A system and method for non-invasive in situ detection of stress in fish individuals and groups

Abstract

The application provides a fish individual and group stress nondestructive in-situ detection system and method, and relates to the technical field of stress evaluation.The system comprises: a multi-modal sensing module using a flexible sensing chip and an impedance hydrogel electrode to collect real-time breathing, environmental and physiological impedance data of the gill and back of a fish body; a rear-end circuit module to complete data collection, processing and wireless transmission; a stress data determination module to perform feature extraction and nonlinear fusion on the screened data based on a multi-modal data fusion algorithm, to quantify fish stress levels and calculate stress fatigue, and to trigger an early warning and generate an environmental control instruction when a high stress threshold is reached.The application realizes non-invasive, high-precision real-time monitoring and closed-loop feedback control of the stress state of fish during transportation, and significantly improves the survival rate of fish transportation.

Description

Technical Field

[0001] This invention relates to the field of stress assessment technology, and in particular to a non-destructive in-situ detection system and method for stress in individual fish and fish populations. Background Technology

[0002] Fish are susceptible to environmental disturbances (such as temperature and humidity) and physical stimuli (such as collisions and compression) during long-distance transportation, leading to physiological stress responses that affect their survival rate and transportation quality. Traditional stress assessment methods often rely on endpoint detection methods, such as blood biochemical parameter testing (such as blood glucose and cortisol levels), which are not only invasive but also suffer from poor real-time performance and delayed feedback.

[0003] Most existing wearable or adhesive sensors are designed for humans or terrestrial animals, making it difficult to operate stably for extended periods in the slippery, turbulent, and confined environments of aquatic transport. The adaptability of flexible sensors and their electrodes to fish remains to be improved, and multimodal data fusion algorithms cannot effectively identify the dynamic coupling relationships between complex modes. Particularly in group stress assessment, there is a lack of a method that can simultaneously assess individual dynamics and predict group fatigue trends. Furthermore, issues such as data anomalies and missing modes can also lead to decreased model robustness and prevent sustained stable operation. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a non-destructive in-situ detection system and method for individual and group stress in fish. This invention solves the technical problems of existing fish transportation stress assessment technologies, such as the invasiveness and lag of detection methods, poor adaptability and stability of sensors in slippery aquatic environments, and the difficulty of multimodal fusion algorithms in taking into account both individual dynamics and group fatigue prediction, as well as insufficient robustness against interference.

[0005] To achieve the above objectives, the present invention provides the following solution: A non-destructive in-situ detection system for individual and group stress in fish includes: Multimodal sensing module, back-end circuit module, host computer and mobile terminal module, and stress data determination module; The multimodal sensing module is installed on the fish body and is used to collect signals from the gills and dorsal sidewall of the fish body using a flexible sensing chip and an impedance hydrogel electrode to obtain respiratory stress data, environmental stress data, and impedance stress data. The flexible sensing chip includes a flexible substrate, an attitude chip, a temperature and humidity sensing chip, a gas sensing chip, electrode circuitry, and an encapsulation layer. The impedance hydrogel electrode includes a patch substrate, a colloidal electrode, and conductive wires. The back-end circuit module is connected to the multimodal sensing module and the host computer and mobile terminal module respectively, and is used to collect the respiratory stress data, the environmental stress data and the impedance stress data in real time, and convert the respiratory stress data, the environmental stress data and the impedance stress data into digital signals for wireless transmission. The host computer and mobile terminal module are used to receive the digital signal, record the stress data corresponding to the digital signal, and generate a stress data table based on the recorded data. The stress data determination module is used to perform feature filtering and fusion processing based on the stress data table to determine the stress level of the fish, calculate the stress fatigue of the fish based on the stress level, and trigger an alarm and generate an environmental parameter adjustment command when the stress level of the fish reaches the high stress state threshold.

[0006] A non-destructive in-situ detection method for stress in individual fish and fish groups includes: Signals from the gills and dorsal sidewalls of fish are acquired using a flexible sensor chip and impedance hydrogel electrodes to obtain respiratory stress data, environmental stress data, and impedance stress data. The flexible sensor chip includes a flexible substrate, an attitude chip, a temperature and humidity sensor chip, a gas sensor chip, electrode circuitry, and an encapsulation layer. The impedance hydrogel electrode includes a patch substrate, a colloidal electrode, and conductive wires. The respiratory stress data, environmental stress data, and impedance stress data are collected in real time, and the respiratory stress data, environmental stress data, and impedance stress data are converted into digital signals and then transmitted wirelessly. Receive the digital signal, record the stress data corresponding to the digital signal, and generate a stress data table based on the recorded data; The stress data table is used to perform feature filtering and fusion processing to determine the stress level of the fish. The stress fatigue of the fish is calculated based on the stress level. When the stress level of the fish reaches the high stress state threshold, an alarm is triggered and an environmental parameter adjustment command is generated.

[0007] The present invention discloses the following technical effects: This invention provides a non-destructive in-situ detection system and method for individual and group stress in fish. Through the collaborative deployment and signal fusion of multimodal sensing modules (flexible sensing chip + impedance hydrogel electrode), this invention achieves simultaneous and accurate acquisition of fish respiration, impedance, and environmental stress data, overcoming the technical limitations of incomplete measurements by single sensors. Based on a dynamic weighted fusion algorithm and a fatigue accumulation model, a hierarchical quantitative evaluation system is constructed, breaking through the limitations of traditional subjective experience-based judgment. The adoption of an abnormal mode adaptive switching mechanism and a jumping sliding window sampling strategy significantly improves the system's robustness and real-time performance in complex transportation environments. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of a non-destructive in-situ detection system for individual and group stress in fish, provided by an embodiment of the present invention. Figure 2 This is a schematic diagram of the stress data determination module structure provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the flexible multimodal sensing module structure provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the multimodal fusion network construction process provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the integrated structure of a flexible multimodal sensing system provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the system transmission architecture provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the system edge-cloud collaborative processing mechanism provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the full-process intelligent control mechanism of the system provided in the embodiments of the present invention; Figure 9 This is a schematic diagram illustrating the principle of system failure testing during continuous monitoring, as provided in an embodiment of the present invention.

[0010] Figure label: 1-Multimodal sensing module, 2-Back-end circuit module, 3-Host computer and mobile terminal module, 4-Stress data determination module, 5-Data filtering sub-module, 6-Data fusion sub-module, 7-Fatigue calculation sub-module, 8-Feedback correction sub-module, 9-Early warning sub-module; 101-Attitude chip, 102-Gas sensor chip, 103-Temperature and humidity sensor chip, 104-Main controller, 105-Signal conditioning circuit, 106-Bluetooth communication sub-module, 107-Self-powered electronic module, 108-Impedance hydrogel electrode, 201-Terminal device. Detailed Implementation

[0011] 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.

[0012] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0013] like Figure 1 As shown, the present invention provides a non-destructive in-situ detection system for individual and group stress in fish, comprising: Multimodal sensing module 1, back-end circuit module 2, host computer and mobile terminal module 3, and stress data determination module 4; The multimodal sensing module 1 is disposed on the fish body and is used to collect signals from the gills and dorsal sidewall of the fish body using a flexible sensing chip and an impedance hydrogel electrode 108 to obtain respiratory stress data, environmental stress data, and impedance stress data. The flexible sensing chip includes: a flexible substrate, an attitude chip 101, a temperature and humidity sensing chip, a gas sensing chip 102, an electrode circuit, and an encapsulation layer. The impedance hydrogel electrode 108 includes: a patch substrate, a colloidal electrode, and conductive wires. The flexible substrate is made of biocompatible material, conforming to the fish's body curves and adapting to swimming deformations, providing stable support for each component without damaging the fish. The posture chip 101 monitors respiratory posture changes in real time, reflecting behavioral stress states. The temperature and humidity sensor chip collects temperature and humidity data to obtain environmental stress parameters. The gas sensor chip 102 detects the O2 / CO2 concentration inside the packaging and correlates it with gill activity to assess respiratory stress. The electrode circuit integrates multi-source analog signals and connects to the hydrogel electrode for unified transmission. The encapsulation layer seals the internal components with waterproof and wear-resistant material to ensure long-term stable operation in the aquatic environment. The impedance hydrogel electrode 108 is attached to the fish's back to establish low-impedance contact for collecting tissue electrical signals, effectively suppressing motion artifacts and ensuring continuous data acquisition. The back-end circuit module 2 is connected to the multimodal sensing module 1 and the host computer and mobile terminal module 3 respectively, and is used to collect the respiratory stress data, the environmental stress data and the impedance stress data in real time, and convert the respiratory stress data, the environmental stress data and the impedance stress data into digital signals for wireless transmission. The host computer and mobile terminal module 3 are used to receive the digital signal, record the stress data corresponding to the digital signal, and generate a stress data table based on the recorded data. The stress data determination module 4 is used to perform feature filtering and fusion processing based on the stress data table to determine the stress level of the fish, calculate the stress fatigue of the fish based on the stress level, and trigger an alarm and generate an environmental parameter adjustment command when the stress level of the fish reaches the high stress state threshold.

[0014] Furthermore, the back-end circuit module 2 includes: The main controller 104 is used to collect the respiratory stress data, the impedance stress data and the environmental stress data in real time, and record the corresponding detection time. The signal conditioning circuit 105 is connected to the main controller 104 and is used to receive analog signals of the respiratory stress data, the impedance stress data and the environmental stress data, and convert the analog signals into digital signals. Bluetooth communication submodule 106 is connected to the signal conditioning circuit 105 and is used to wirelessly transmit the digital signal to the host computer and mobile terminal module. The self-powered electronic module 107 is connected to the main controller 104 and the signal conditioning circuit 105 respectively, and is used to provide power to the back-end circuit module 2; the self-powered electronic module 107 includes a flexible rechargeable lithium battery and an energy harvesting circuit.

[0015] Furthermore, such as Figure 2 As shown, the stress data determination module 4 includes: Data filtering submodule 5 is used to perform feature filtering on the data in the stress data table to obtain stress data after feature filtering. The data fusion submodule 6 is connected to the data filtering submodule 5 and is used to fuse the stress data after feature filtering to determine the stress level of the fish. The fatigue calculation submodule 7 is connected to the data fusion submodule 6 and is used to calculate the stress fatigue of the fish based on the stress level of the fish. The feedback correction submodule 8, connected to the data fusion submodule 6, is used to generate environmental parameter adjustment instructions based on the feedback model when the fish stress level reaches the high stress state threshold. The environmental parameters include temperature parameters and oxygen supply parameters. The early warning submodule 9, connected to the feedback correction submodule 8, is used to trigger an alarm and generate a frequency adjustment command for adjusting the data acquisition frequency of the multimodal sensing module 1 when the stress level of the fish reaches the high stress state threshold.

[0016] Furthermore, the fabrication methods of the flexible sensing chip and the impedance hydrogel electrode 108 are as follows: Polydimethylsiloxane prepolymer and curing agent were mixed at a mass ratio of 10:1, vacuum degassing was performed for 30 minutes, and then injected into a 3D printing mold. The layer thickness was controlled to be 0.2 mm, and the mixture was cured at 80°C for 2 hours to obtain a flexible substrate. Using a pneumatic direct writing process, an electrode circuit pattern is etched onto the flexible substrate using a PEDOT:PSS solution doped with 6% dimethyl sulfoxide by mass, and the thickness of the electrode circuit pattern is controlled to be 0.1 mm. The attitude sensing chip, temperature and humidity sensing chip 103 and gas sensing chip 102 are fixed to the electrode circuit pattern and electrically connected using conductive silver paste, and then cured by heating at 250°C for 10 seconds using a hot air gun. A 0.5 mm thick medical-grade silicone is coated on the surface of the flexible substrate with integrated chip. After removing air bubbles using a vacuum degassing machine, the substrate is cured at room temperature for 24 hours to obtain the flexible sensing chip. Carboxymethyl cellulose and deionized water were mixed at a mass ratio of 1:10, and the pH was adjusted to 10. Then, 2% dopamine hydrochloride was added, and the mixture was mechanically stirred for 120 minutes to prepare a homogeneous CMC-PDA solution. The CMC-PDA solution was injected into a mold and dried at 60°C for 12 hours to form a chip substrate. Polystyrene sulfonate and the CMC-PDA solution obtained in the chip substrate preparation step are mixed at a volume ratio of 1:8, 5% phytic acid solution is added, the pH value is adjusted to 6.5, and a colloidal electrode with a layer thickness of 0.1 mm is printed using 3D printing technology. After 12 hours of crosslinking reaction, the electrode is cleaned and dried. The silver-plated copper wire is fixed to the surface of the colloidal electrode using conductive silver paste, and epoxy resin is coated at the solder joint for protection. The coating is cured for 30 minutes to form a conductive connection structure with mechanical stability. The colloidal electrode is bonded to the chip substrate, and its impedance characteristics are tested. The process parameters are adjusted until the measurement error of impedance amplitude and phase angle in the 1kHz frequency range is less than 5%, thus obtaining the impedance hydrogel electrode 108.

[0017] Furthermore, the data filtering submodule 5 includes: Feature extraction unit and filtering execution unit; The feature extraction unit is used to extract respiratory stress data, environmental stress data, and impedance stress data from the stress data table. The respiratory stress data includes respiratory rate stress data and respiratory intensity stress data. The environmental stress data includes temperature stress data, humidity stress data, oxygen stress data, and carbon dioxide stress data. The impedance stress data includes amplitude stress data and phase angle stress data. Where T represents ambient temperature stress data, RH represents ambient humidity stress data, O2 represents oxygen stress data, CO2 represents carbon dioxide stress data, T0, RH0, O20, and CO20 represent the ambient temperature, ambient humidity, oxygen concentration, and carbon dioxide concentration values ​​at the beginning of a certain moment, respectively; R represents the chip resistance value at the beginning of the next moment; and R0 represents the chip resistance value at the beginning of a certain moment. ST, SRH, SO2, and SCO2 represent the chip resistance sensitivity for temperature, humidity, oxygen, and carbon dioxide, respectively. The formula for extracting the respiratory stress data is: ; Where ΔT is one respiratory cycle of the fish, θmax is the maximum value of the respiratory angle within one respiratory cycle, and θmin is the minimum value of the respiratory angle within one respiratory cycle; The formula for extracting the environmental stress data is: ; Where T represents ambient temperature stress data, RH represents ambient humidity stress data, O2 represents oxygen stress data, CO2 represents carbon dioxide stress data, T0, RH0, O20, and CO20 represent the ambient temperature, ambient humidity, oxygen concentration, and carbon dioxide concentration values ​​at the beginning of a certain moment, respectively; R represents the chip resistance value at the beginning of the next moment; and R0 represents the chip resistance value at the beginning of a certain moment. ST, SRH, SO2, and SCO2 represent the chip resistance sensitivity for temperature, humidity, oxygen, and carbon dioxide, respectively. The formula for extracting the impedance stress data is: ; Where fk represents the continuous function sequence of the stress signal; R is the real part of the impedance stress data, X is the imaginary part of the impedance stress data, j is the imaginary part coefficient of the impedance stress data, Ae is the amplitude stress data, and Pe is the phase angle stress data; The screening execution unit is used to perform quality assessment on the extracted feature data, calculate the coefficient of variation of each feature data and the correlation coefficient with the preset stress standard value, and remove feature data with a coefficient of variation greater than 0.2 and a correlation coefficient less than 0.8. The application formula for the filtering execution unit is: ; in, CV Let be the coefficient of variation for each feature data, Std be the standard deviation for each feature data, and M be the mean for each feature data. |r| The correlation coefficient is the preset stress standard value. Specifically, the extraction method of the respiratory data extraction module is as follows: the respiratory rate is the reciprocal of one respiratory cycle, and the respiratory intensity is the difference between the maximum and minimum respiratory angles within one respiratory cycle. The impedance data extraction module uses the following method: The complex impedance Z is measured at 10 characteristic frequency points (10Hz-50kHz), and the real part is extracted using Discrete Fourier Transform (DFT). R ) and imaginary part ( i ), and extract amplitude features and phase.

[0018] Furthermore, such as Figure 3-4 As shown, the data fusion submodule 6 includes: The cross-fusion unit is used to perform temporal feature extraction and intermodal nonlinear fusion on the respiratory stress data and impedance stress data after feature screening. The cross-fusion unit includes four temporal fusion sub-units. The output of each temporal fusion sub-unit interacts with other temporal fusion sub-units through hidden states and cell states. The cross-fusion calculation formula is as follows: ; in, hj , HK , hl and cj , ck , cl Let be the hidden state and cell state of the other temporally fused units, respectively, and σ be the activation function; The convolutional processing unit, connected to the cross-fusion unit, is used to perform deep feature extraction on the cross-fused feature data. The convolutional processing unit is configured with two one-dimensional convolutional layers, and the convolution operation formula for each one-dimensional convolutional layer is: ; in, wk For convolution kernel weights, K The kernel size is [size]. xt+k-1 Input data; The fully connected unit is used to process the environmental stress data after feature filtering and aggregate it with the respiratory stress data and the impedance stress data. The fully connected unit includes two linear layers and an aggregation activation layer. The linear layers are used to extract features from the environmental stress data, and the aggregation activation layer is used to perform aggregation activation processing on the respiratory stress data, the impedance stress data, and the environmental stress data. The operation formula of the linear layer is: ; The operation formula for the aggregation activation layer is: ; The classification unit, connected to the convolutional processing unit and the fully connected unit, is used to output the fused features to the classifier for classification, outputting the fish stress level. The classification score is calculated using the following formula: ; in, zi For the first i The class output score, C This represents the total number of categories.

[0019] Furthermore, the fatigue calculation submodule 7 includes: Individual fatigue calculation unit, group fatigue calculation unit, and group state estimation unit; The individual fatigue calculation unit is used to calculate the individual fatigue value of fish, and the calculation formula is as follows: ; Where SF is the individual fatigue value, b is the interval length of the time period, SLtn is the individual stress level at the beginning of a certain moment, and SLtn+1 is the individual stress level at the beginning of the next moment. The group state estimation unit is used to estimate the state of the fish group based on a preset sampling ratio parameter; the sampling ratio parameter is the pairing ratio of the experimental group and the control group, and only the experimental group is equipped with the multimodal sensing module; the group state estimation unit estimates the group stress level by interpolating the data of the experimental group, and the interpolation estimation formula is: ; Among them, TSF is the group stress level, SF p This represents the individual fatigue value of the p-th fish, where p represents the number of fish in a certain transport batch; The group fatigue calculation unit is used to calculate the group fatigue of fish, and its calculation formula is as follows: ; Where TSL represents the group fatigue level, and SLp represents the individual stress level value of the p-th fish; Furthermore, the stress data determination module 4 also includes: The model building and training submodule is configured to build a stress classification multimodal fusion network based on data in a preset knowledge base module. The knowledge base module stores a stress prediction dataset and a stress reference dataset. The stress prediction dataset includes respiratory stress data, impedance stress data, and environmental stress data collected using the multimodal sensing module. The stress reference dataset includes blood glucose and cortisol parameters collected synchronously as standard parameters. When constructing the stress classification multimodal fusion network, the model building and training submodule executes the following data processing logic to generate a fish stress label set: Standardized stress values ​​are defined using the blood glucose and cortisol parameters as baseline signals. The Pearson correlation coefficient between the detected stress value and the standardized stress value is calculated, and this Pearson correlation coefficient is defined as a weighting coefficient for the respiratory stress data, the environmental stress data, and the impedance stress data. The calculation formula is as follows: ; The respiratory stress data, environmental stress data, and impedance stress data are discretized and categorized. Within the comfort zone, the stress level is set to 0, and beyond that threshold, it increases linearly based on the intensity of the characteristic response. The respiratory stress data corresponds to 3 levels, the environmental stress data to 5 levels, and the impedance stress data to 21 levels. The categorization calculation formula is as follows: ; Using the aforementioned weighting coefficients, a multi-factor arithmetic incremental step accumulation is performed on each of the discretized and graded data to obtain the fused stress level label. The calculation formula is as follows: ; The fused stress level labels were normalized and divided into 5 stress levels with an interval increment of 0.2. The classification formula is as follows: ; The model building and training submodule uses the stress prediction dataset as input and the fish stress label set as the output ground truth to train the stress classification multimodal fusion network, so that the stress data determination module 4 can output the stress detection results of the fish to be tested.

[0020] Specifically, the quantitative assessment method for fish stress is as follows: Several experimental samples were obtained by randomly grouping live fish purchased from the same batch, subjecting them to waterless low-temperature dormancy and live transport treatment; The multimodal sensing module 1 is used to collect data at fixed time intervals to obtain the original stress dataset, and the original stress dataset of the experiment is saved to the preset knowledge base module. For the data collected from each group within the original stress dataset, fish stress standard parameters are also collected to obtain a fish stress prediction dataset and a fish stress reference dataset. A fish stress label set is constructed based on the fish stress prediction dataset and stress reference dataset, and the fish stress label set is saved to a preset knowledge base module; A stress classification multimodal fusion network is constructed based on the fish stress prediction dataset and the fish stress label set; the fish stress hierarchical fusion network includes: a cross-fusion layer, a convolutional processing layer, a fully connected layer, and a classification layer; The multimodal sensing module 1 collects stress detection datasets at fixed time intervals; the stress detection datasets are input into a stress classification multimodal fusion network to detect stress in the fish and obtain stress detection results for the fish under test.

[0021] The fish stress standard parameters include a stress reference dataset; the fish stress standard parameters include blood glucose parameters and cortisol parameters; the stress prediction parameters include fish respiratory stress data, impedance stress data, and environmental stress data; Furthermore, the deployment location of the multimodal sensing module 1 within the fish and transportation environment is determined by a multimodal sensing module location determination method based on prior knowledge. This location determination method includes the following steps: A multimodal location optimization objective function was constructed; a prior model of fish physiological response was established, and historical data was used to fit the spatial distribution function of respiratory rate and body surface impedance, the formula of which is: ; The gradient field distribution of temperature and humidity in a transportation environment is calculated based on fluid dynamics simulation. The calculation formula is as follows: ; A location optimization objective function is constructed, and the optimal deployment location is solved using a particle swarm optimization algorithm. The location optimization objective function is as follows: ; Determine the location for respiratory monitoring using the flexible sensor chip; By capturing the movement trajectory of fish gill covers using high-speed photography, motion equations are established: ; The maximum angular displacement region is calculated, and the maximum angular displacement region is determined as the optimal attachment position of the flexible sensing chip. The calculation criterion is as follows: ; Determine the impedance monitoring location of the impedance hydrogel electrode 108; The deployment location was determined through conductivity field simulation; a three-dimensional finite element model of the fish was established, and the Poisson equation was solved. ; The region with the largest surface potential gradient and the minimum distance constraint between it and the respiratory monitoring location is selected as the deployment location of the impedance hydrogel electrode 108. The calculation criterion is as follows: ; Determine the placement of environmental sensors; The layout is determined using the Voronoi diagram segmentation method; a three-dimensional spatial segmentation is established based on the dimensions of the transport packaging, and the segmentation set is defined as follows: ; Sensors are deployed at the center of each Voronoi unit to ensure full coverage of environmental parameter acquisition, with the coverage criterion being: .

[0022] This invention also provides a method for quantitative assessment and decision-making regarding stress during fish survival and transportation: Each group of fish is numbered. During the live transport process, wearable sensors are paired with the fish, and the corresponding number is activated by operating the corresponding multimodal sensor module 1. The multimodal sensing module 1 is initialized to determine whether the system is working properly. If it is not working properly, the multimodal sensing module 1 is restarted for adjustment. If it still cannot work properly after adjustment, an alarm is triggered to remind manual inspection. The multimodal sensing module 1 collects stress detection parameters of fish, and collects data multiple times within the same time period and from multiple fish to ensure data stability and generalization. It compares the stress detection signals with the knowledge base to determine whether the collected signals are abnormal data. If the data is abnormal, an alarm is triggered and manual inspection is performed. If the data is normal, the multimodal fusion network and fatigue calculation submodule 7 are activated at the terminal to perform stress classification and fatigue calculation, and the results are transmitted to the terminal via Bluetooth module and further displayed through a visual interface. Simultaneous application of all scales across the three modalities allows for more accurate assessment of the current stress level and fatigue level of fish. In single-modal, dual-modal, or trimodal partial-scale modes, basic quality assessments of fish stress level and fatigue level can be made, suitable for working environments where all scales of the three modalities cannot be simultaneously met, and where rapid screening of fish stress level and fatigue level is required. Under the overall fusion modality, fish stress level can be accurately measured, and the current stress level and fatigue level of fish can be predicted, making it suitable for transportation research of high-end, expensive, and research-grade fish.

[0023] Furthermore, when data anomalies are detected (anomaly in one scale of a mode, anomalies in all scales of a mode, or anomalies in multiple scales of multiple modes), while some modes stop working, other modes can still function normally through the adaptive mode module and can be input into the model normally. The working modes include: respiratory mode, environmental mode, impedance mode, respiratory-environment fusion mode, environmental-impedance fusion mode, respiratory-impedance fusion mode, and total fusion mode. Among them, the respiratory mode includes respiratory frequency scale and respiratory intensity scale, the environmental mode includes temperature scale, humidity scale, oxygen scale, and carbon dioxide scale, and the impedance mode includes amplitude scale and phase scale. The amplitude scale includes 10 amplitude scales at 10 frequencies, and the phase scale includes 10 phase scales at 10 frequencies.

[0024] Determine if the three modes are affected by environmental interference. If none are affected, activate the overall fusion mode; if the respiratory mode is affected, activate the environmental mode and impedance mode; if the environmental mode is affected, activate the respiratory mode and impedance mode; if the impedance mode is affected, activate the environmental mode and respiratory mode. Respiratory disturbances specifically refer to: if the fish is not fully in a dormant state and its breathing is pulsating, then the respiratory mode will be disturbed; environmental disturbances specifically refer to: if the environment inside the transport packaging is too harsh or the fish bag is ruptured, then the environmental mode will be disturbed; impedance disturbances specifically refer to: when the fish secretes too much mucus, then the impedance mode will be disturbed. The formula for data anomaly detection is to detect a scale anomaly in a single modality: ; in, A m，s Representing modes m scale s The abnormal flag (1 indicates abnormal, 0 indicates normal). x m，s This is the current data value. μ m，s and σ m，s respectively scale s The mean and standard deviation, k This is the abnormal threshold factor.

[0025] Detecting multi-scale anomalies across multiple modalities: ; in, A total This indicates the total number of anomalies. M For the total number of modes, S m For modality m The scale number is used to determine whether the system has entered a state of complete anomalousness.

[0026] Furthermore, when an anomaly is detected, the operating mode is adaptively selected: in, W m Representing modes m The working status (1 indicates working, 0 indicates stopped) is such that when all scales of a mode are abnormal or the mode stops working, the mode is disabled, while other normal modes remain working.

[0027] After selecting a specific mode, input the data from the normal operating mode into the model: The fused data is used as input to the model. X m For the data of mode m, w m For modality m The weights (satisfying ∑w) m =1), W m Ensure that only the working mode contributes data.

[0028] For the overall fusion mode, the scale characteristics of all working modes are combined: in, F total As a feature of overall integration, F RS , F E N S , F ZS These are the eigenvectors for the respiratory mode, environmental mode, and impedance mode, respectively. F RS-E N S , F RS-ZS , F E N S-ZS Features of the corresponding fusion mode, CoNcat This indicates a feature splicing operation.

[0029] When a high stress state is detected, an automatic feedback mechanism is triggered to adjust the transport vehicle's temperature, oxygen supply, and deceleration. This feedback mechanism is based on a formula... Adjusting the temperature, among which e(t) For error, k p , ki , k d The control parameter is used to adjust the oxygen supply based on an optimization algorithm, using the following formula: ,in w For inertial weights, c 1 , c 2 For acceleration coefficient, r 1 , r 2 It is a random factor.

[0030] Furthermore, this embodiment provides a real-time classification method for fish stress, which achieves high timeliness and high accuracy stress monitoring by dynamically adjusting the sampling window and frequency. Specifically, it includes the following steps: Jumping sliding window design: Define the sliding window as overlapping time intervals (previous window: 0~). Δt Seconds; next window: Δt ~2 Δt (seconds), suitable for real-time classification scenarios.

[0031] The respiratory signal was extracted using an integral sliding window to extract the mean feature, the environmental signal was filtered using Gaussian filtering, and the impedance signal was normalized using the range.

[0032] ; ; ; Dynamic window frequency adjustment: Based on the current fish stress level ( SL Adjust window duration in real time Δt The calculation formula is: ; The window duration is dynamically adjusted based on signal stability. The CV value of each type of stress detection signal is determined through multiple experiments. When the signal fluctuates greatly (high CV value), the window is shortened to improve timeliness, and when the signal is stable, the window is extended to reduce noise.

[0033] Stress-response sampling: Construct a dynamic frequency adjustment formula for high stress levels ( Automatically increase the sampling frequency to ensure transient feature capture: ; f base The reference sampling frequency is denoted by , and is the reciprocal of the time window. The adjustment coefficient is between [0, 1] and was determined through multiple experiments. SL This represents the current stress level of the fish.

[0034] Real-time classification output: Input the multimodal data (respiration, impedance, environment) in the window into the stress classification network and output the real-time stress level (0~5).

[0035] This invention provides a fish survival transport stress monitoring and decision support system based on multimodal sensing, the specific implementation of which is as follows: Figure 5 The system's core sensing components were showcased. During the transportation of live fish, oxygenated packaging bags integrating a flexible sensing system were used, with a specially designed flexible chip precisely attached to the fish's gills to monitor key parameters such as respiratory rate and intensity. The flexible sensing chip consists of a flexible substrate (PDMS material, 10:1 ratio, cured at 80℃), an MPU-6050 attitude sensor, an SHT31 temperature and humidity sensor, an MQ-135 gas sensor, a conductive silver circuit, and a silicone encapsulation layer. During fabrication, a 0.2mm thick substrate is first formed in a 3D printing mold, then a 0.1mm thick PEDOT:PSS circuit is printed using pneumatic direct writing technology, and finally encapsulated and cured with medical-grade silicone, with the total thickness controlled within 0.8mm. Meanwhile, an array of 108 impedance hydrogel electrodes is positioned on the back of the fish to detect changes in impedance amplitude and phase angle. These multimodal sensors collect real-time data on respiration, impedance, and environmental stress (temperature, humidity, oxygen, carbon dioxide).

[0036] The impedance hydrogel electrode 108 consists of a CMC thin film substrate, a PEDOT:PSS colloidal electrode, and silver-plated copper wire. During fabrication: The substrate was prepared by mixing CMC with deionized water (1:10), adjusting the pH to 10, adding 2% dopamine hydrochloride, and drying at 60°C for 12 hours. PEDOT:PSS and CMC-PDA solution (1:8) were mixed, and 5% phytic acid was added to adjust the pH to 6.5. A 0.1 mm thick electrode was 3D printed and crosslinked for 12 hours. Fix the silver-plated copper wire with conductive silver paste, protect the solder joints with epoxy resin, and cure for 30 minutes. After assembly, testing was conducted to ensure that the impedance error was <5% at 1kHz.

[0037] Figure 6 The system's data transmission architecture is described. The acquired multimodal data is processed in real-time by the main controller 104, converted into digital signals by the signal conditioning circuit 105, and then transmitted to the terminal devices 201 and 201 (including...) by the Bluetooth Low Energy module. N(Book, PC, and dedicated APP). Users can view real-time monitoring data and early warning information on any terminal device 201 using a unique transport batch number, enabling remote intelligent management of the transport process.

[0038] This embodiment also provides a specific method for optimizing the position of the multimodal sensing module 1 based on prior knowledge, ensuring the accurate deployment of respiratory, impedance, and environmental sensors during the live transport of fish. The method specifically includes the following steps: By fitting the spatial distribution function of respiratory rate and body surface impedance using historical data, the coordinates of the gill center and the anisotropic diffusion coefficient were determined.

[0039] Based on fluid dynamics simulation calculations of temperature and humidity gradient field distribution in the transportation environment, a spatial variation model of environmental parameters is established.

[0040] The movement trajectory of the gill cover is obtained by high-speed imaging, and the motion equation related to the damping ratio ζ and the stress level is established. The region of maximum angular displacement is calculated as the optimal attachment position.

[0041] A three-dimensional finite element model of the fish was established, and the Poisson equation was solved to simulate the tissue conductivity distribution. Electrodes were deployed in the region with the maximum potential gradient on the body surface, while simultaneously satisfying the minimum distance constraint (∥p) to the respiratory sensing point. imp - p resp ∥ ≥d min ).

[0042] Based on the dimensions of the transport packaging, a three-dimensional spatial segmentation is performed using a Voronoi diagram. Environmental sensors are deployed at the center of each Voronoi unit to ensure comprehensive monitoring of temperature, humidity, oxygen, and carbon dioxide.

[0043] An objective function is constructed, and the optimal deployment location is determined by considering the deployment requirements of respiratory, impedance, and environmental sensors through an optimization algorithm. p Weighting coefficient α , β , γ satisfy α + β + γ =1.

[0044] Figure 7The system's edge-cloud collaborative processing mechanism was demonstrated. At the edge computing layer, a multimodal fusion network and a deep learning inference module were deployed, dynamically optimizing the data acquisition frequency using an adaptive sampling strategy. Data was uploaded to a distributed storage cluster in the cloud service layer via the network module, generating a 3D visualized stress heatmap based on a big data analytics platform, and then visualized via a mobile app and a web console. The system employs an innovative hybrid power supply scheme (flexible lithium battery + piezoelectric / thermal energy harvesting), achieving intelligent power management through a dynamic power allocation algorithm (P_allocate=α·P_battery+(1-α)·P_harvest).

[0045] Figure 8 The system's end-to-end intelligent control mechanism is presented. At the hardware control layer, the main controller 104 and signal conditioning circuit 105 complete signal acquisition and processing; the data processing layer ensures data reliability through anomaly tolerance mechanisms (including command reassembly and mode switching); the feedback control layer, based on error analysis of temperature / oxygen concentration regulation and blockchain data storage, combines CCP fluid simulation and 3D spatial segmentation to optimize sensor placement and dynamically adjust environmental parameters; finally, through edge A... i The calculation realizes multimodal data fusion and hierarchical correction, and the results are synchronized to the host computer (generating dynamic tables) and mobile terminals (displayed in real time), and trigger the calculation of group fatigue and threshold warning, forming a complete monitoring-analysis-control closed loop.

[0046] This system, through the collaborative operation of a flexible sensing chip and impedance hydrogel electrodes 108, achieves precise monitoring of multi-dimensional stress parameters in fish, including respiration, impedance, and environmental factors. Its innovative edge-cloud collaborative architecture and intelligent control mechanism ensure closed-loop management throughout the entire process, from data acquisition and intelligent analysis to decision-making and control, significantly improving the survival rate and quality assurance capabilities of live fish transportation.

[0047] like Figure 9 As shown, this embodiment of the invention provides a method for determining survival rate and economic benefits based on specific values ​​of individual stress and specific values ​​of group stress.

[0048] As an optional embodiment, taking sturgeon as an example, in a waterless cryogenic transport experiment of 100 sturgeon, individual stress data and group stress data were collected based on the multimodal sensing module 1.

[0049] First, calculate the average group stress level and average group stress fatigue: ; Based on the comparison between experimental group data and actual mortality rate, a nonlinear regression model was constructed to predict the survival rate Sr of the fish population per unit time. in: λ 1 , λ 2 These are the experimental fitting coefficients, obtained from multiple experiments; Next, considering factors such as survival rate during transportation, value per fish, and loss costs, the total economic benefit of transportation, E, is defined as follows: ; Where: V f i sh The selling price per unit of fish; c loss The economic loss caused by the death of a single fish; N : Total quantity transported; C tra N sport This is the cost of this transportation.

[0050] Furthermore, as an optional implementation, the system supports dynamic optimization of transport density by calculating the average stress level based on partial sturgeon sensor data, optimizing the density of the next transport batch, reducing costs and improving efficiency.

[0051] In the waterless cryogenic transport of sturgeon, multimodal sensing module 1 was used to monitor the gills and back of 20 sturgeon in real time. Based on data fusion submodule 6, a cross-fusion module was used to extract temporal features, a convolutional processing module (convolutional kernels 3 and 5) was used for deep feature extraction, and a fully connected module fused the multimodal signals to output the stress level. SL ). This consists of a subset of sensor samples (numbered as set ). i The quantity is N Stress values ​​are collected and the average stress level (AE) of the current transport batch is calculated. SL ): ; in, SLi For the first i Multimodal fusion stress value of a fish during transportation.

[0052] Empirical response function modeling of transport density and stress level: A regression model was established based on historical experimental data to describe the relationship between loading density D and average stress level AE. SL Experience relationship: Where a>0, the higher the density, the stronger the stress.

[0053] ; Define the economic benefit function for each vehicle transported per unit time: ; in: P s (D) =1-λ·AE SL (D): Based on AE SL Estimated predicted survival rate; V fish Price per unit of fish; C tra N sport Unit transportation cost.

[0054] Optimal density solution and adaptive update mechanism: Optimal transport density D The density corresponding to the maximum value of the objective function: The system can be based on the current batch of AEs SL The predictive density model automatically derives the target density to be set for the next transport. DN ext =D And it serves as an input parameter for the scheduling system.

[0055] Multiple feedback loop updates based on an adaptive mechanism: As transportation batches accumulate, the system can use a sliding window to perform real-time regression updates on a, b, and c, thereby continuously optimizing the density prediction accuracy. ; Where, θ=( (a, b, c) , N This represents the regression window length.

[0056] This embodiment also provides a method for quantitatively assessing individual and group stress during fish survival and transportation based on a flexible sensor chip, including: Signals from the gills and dorsal sidewalls of fish are acquired using a flexible sensor chip and impedance hydrogel electrodes to obtain respiratory stress data, environmental stress data, and impedance stress data. The flexible sensor chip includes a flexible substrate, an attitude chip, a temperature and humidity sensor chip, a gas sensor chip, electrode circuitry, and an encapsulation layer. The impedance hydrogel electrode includes a patch substrate, a colloidal electrode, and conductive wires. The respiratory stress data, environmental stress data, and impedance stress data are collected in real time, and the respiratory stress data, environmental stress data, and impedance stress data are converted into digital signals and then transmitted wirelessly. Receive the digital signal, record the stress data corresponding to the digital signal, and generate a stress data table based on the recorded data; The stress data table is used to perform feature filtering and fusion processing to determine the stress level of the fish. The stress fatigue of the fish is calculated based on the stress level. When the stress level of the fish reaches the high stress state threshold, an alarm is triggered and an environmental parameter adjustment command is generated.

[0057] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0058] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A non-destructive in-situ detection system for individual and group stress in fish, characterized in that, include: Multimodal sensing module, back-end circuit module, host computer and mobile terminal module, and stress data determination module; The multimodal sensing module is disposed on the fish body and is used to collect signals from the gills and dorsal sidewall of the fish body using a flexible sensing chip and an impedance hydrogel electrode to obtain respiratory stress data, environmental stress data, and impedance stress data. The flexible sensing chip includes: a flexible substrate, an attitude chip, a temperature and humidity sensing chip, a gas sensing chip, an electrode circuit, and an encapsulation layer. The impedance hydrogel electrode includes: a patch substrate, a colloidal electrode, and conductive wires. The back-end circuit module is connected to the multimodal sensing module and the host computer and mobile terminal module respectively, and is used to collect the respiratory stress data, the environmental stress data and the impedance stress data in real time, and convert the respiratory stress data, the environmental stress data and the impedance stress data into digital signals for wireless transmission. The host computer and mobile terminal module are used to receive the digital signal, record the stress data corresponding to the digital signal, and generate a stress data table based on the recorded data. The stress data determination module is used to perform feature filtering and fusion processing based on the stress data table to determine the stress level of the fish, calculate the stress fatigue of the fish based on the stress level, and trigger an alarm and generate an environmental parameter adjustment command when the stress level of the fish reaches the high stress state threshold.

2. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The back-end circuit module includes: The main controller is used to collect the respiratory stress data, the impedance stress data, and the environmental stress data in real time, and record the corresponding detection time. A signal conditioning circuit, connected to the main controller, is used to receive analog signals of the respiratory stress data, the impedance stress data, and the environmental stress data, and convert the analog signals into digital signals. A Bluetooth communication submodule, connected to the signal conditioning circuit, is used to wirelessly transmit the digital signal to the host computer and mobile terminal module; The self-powered electronic module is connected to the main controller and the signal conditioning circuit respectively, and is used to provide power to the back-end circuit module; the self-powered electronic module includes a flexible rechargeable lithium battery and an energy harvesting circuit.

3. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The stress data determination module includes: The data filtering submodule is used to perform feature filtering on the data in the stress data table to obtain stress data after feature filtering. The data fusion submodule, connected to the data filtering submodule, is used to fuse the stress data after feature filtering to determine the stress level of the fish. The fatigue calculation submodule, connected to the data fusion submodule, is used to calculate the stress fatigue of the fish based on the stress level of the fish. The feedback correction submodule, connected to the data fusion submodule, is used to generate environmental parameter adjustment instructions based on the feedback model when the fish stress level reaches the high stress state threshold. The environmental parameters include temperature parameters and oxygen supply parameters. The early warning submodule, connected to the feedback correction submodule, is used to trigger an alarm and generate a frequency adjustment command for adjusting the data acquisition frequency of the multimodal sensing module when the stress level of the fish reaches the high stress state threshold.

4. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The fabrication methods for the flexible sensing chip and the impedance hydrogel electrode are as follows: Polydimethylsiloxane prepolymer and curing agent were mixed at a mass ratio of 10:1, vacuum degassing was performed for 30 minutes, and then injected into a 3D printing mold. The layer thickness was controlled to be 0.2 mm, and the mixture was cured at 80°C for 2 hours to obtain a flexible substrate. Using a pneumatic direct writing process, an electrode circuit pattern is etched onto the flexible substrate using a PEDOT:PSS solution doped with 6% dimethyl sulfoxide by mass, and the thickness of the electrode circuit pattern is controlled to be 0.1 mm. The attitude sensing chip, temperature and humidity sensing chip and gas sensing chip are fixed to the electrode circuit pattern and electrically connected by conductive silver paste, and then cured by heating at 250°C for 10 seconds using a hot air gun. A 0.5 mm thick medical-grade silicone is coated on the surface of the flexible substrate with integrated chip. After removing air bubbles using a vacuum degassing machine, the substrate is cured at room temperature for 24 hours to obtain the flexible sensing chip. Carboxymethyl cellulose and deionized water were mixed at a mass ratio of 1:10, and the pH was adjusted to 10. Then, 2% dopamine hydrochloride was added, and the mixture was mechanically stirred for 120 minutes to prepare a homogeneous CMC-PDA solution. The CMC-PDA solution was injected into a mold and dried at 60°C for 12 hours to form a patch substrate. Polystyrene sulfonate and the CMC-PDA solution obtained in the patch substrate preparation step are mixed at a volume ratio of 1:8, 5% phytic acid solution is added, the pH value is adjusted to 6.5, and a colloidal electrode with a layer thickness of 0.1 mm is printed using 3D printing technology. After 12 hours of crosslinking reaction, the electrode is cleaned and dried. The silver-plated copper wire is fixed to the surface of the colloidal electrode using conductive silver paste, and epoxy resin is coated at the solder joint for protection. The coating is cured for 30 minutes to form a conductive connection structure with mechanical stability. The colloidal electrode is bonded to the patch substrate, and its impedance characteristics are tested. The process parameters are adjusted until the measurement error of impedance amplitude and phase angle in the 1kHz frequency range is less than 5%, thus obtaining the impedance hydrogel electrode.

5. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The data filtering submodule includes: Feature extraction unit and filtering execution unit; The feature extraction unit is used to extract respiratory stress data, environmental stress data, and impedance stress data from the stress data table. The respiratory stress data includes respiratory rate stress data and respiratory intensity stress data. The environmental stress data includes temperature stress data, humidity stress data, oxygen stress data, and carbon dioxide stress data. The impedance stress data includes amplitude stress data and phase angle stress data. ； Where RS is the respiratory stress dataset, RSf For respiratory rate data, RSi The data represents respiratory intensity; ENS represents the environmental stress dataset, with ENST, ENSRH, ENSO2, and ENSCO2 representing environmental temperature, humidity, oxygen, and carbon dioxide stress data, respectively; ZS represents the impedance stress dataset, with ZSAe and ZSPe representing the amplitude stress dataset and phase angle stress data, respectively. The formula for extracting the respiratory stress data is: ； Where ΔT is one respiratory cycle of the fish, θmax is the maximum value of the respiratory angle within one respiratory cycle, and θmin is the minimum value of the respiratory angle within one respiratory cycle; The formula for extracting the environmental stress data is: ； Where T represents ambient temperature stress data, RH represents ambient humidity stress data, O2 represents oxygen stress data, CO2 represents carbon dioxide stress data, T0, RH0, O20, and CO20 represent the ambient temperature, ambient humidity, oxygen concentration, and carbon dioxide concentration values ​​at the beginning of a certain moment, respectively; R represents the chip resistance value at the beginning of the next moment; and R0 represents the chip resistance value at the beginning of a certain moment. ST, SRH, SO2, and SCO2 represent the chip resistance sensitivity for temperature, humidity, oxygen, and carbon dioxide, respectively. The formula for extracting the impedance stress data is: ； Where fk represents the continuous function sequence of the stress signal; R is the real part of the impedance stress data, X is the imaginary part of the impedance stress data, j is the imaginary part coefficient of the impedance stress data, Ae is the amplitude stress data, and Pe is the phase angle stress data; The screening execution unit is used to perform quality assessment on the extracted feature data, calculate the coefficient of variation of each feature data and the correlation coefficient with the preset stress standard value, and remove feature data with a coefficient of variation greater than 0.2 and a correlation coefficient less than 0.

8. The application formula for the filtering execution unit is: ； in, CV Let be the coefficient of variation for each feature data, Std be the standard deviation for each feature data, and M be the mean for each feature data. |r| The correlation coefficient is the preset stress standard value.

6. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The data fusion submodule includes: A cross-fusion unit is used to perform temporal feature extraction and intermodal nonlinear fusion on the respiratory stress data and impedance stress data after feature screening. The cross-fusion unit includes four temporal fusion sub-units. The output of each temporal fusion sub-unit interacts with other temporal fusion sub-units through hidden states and cell states. The cross-fusion calculation formula is as follows: ; in, hj , HK , hl and cj , ck , cl Let be the hidden state and cell state of the other temporally fused units, respectively, and σ be the activation function; The processing unit, connected to the cross-fusion unit, is used to perform deep feature extraction on the cross-fused feature data; the convolution processing unit is configured with two one-dimensional convolutional layers, and the convolution operation formula for each one-dimensional convolutional layer is: ; in, wk For convolution kernel weights, K The kernel size is [size]. xt+k-1 Input data; A fully connected unit is used to process the environmental stress data after feature filtering and aggregate it with the respiratory stress data and the impedance stress data. The fully connected unit includes two linear layers and an aggregation activation layer. The linear layers are used to extract features from the environmental stress data, and the aggregation activation layer is used to perform aggregation activation processing on the respiratory stress data, the impedance stress data, and the environmental stress data. The operation formula of the linear layer is: ; The operation formula for the aggregation activation layer is: ; The classification unit, connected to the convolutional processing unit and the fully connected unit, is used to output the fused features to the classifier for classification, outputting the fish stress level. The classification score is calculated using the following formula: ; in, zi For the first i The class output score, C This represents the total number of categories.

7. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The fatigue calculation submodule includes: Individual fatigue calculation unit, group fatigue calculation unit, and group state estimation unit; The individual fatigue calculation unit is used to calculate the individual fatigue value of fish, and the calculation formula is as follows: ； Where SF is the individual fatigue value, b is the interval length of the time period, SLtn is the individual stress level at the beginning of a certain moment, and SLtn+1 is the individual stress level at the beginning of the next moment. The group state estimation unit is used to estimate the state of the fish group based on a preset sampling ratio parameter; the sampling ratio parameter is the pairing ratio of the experimental group and the control group, and only the experimental group is equipped with the multimodal sensing module; the group state estimation unit estimates the group stress level by interpolating the data of the experimental group, and the interpolation estimation formula is: ； Among them, TSF is the group stress level, SF p This represents the individual fatigue value of the p-th fish, where p represents the number of fish in a certain transport batch; The group fatigue calculation unit is used to calculate the group fatigue of fish, and its calculation formula is as follows: ； Where TSL represents the group fatigue level, and SLp represents the individual stress level of the p-th fish.

8. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The stress data determination module also includes: The model building and training submodule is configured to build a stress classification multimodal fusion network based on data in a preset knowledge base module. The knowledge base module stores a stress prediction dataset and a stress reference dataset. The stress prediction dataset includes respiratory stress data, impedance stress data, and environmental stress data collected using the multimodal sensing module. The stress reference dataset includes blood glucose and cortisol parameters collected synchronously as standard parameters. When constructing the stress classification multimodal fusion network, the model building and training submodule executes the following data processing logic to generate a fish stress label set: Standardized stress values ​​are defined using the blood glucose and cortisol parameters as baseline signals. The Pearson correlation coefficient between the detected stress value and the standardized stress value is calculated, and this Pearson correlation coefficient is defined as a weighting coefficient for the respiratory stress data, the environmental stress data, and the impedance stress data. The calculation formula is as follows: ； The respiratory stress data, environmental stress data, and impedance stress data are discretized and categorized. Within the comfort zone, the stress level is set to 0, and beyond that threshold, it increases linearly based on the intensity of the characteristic response. The respiratory stress data corresponds to 3 levels, the environmental stress data to 5 levels, and the impedance stress data to 21 levels. The categorization calculation formula is as follows: ; Using the aforementioned weighting coefficients, a multi-factor arithmetic incremental step accumulation is performed on each of the discretized and graded data to obtain the fused stress level label. The calculation formula is as follows: ; The fused stress level labels are normalized and divided into 5 stress levels with an interval increment of 0.

2. The classification formula is as follows: ； The model building and training submodule uses the stress prediction dataset as input and the fish stress label set as the output ground truth to train the stress classification multimodal fusion network, so that the stress data determination module can output the stress detection results of the fish to be tested.

9. The non-destructive in-situ detection system for individual and group stress in fish according to claim 1, characterized in that, The deployment location of the multimodal sensing module in the fish and transportation environment is determined by a multimodal sensing module location determination method based on prior knowledge. The location determination method includes the following steps: A multimodal location optimization objective function was constructed; a prior model of fish physiological response was established, and historical data was used to fit the spatial distribution function of respiratory rate and body surface impedance, the formula of which is: ； The gradient field distribution of temperature and humidity in a transportation environment is calculated based on fluid dynamics simulation. The calculation formula is as follows: ； A location optimization objective function is constructed, and the optimal deployment location is solved using a particle swarm optimization algorithm. The location optimization objective function is as follows: ； Determine the location for respiratory monitoring using the flexible sensor chip; By capturing the movement trajectory of fish gill covers using high-speed photography, motion equations are established: ； The maximum angular displacement region is calculated, and the maximum angular displacement region is determined as the optimal attachment position of the flexible sensing chip. The calculation criterion is as follows: ； Determine the impedance monitoring location of the impedance hydrogel electrode; The deployment location was determined through conductivity field simulation; a three-dimensional finite element model of the fish was established, and the Poisson equation was solved. ; The region with the largest surface potential gradient and the minimum distance constraint between it and the respiratory monitoring location is selected as the deployment location of the impedance hydrogel electrode. The calculation criterion is as follows: ； Determine the placement of environmental sensors; The layout is determined using the Voronoi diagram segmentation method; a three-dimensional spatial segmentation is established based on the dimensions of the transport packaging, and the segmentation set is defined as follows: ； Sensors are deployed at the center of each Voronoi unit to ensure full coverage of environmental parameter acquisition, with the coverage criterion being: 。 10. A non-destructive in-situ detection method for stress in individual fish and fish groups, characterized in that, include: Signals from the gills and dorsal sidewalls of fish are acquired using a flexible sensor chip and an impedance hydrogel electrode to obtain respiratory stress data, environmental stress data, and impedance stress data. The flexible sensor chip includes a flexible substrate, an attitude chip, a temperature and humidity sensor chip, a gas sensor chip, an electrode circuit, and an encapsulation layer. The impedance hydrogel electrode includes a patch substrate, a colloidal electrode, and conductive wires. The respiratory stress data, environmental stress data, and impedance stress data are collected in real time, and the respiratory stress data, environmental stress data, and impedance stress data are converted into digital signals and then transmitted wirelessly. Receive the digital signal, record the stress data corresponding to the digital signal, and generate a stress data table based on the recorded data; The stress data table is used to perform feature filtering and fusion processing to determine the stress level of the fish. The stress fatigue of the fish is calculated based on the stress level. When the stress level of the fish reaches the high stress state threshold, an alarm is triggered and an environmental parameter adjustment command is generated.

Images