture equipment, feeding path optimization strategies and aeration intensity adjustment strategies are generated, further improving the accuracy and effectiveness of aquaculture management. Regarding feeding, the optimized feeding path combines larval density distribution and water quality conditions, ensuring reasonable feed placement in areas with high larval density, avoiding overfeeding or underfeeding, reducing water pollution caused by feed residue, and ensuring larvae receive sufficient nutrition. Regarding aeration, adjusting the aeration intensity based on dissolved oxygen fluctuations maintains the dissolved oxygen content in the water within a suitable range for larval survival, avoiding harm caused by excessively high or low dissolved oxygen levels. Through the implementation of these optimization strategies, a more suitable living environment can be created for the planktonic larvae of the wavy lobster, reducing the impact of environmental stress on the larvae and thus helping them successfully survive the planktonic stage. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the working principle of the method for improving the survival rate of planktonic larvae of the wavy lobster as described in this invention.

[0054] Figure 2 A schematic diagram illustrating the working principle of a method for creating time-stamped water quality change maps.

[0055] Figure 3 A schematic diagram illustrating the working principle of the environmental pressure matrix method for larvae survival.

[0056] Figure 4 This is a diagram illustrating the working principle of the environmental stress response rule base method. Detailed Implementation

[0057] 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.

[0058] Please see Figure 1 This invention provides a method for improving the survival rate of planktonic larvae of the wavy lobster, the method comprising:

[0059] The physical structural parameters of the aquaculture pond are set, including the pond's geometry, depth distribution, the relative positions of the inlet and outlet, and the placement of internal obstacles or functional zones. Based on these parameters, the pond is divided into multiple juvenile density monitoring zones. The division principle must ensure that the aquatic environment in each zone is relatively uniform and facilitates sensor deployment. Initial aquatic environmental parameter acquisition is achieved through a distributed sensor network, including but not limited to dissolved oxygen sensors, temperature sensors, pH sensors, ammonia nitrogen sensors, and turbidity sensors. The sensors collect raw data at a fixed frequency, for example, every 5 minutes, and transmit it to the central processing unit (CPU) via a wireless transmission module. The CPU filters and denoises the raw data, uses a moving average algorithm to eliminate transient interference, and stores the processed data in a time-series database, forming a time-series aquatic environmental data. This time-series data contains multi-parameter water quality readings at each monitoring point at different timestamps, providing a structured data source for subsequent analysis.

[0060] The dynamic timeline generation technology employs a piecewise linear fitting algorithm to process the dissolved oxygen concentration curve, calculating the concentration change gradient between adjacent time points. When the gradient exceeds a preset threshold (e.g., 0.3 mg / L / min) and lasts for more than 5 minutes, it is marked as a first-category critical water quality fluctuation node. The temperature change curve is analyzed using Fourier transform to extract its periodicity, identifying periods of abnormal intraday temperature fluctuations. The intersection of these abnormal temperature periods with the timestamps of the first-category critical water quality fluctuation nodes is calculated; if overlapping time windows exist, they are marked as composite water quality anomaly event nodes. Time-series alignment is performed by matching water quality event nodes with biological learning data from a larval developmental stage database, including larval molting cycles and changes in feeding intensity. Different levels of anomalies are then color-coded in the water quality change map; for example, red represents high-risk composite events, and yellow represents single-parameter anomalies.

[0061] The spatial distribution information of the larval density monitoring area is digitally represented by the area identification code and its geometric center coordinates obtained through GPS mapping or relative coordinate system conversion. The boundary topology is stored using a graph structure, with each area as a node and the connections between adjacent areas as edges. Historical dissolved oxygen fluctuation data in the water quality change map are mapped to the geometric center points of each area using spatial interpolation algorithms such as Kriging interpolation, forming a dissolved oxygen fluctuation intensity distribution map. The environmental parameter gradient difference is calculated using the Laplace operator to handle parameter differences between adjacent areas; for example, the difference in dissolved oxygen concentration between area A and area B is calculated and divided by their boundary distance. The fusion process integrates the dissolved oxygen fluctuation intensity and gradient difference into three-dimensional grid data using a weighted superposition algorithm. The grid cell value represents the environmental pressure index, generating a larval survival environmental pressure matrix. This matrix is ​​stored in the form of a three-dimensional array, with dimensions corresponding to the time axis, spatial X-axis, and spatial Y-axis, respectively.

[0062] The adaptive learning mechanism employs an online machine learning framework. During initialization, the dissolved oxygen critical threshold is preset to 4.2 mg / L, and the ammonia nitrogen safety value is 0.1 mg / L. The system continuously monitors continuous regions in the pressure matrix where the dissolved oxygen concentration is below the critical threshold. When such regions persist for more than 30 minutes and their spatial extent exceeds 15% of the total monitored area, they are marked as dissolved oxygen fluctuation zones. Ammonia nitrogen accumulation point identification uses density clustering algorithms such as DBSCAN to find sensor clusters where ammonia nitrogen concentrations continuously exceed limits and are spatially close. Spatiotemporal correlation analysis calculates the spatial overlap and temporal synchronization between fluctuation zones and accumulation points. If the overlap exceeds 60% and the time deviation is less than 10 minutes, a "hypoxia-ammonia toxicity combined stress" event label is generated. The incremental learning algorithm uses stochastic gradient descent to update the decision weight parameters, recalculating the weights each time a new stress event is added, allowing the rule base to gradually adapt to the actual aquaculture environment.

[0063] The decision weight parameters of the environmental stress response rule base include the priority coefficient of feeding equipment and the response intensity coefficient of aeration equipment. Real-time larval distribution density data are collected through an underwater camera system, and the number of larvae in each area is calculated using an image recognition algorithm. The feeding path optimization strategy generation process is as follows: the demand weight of each area is calculated based on the density data, and then combined with parameters such as equipment movement speed constraints and feeding amount limits in the decision weight parameters, a greedy algorithm is used to generate the movement path of the feeding device, in which the dwell time in high-density areas is allocated according to the weight ratio. The aeration intensity adjustment strategy is generated based on the ammonia nitrogen accumulation point heat map through kernel density estimation, dividing the aquaculture pond into a core aeration zone, a buffer zone, and a safe zone, and generating a corresponding intensity command sequence for the aeration equipment. For example, the core zone uses pulsed strong aeration at 2.5 mg / L / min, and the buffer zone uses continuous low-intensity aeration at 0.5 mg / L / min.

[0064] Example 1: See Figure 2During implementation, optical dissolved oxygen sensors deployed in the aquaculture ponds collect water data every thirty seconds. These sensors operate on the principle of quenching effect between fluorescent substances and oxygen molecules, inferring dissolved oxygen concentration by measuring changes in fluorescence lifetime. Temperature monitoring uses immersion platinum resistance probes, whose resistance changes linearly with temperature, and is converted into precise readings via a bridge circuit. All sensor signals are transmitted via shielded cables to a data acquisition box, where an analog-to-digital converter converts the analog signals into digital signals, which are then sent to the central processing unit via industrial Ethernet. The central processing unit runs a real-time operating system and performs moving average filtering on the input data, taking the arithmetic mean of ten consecutive samples as the effective output to eliminate abnormal pulses caused by instantaneous water flow or bubble interference. The processed data is stored in a time-series database with timestamps, forming a dataset containing two main curves: dissolved oxygen and temperature. Each data record includes a millisecond-level timestamp, sensor number, and parameter value, providing structured data support for subsequent analysis.

[0065] The abrupt gradient detection of the dissolved oxygen concentration curve employs a discrete differential algorithm, calculating the difference in concentration values ​​between adjacent time points and dividing it by the time interval. When this gradient value consistently exceeds 0.25 mg / L / min and persists for more than eight minutes, the system automatically marks a red warning on the time axis. These marked points are referred to as the first type of critical water quality fluctuation nodes. Temperature change curve analysis utilizes spectral analysis, converting continuous 24-hour temperature data into frequency components to identify abnormal fluctuation periods deviating from the normal daily cycle. The system compares the time information of abnormal temperature periods with that of dissolved oxygen fluctuation nodes. When the overlap exceeds 70%, a composite water quality anomaly event node is generated, representing a situation where multiple environmental parameters are simultaneously abnormal. The event correlation engine employs a multi-threaded parallel processing approach, capable of processing hundreds of sensor data streams simultaneously, ensuring the completion of correlation analysis of the entire day's data within five minutes.

[0066] The generation of the water quality change map requires combining the aforementioned event nodes with larval biological learning data. The larval development database stores the standard developmental timeline of planktonic larvae of the wavy lobster, including biological parameters such as molting cycles and feeding activity patterns. The system uses a dynamic time warping algorithm to align and match the water quality event timeline with the biological development timeline. This algorithm can automatically compensate for rate differences between the two time series. The output map is visualized as a heatmap, with the horizontal axis representing the time series and the vertical axis representing different monitoring area numbers. The color depth of the blocks reflects the severity of the event, with dark red indicating high-risk complex events and light yellow indicating slight anomalies in a single parameter. This map is updated every ten minutes and can be viewed in real time via a monitoring terminal or exported as a structured data file for subsequent analysis.

[0067] Example 2: See Figure 3 During implementation, precise spatial measurements and data acquisition of the aquaculture ponds are required. A total station is used to map the pond boundaries. This instrument calculates distances and angles by emitting infrared laser beams and receiving reflected signals, thus obtaining the three-dimensional coordinates of each corner point of the pond. A laser rangefinder is used to measure the straight-line distances of each section of the pond wall. These raw measurement data are recorded in an electronic handbook and transmitted to computer-aided design software. A two-dimensional plane coordinate system for the aquaculture pond is established in the software, typically with a corner of the pond as the origin, and the X and Y axes extending along the pond wall. Based on the distribution characteristics of planktonic lobster larvae and the water flow characteristics within the pond, the pond area is divided into several monitoring zones. The area of ​​each zone is controlled between four and six square meters, and the shape of the zone can be rectangular or an irregular polygon depending on the actual pond structure. The geometric center coordinates of each monitoring zone are obtained by calculating the arithmetic mean of all boundary points of that zone. These coordinate values ​​are stored as floating-point numbers in the coordinate table of the spatial database. The boundary topology is represented by an adjacency matrix in graph theory. Each monitoring area is regarded as a node. If two areas share a common boundary, they are marked as 1 in the corresponding position of the matrix, otherwise they are marked as 0. This representation method can clearly reflect the spatial adjacency relationship of each area.

[0068] After acquiring spatial information, environmental data mapping processing begins. The system extracts historical dissolved oxygen fluctuation data for the most recent 72 hours from the water quality change map database. This data originates from water quality sensors deployed in various monitoring areas. A spatial matching algorithm associates the geometric center coordinates of each monitoring area with the nearest sensor number, finding the sensor node with the smallest Euclidean distance. The historical dissolved oxygen fluctuation data includes two indicators: fluctuation amplitude and fluctuation frequency. Fluctuation amplitude refers to the range of dissolved oxygen concentration within the time period, while fluctuation frequency refers to the number of times the concentration exceeds a set threshold per unit time. The calculation of the environmental parameter gradient difference uses the finite difference method. For any two adjacent monitoring areas, the difference between their daily average dissolved oxygen values ​​is divided by the boundary line length, and the result is stored in the edge attributes of the topology database. The three-dimensional environmental pressure distribution model is constructed using a gridding method. The entire aquaculture pond is divided into several cubic units using a 1-meter by 1-meter horizontal grid and 0.5-meter vertical layers. The environmental pressure value of each unit is calculated using an inverse distance weighted interpolation algorithm, which assumes that closer monitoring points have a greater impact on the grid unit. The generated environmental pressure distribution model is stored as a three-dimensional matrix. The three dimensions of the matrix represent the east-west grid number, the north-south grid number, and the vertical layer number, respectively. Each matrix element represents the environmental pressure index for that location. This model can be rendered and displayed using 3D visualization software, allowing operators to observe the pressure distribution from different angles. The formula for calculating the environmental pressure index is:

[0069] Pij =α×D ij +β×G ij

[0070] Where: P ij D represents the environmental pressure index of the grid cell in the i-th row and j-th column; ij The dissolved oxygen fluctuation intensity of this unit is represented by G, calculated using inverse distance weighted interpolation; ij The gradient difference of environmental parameters in this unit is obtained by interpolating the gradient difference between adjacent monitoring areas; α and β are weighting coefficients, initially set to 0.6 and 0.4 respectively, and these coefficients can be adjusted according to actual aquaculture conditions. Cross-validation is used in the model validation phase, using a subset of sensor data as the validation set to compare the degree of agreement between model predictions and actual measurements. The model update frequency is set to once every two hours to ensure timely reflection of changes in the aquatic environment. In practical applications, operators can adjust the grid size and layer thickness through the human-computer interface to meet different monitoring accuracy requirements. The system also automatically records various parameters during model generation, including interpolation algorithm type, weighting coefficient values, and grid division accuracy. This log information helps trace the model construction process and facilitate subsequent optimization. The generated juvenile survival environment pressure matrix is ​​stored in a distributed file system, providing a data foundation for subsequent environmental stress analysis.

[0071] Example 3: See Figure 4 During implementation, the system scans and analyzes the environmental pressure matrix of larvae to identify regions where dissolved oxygen concentrations are below a set threshold of four milligrams per liter. The system checks the values ​​of each grid cell layer by layer in the three-dimensional matrix. When a cell's dissolved oxygen value is found to be consistently below this threshold, a connected component analysis algorithm is activated, which checks the status of adjacent grid cells. If adjacent cells are also below the threshold and are continuous in the time dimension, these cells are marked as the same spatiotemporal block. Each block's record includes its spatial extent, duration, and average degree of hypoxia. Ammonia nitrogen concentration monitoring uses a similar method, with a safety value set at 0.1 milligrams per liter. The system uses a density clustering algorithm to find spatial clusters of ammonia nitrogen exceeding the limit. This algorithm identifies clustered regions by calculating the spatial distance and concentration similarity between sensor nodes. For each cluster, the system records its spatial coordinates, duration, and concentration change trend. The spatiotemporal correlation analysis module matches dissolved oxygen fluctuation blocks with ammonia nitrogen clusters, calculating their spatial overlap and temporal synchronization. When the spatial overlap exceeds 60% and the time deviation is less than 10 minutes, the system generates an environmental stress coupling event label, which contains information such as event type, severity level, and scope of impact.

[0072] The adaptive learning mechanism employs an online machine learning framework to handle these environmental stress events. The system uses each detected event as a training sample, extracting feature vectors including parameters such as dissolved oxygen decline rate, ammonia nitrogen accumulation rate, and temperature change amplitude. These feature vectors are correlated with the larval stress response level, which is quantified by the abnormal swimming rate recorded by an underwater camera system. An incremental learning algorithm uses stochastic gradient descent to update model parameters; each new event leads to minor adjustments to the weight parameters. Decision weight parameters include the contribution coefficients of each environmental factor and the equipment response priority, stored in an environmental stress response rule base. The rule base uses a hierarchical structure, divided into three levels: event identification rules, response priority rules, and equipment control rules. The system performs an integrity check on the rule base every 24 hours to verify the logical consistency and timeliness of the rules. Historical versions are retained during rule updates for rollback when necessary.

[0073] Real-time larval distribution density data is acquired through an image processing system. Underwater cameras capture panoramic images every fifteen minutes, which are then transmitted to an image processing server. Deep learning algorithms are used to identify and count individual larvae. The identification results are statistically analyzed by monitoring area, and a larval density weight is calculated for each area. The environmental stress response rule base receives the density data and, combined with decision weight parameters, calculates the feeding plan. The formula for calculating the duration of stay at the feeding equipment is:

[0074]

[0075] Wherein: T k W represents the duration of stay in the k-th region. k Q represents the density weight of the region. k The system prioritizes regions, where C is a constant for total feeding time and N represents the total number of regions. The aeration strategy is generated based on a heat map of ammonia nitrogen accumulation points. The system generates a circular buffer zone centered on each accumulation point and calculates the integral value of the ammonia nitrogen concentration within the buffer zone. Based on the integral value, the aeration intensity is divided into three levels, each corresponding to a different equipment power output. The command sequence is ordered spatially and sent to the equipment controller to ensure the execution order conforms to the water flow characteristics. The system recalculates the strategy parameters every thirty minutes to adapt to dynamic changes in environmental conditions.

[0076] Example 4: In implementation, a track-mounted mobile feeder was used. This device was installed on a double-track system above the rearing pond and moved along a predetermined path driven by a servo motor. The feeding path optimization strategy was converted into equipment control commands, including the movement path coordinate sequence and the dwell time in each monitoring area. The movement trajectory was generated using a piecewise linear interpolation algorithm, converting the path coordinates into motor control pulse signals. Equipment positioning was achieved through real-time feedback of position information from an encoder, ensuring a movement accuracy error of no more than five centimeters. Image data acquisition of larval aggregation behavior was performed using high-speed underwater cameras. These cameras acquired images at a rate of thirty frames per second and transmitted the image data to an image processing server via gigabit Ethernet. Image processing employed a background subtraction algorithm to establish a background model excluding larvae, and extracted moving targets using inter-frame differencing. Motion vector calculation used optical flow, estimating the larvae's movement direction and velocity by analyzing the displacement of pixels between adjacent frames. The system divided the image into ten-by-ten grid cells and performed statistical analysis on the motion vectors within each grid.

[0077] The consistency index of movement direction is calculated by analyzing the angular variance of the movement direction of larvae within each grid cell. The system calculates the movement direction angle of each individual larva and statistically analyzes the angular variance of all individuals within the grid. When the angular variance of a grid is below 0.15, that grid is considered to have high movement direction consistency. The aggregation threshold is set at 40% of the total number of grids. When the number of highly consistent grids exceeds this threshold, the system triggers a high-density area feeding delay mechanism. This mechanism adds an extra delay to the original dwell time, dynamically calculated based on the proportion of consistent grids. Specifically, the delay time equals the proportion of consistent grids minus 0.4, multiplied by the base feeding time. Simultaneously, the system monitors the operating status of the aeration equipment and calculates the dissolved oxygen compensation coefficient based on the current aeration intensity. This coefficient is used to adjust the output power of the aeration equipment to ensure sufficient dissolved oxygen levels are maintained during the feeding delay.

[0078] During system operation, a series of monitoring data will be generated. This data is recorded in the operation log for subsequent analysis. For the main monitoring parameters and their threshold settings, please refer to Table 1.

[0079] Table 1: Monitoring Parameters for Feeding Path Optimization

[0080] Parameter name Threshold type Value / Range Motion direction consistency threshold Variance threshold 0.15 Grid aggregation threshold Percentage threshold 40％ Basic feeding time time constant 120 seconds oxygenation intensity ratio coefficient Coefficient range 0.5-1.2 Position error tolerance Distance threshold 5 centimeters Image processing frame rate Frequency indicators 30 frames per second Delay calculation coefficient proportionality coefficient 1.0

[0081] During operation, the system recalculates the movement direction consistency index every two minutes and dynamically adjusts the feeding delay based on the latest results. The aeration equipment control employs a closed-loop regulation method, monitoring dissolved oxygen concentration in real time and feeding it back to the control system. Equipment operating status data, including motor current, movement speed, and feed quantity, are recorded in real time for system performance evaluation and fault diagnosis. All operating commands and execution results are stored in the operation log database, which can be used for subsequent system optimization and operational analysis. The system also features an alarm function; when abnormal equipment operation or environmental parameters exceeding safe ranges are detected, an alarm will be immediately issued and corresponding safety measures will be taken.

[0082] Example 5: During implementation, the system comprehensively scans and analyzes the environmental pressure matrix of larvae to identify specific areas where dissolved oxygen concentration is below a set threshold. The critical dissolved oxygen threshold is set at four milligrams per liter, based on the physiological characteristics of planktonic lobster larvae. The system checks the numerical records of each grid cell layer by layer in the three-dimensional matrix data structure. When the dissolved oxygen value of a cell is detected to be consistently below the preset threshold, the system automatically activates a connected component analysis algorithm, which systematically checks the real-time status of adjacent grid cells. If adjacent cells are also detected to be below the threshold and maintain continuity in the time dimension, these cells are marked as the same spatiotemporal block. Detailed records for each block include key parameters such as its spatial coordinates, duration, and average degree of hypoxia. Ammonia nitrogen concentration monitoring uses a similar principle and method, with the safety value set at 0.1 milligrams per liter based on larval tolerance studies.

[0083] The system uses a density-based clustering algorithm to locate spatial clusters of ammonia nitrogen exceedances. This algorithm identifies clusters by calculating spatial distances and concentration similarity indices between sensor nodes. For each identified cluster, the system records key data such as its spatial coordinates, duration, and concentration change trends. A spatiotemporal correlation analysis module precisely matches dissolved oxygen fluctuation zones with ammonia nitrogen clusters, calculating their spatial overlap and temporal synchronization indices. When the spatial overlap exceeds 60% and the time deviation is less than ten minutes, the system automatically generates an environmental stress coupling event label, which includes detailed information such as event type, severity level, and impact range. The entire monitoring process employs a distributed computing architecture, with multiple computing nodes processing data from different regions in parallel to ensure real-time monitoring and accuracy. The system also establishes an abnormal data filtering mechanism. By setting reasonable data fluctuation range thresholds, it effectively eliminates data anomalies caused by occasional sensor malfunctions or transient interference, ensuring the reliability of monitoring results.

[0084] The adaptive learning mechanism employs an advanced online machine learning framework to handle these environmental stress events. The system uses each detected event as a training sample, extracting feature vectors including key parameters such as dissolved oxygen decline rate, ammonia nitrogen accumulation rate, and temperature change amplitude. These feature vectors are precisely correlated with the larval stress response level, which is quantified using the abnormal swimming rate recorded by an underwater camera system. The quantification process utilizes image recognition technology, analyzing indicators such as the degree of disorder in larval swimming trajectories and changes in aggregation patterns to establish a stress response assessment model. An incremental learning algorithm uses stochastic gradient descent to update the model parameters; each new event leads to subtle adjustments and optimizations to the weight parameters. Decision weight parameters include important indicators such as the contribution coefficients of various environmental factors and equipment response priorities, stored in an environmental stress response rule base. The rule base uses a hierarchical structure, divided into three main levels: event identification rules, response priority rules, and equipment control rules. The system performs a comprehensive integrity check on the rule base every 24 hours to verify the logical consistency and timeliness of the rules. The system retains historical version records during rule updates to allow for rollback operations when necessary, ensuring system stability and reliability. During the learning process, the system pays particular attention to the seasonal variations in environmental parameters. By establishing time-series prediction models, it anticipates potential environmental stress events, shifting from passive response to proactive prevention. Furthermore, the system incorporates a transfer learning mechanism, enabling rapid adaptation to new environmental characteristics when pond structure or farming scale changes, reducing the time cost of retraining.

[0085] Real-time larval distribution density data is acquired through a high-precision image processing system. Underwater cameras capture panoramic images every fifteen minutes, which are transmitted to an image processing server via a high-speed network. Deep learning algorithms are used to automatically identify and count individual larvae. The image processing process includes four main steps: background modeling, target segmentation, feature extraction, and classification. First, a dynamic background is established using a Gaussian mixture model to effectively eliminate the effects of water ripples and light changes. Then, a watershed algorithm is used for target segmentation to accurately separate overlapping larvae. Next, morphological and texture features of the larvae are extracted. Finally, a convolutional neural network is used for classification. The identification results are statistically analyzed by monitoring area, and the larval density weight for each area is calculated. After receiving the density data, the environmental stress response rule base calculates the optimal feeding plan based on decision weight parameters. The formula for calculating the dwell time of the feeding equipment is:

[0086]

[0087] Wherein: T k W represents the duration of stay in the k-th region. k Q represents the density weight of the region. kThe system represents the regional priority coefficient, where C is a constant for total feeding time, and N represents the total number of regions. The aeration strategy is generated based on a heat map of ammonia nitrogen accumulation points. The system generates a circular buffer zone centered on each accumulation point and calculates the integral value of the ammonia nitrogen concentration within the buffer zone. Based on the integral value, the aeration intensity is divided into three levels, each corresponding to a different equipment power output level. The command sequence is ordered spatially and sent to the equipment controller to ensure that the execution order conforms to the water flow characteristics. The system recalculates the strategy parameters every thirty minutes to adapt to dynamic changes in environmental conditions, ensuring the stability and suitability of the aquaculture environment.

[0088] The entire control process employs a closed-loop control principle, dynamically adjusting control parameters through real-time monitoring of the control effects to achieve precise environmental management. The system also establishes an emergency response mechanism; when a sudden environmental deterioration is detected, the emergency plan can be immediately activated, implementing measures such as enhanced oxygenation and emergency water exchange to minimize the impact of sudden environmental changes on larvae. The entire process achieves closed-loop control from environmental monitoring to intelligent regulation, providing a precise environmental management solution for planktonic larvae of the wavy lobster. All data generated during system operation is fully recorded, forming big data on the aquaculture environment, providing support for subsequent data analysis and system optimization. Through long-term operation and data accumulation, the system can continuously optimize control parameters, improving the accuracy and efficiency of environmental control.

[0089] 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.

[0090] 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 method for improving the survival rate of planktonic larvae of the wavy lobster, characterized in that, include: Based on the physical structure parameters of the aquaculture pond, the larval density monitoring area was divided, and the initial water environment parameters were continuously collected and processed to obtain water environment time series data. Based on the water environment time series data, key water quality fluctuation nodes are marked using dynamic time axis generation technology to obtain a water quality change map with time markers. Based on the spatial distribution information of the larval density monitoring area, and combined with the water quality change map, a multi-dimensional environmental factor mapping process is performed to generate a larval survival environment pressure matrix. An adaptive learning mechanism is introduced to perform pattern recognition processing on dissolved oxygen fluctuation segments and ammonia nitrogen accumulation points in the environmental stress matrix of the larvae's survival environment, and to establish an environmental stress response rule base. Using the environmental stress response rule base, the operation instructions of the aquaculture equipment are dynamically parsed and processed to generate feeding path optimization strategies and oxygenation intensity adjustment strategies; Based on the aforementioned water environment time series data, key water quality fluctuation nodes are marked using dynamic time axis generation technology to obtain a water quality change map with time stamps, including: Extract dissolved oxygen concentration curves and temperature change curves from time-series data of aquatic environment; Based on the abrupt gradient and duration threshold of the dissolved oxygen concentration curve, the first type of critical water quality fluctuation nodes are marked; By combining the periodic characteristics of the temperature change curve and the timestamps of the first type of key water quality fluctuation nodes, the nodes of composite water quality anomaly events are marked. The composite water quality anomaly event nodes are time-series aligned with the data of the larval development stage to generate the water quality change map. Based on the spatial distribution information of the larval density monitoring area, and combined with the water quality change map, multi-dimensional environmental factor mapping processing is performed to generate a larval survival environment pressure matrix, including: Obtain the geometric center coordinates and boundary topology of the larval density monitoring area; Historical dissolved oxygen fluctuation data at the corresponding location in the water quality change map are matched based on the geometric center coordinates. Based on the boundary topology, the gradient difference of environmental parameters between adjacent monitoring areas is calculated; By integrating the historical dissolved oxygen fluctuation data and the gradient difference of environmental parameters, a three-dimensional environmental pressure distribution model is constructed. An adaptive learning mechanism is introduced to perform pattern recognition processing on dissolved oxygen fluctuation segments and ammonia nitrogen accumulation points in the environmental stress matrix of the larvae's survival environment, and to establish an environmental stress response rule base, including: Extract continuous spatiotemporal blocks with dissolved oxygen concentrations below a critical threshold from the environmental pressure matrix of larvae's survival environment; Identify spatial clusters of ammonia nitrogen concentrations exceeding safe levels and their duration characteristics; Based on the spatiotemporal correlation of the continuous spatiotemporal blocks and spatial aggregation points, environmental stress coupling event tags are generated; The decision weight parameters in the environmental stress response rule base are updated using an incremental learning algorithm; Using the aforementioned environmental stress response rule base, the operation instructions of aquaculture equipment are dynamically parsed and processed to generate feeding path optimization strategies and oxygenation intensity adjustment strategies, including: Input the real-time collected larval distribution density data into the environmental stress response rule base; The decision weight parameters in the environmental stress response rule base are called to calculate the allocation scheme for the duration of the feeding device in the larval density monitoring area; Based on the spatial distribution heat map of ammonia nitrogen accumulation points, a multi-level intensity control command sequence for the oxygenation equipment is generated; After generating the feeding path optimization strategy and the oxygenation intensity adjustment strategy, the following is also included: The feeding path optimization strategy drives the movement trajectory of the feeding device, while acquiring image data of larval aggregation behavior. The dwell time allocation scheme is adjusted in real time based on the changes in motion vectors in the juvenile group behavior image data; Based on the changes in motion vectors in the juvenile grouping behavior image data, the dwell time allocation scheme is adjusted in real time, including: Extract motion direction consistency indicators from juvenile group aggregation behavior image data; When the consistency index of movement direction exceeds the aggregation threshold, the feeding delay mechanism in high-density areas is triggered. By combining the multi-level intensity control command sequence of the oxygenation equipment, the dissolved oxygen compensation coefficient under the feeding delay mechanism is calculated.

2. The method for improving the survival rate of planktonic larvae of the wavy lobster according to claim 1, characterized in that, The method further includes: Based on the nodes of complex water quality anomalies in the water quality change map, an emergency water exchange time window is pre-generated; During the emergency water exchange time window, the feeding path optimization strategy is simultaneously disabled and the water replacement protocol is initiated.

3. The method for improving the survival rate of planktonic larvae of the wavy lobster according to claim 2, characterized in that, After initiating the water replacement protocol, the following is also included: Monitor the deviation between the water environment parameters after replacement and the pressure matrix of the larvae's living environment; When the deviation value continues to exceed the fault tolerance threshold, the learning parameters of the environmental stress response rule base are reinitialized.

4. The method for improving the survival rate of planktonic larvae of the wavy lobster according to claim 3, characterized in that, After reinitializing the learning parameters of the environmental stress response rule base, the method further includes: Based on the latest results of the larval density monitoring area division, the spatial resolution of the larval survival environment pressure matrix is ​​iteratively updated. The updated larval survival environment stress matrix is ​​input into the retrained environmental stress response rule base to generate an optimized version of the oxygenation intensity adjustment strategy.

Images