Appearance characteristic-based hippocampus intelligent grading and pond-dividing culture method and system

Abstract

The invention relates to the technical field of intelligent aquaculture and automatic processing, in particular to an appearance characteristic-based hippocampus intelligent grading and pond-dividing culture method and system, and the method comprises the following steps: S1, collecting sparse visual data and multi-source physiological data of hippocampus; s2, generating a three-dimensional digital twinborn model based on the sparse visual data; calculating a growth state grading index based on the three-dimensional digital twinborn model; s3, calculating a physiological stress index based on the multi-source physiological data; s4, comparing the physiological stress index calculated in real time with a preset safety stress threshold value and a critical stress threshold value, and generating a hierarchical adaptive correction strategy; s5, according to a grading self-adaptive correction strategy, controlling a flexible mechanical arm to execute fishing and pond separation operation; according to the method, the appearance and physiological data of the hippocampus are synchronously collected in a non-contact mode, the technical problems that the real condition of the hippocampus cannot be evaluated and damage is possibly caused due to target stress response caused by traditional physical intervention are solved, and the authenticity and effectiveness of original data are guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent aquaculture and automatic processing technology, in particular to a sea horse intelligent grading and pool breeding method and system based on appearance characteristics. BACKGROUND

[0002] In the current high-value live aquatic product, such as sea horse breeding industry, the growth state grading such as size pool is the core link to determine the breeding efficiency and survival rate; In order to ensure the accuracy of grading and subsequent packaging operation, the appearance characteristics of live sea horse need to be evaluated carefully; Traditional fishing and grading process depends on direct physical intervention, which can easily cause sea horse to produce stress reaction, which not only may cause sea horse to be injured, but also may change the key appearance characteristics such as body color and posture, so that the real grade in natural state cannot be accurately evaluated, which seriously affects the quality of automatic pool operation and subsequent breeding benefit.

[0003] In the prior art, fishing and pool operation is usually one-way physical operation, which lacks real-time perception and feedback of the physiological state of the target organism. On the one hand, in order to accurately evaluate the appearance, high-precision features need to be obtained at close range and from multiple angles. On the other hand, any close observation or physical contact will inevitably disturb the sea horse, making it enter a stress state, and thus leading to distortion of the evaluation data.

[0004] The prior art fails to effectively solve the conflict between high-precision feature recognition and maintaining the natural state of the target, resulting in insufficient objectivity and consistency of the grading results, and there is a risk of damaging the target in the fishing process; Therefore, how to provide a technical scheme for non-destructive, accurate and intelligent grading and pool breeding of live sea horses, complete high-fidelity appearance evaluation without physical intervention, and guide adaptive flexible fishing and pool operation by monitoring the physiological state in real time, is a technical problem that needs to be solved by those skilled in the art. SUMMARY

[0005] To solve the above technical problems, the present application discloses a sea horse intelligent grading and pool breeding method and system based on appearance characteristics, in particular, the technical scheme of the present application is:

[0006] A sea horse intelligent grading and pool breeding method based on appearance characteristics, comprising the following steps:

[0007] S1, collecting sparse visual data and multi-source physiological data of sea horse;

[0008] S2, generating a three-dimensional digital twin model based on the sparse visual data; calculating the growth state grading index based on the three-dimensional digital twin model;

[0009] S3, calculating a physiological stress index based on the multi-source physiological data;

[0010] S4, comparing the real-time calculated physiological stress index with preset safe stress threshold and critical stress threshold, and generating a hierarchical adaptive correction strategy;

[0011] S5, controlling the flexible manipulator to perform fishing and pool sorting operation according to the hierarchical adaptive correction strategy.

[0012] Preferably, S1 specifically includes:

[0013] A low-resolution two-dimensional image of the target is collected from multiple preset azimuth angles by a long-distance visual sensor to construct a sparse visual data set.

[0014] Multi-source physiological data is collected by a non-contact visual sensor; wherein the multi-source physiological data includes gill movement frequency, limb movement frequency and amplitude, and instantaneous body color change rate.

[0015] Preferably, S2 specifically includes:

[0016] The sparse visual data set is input into a pre-trained deep generation model to generate a three-dimensional digital twin model;

[0017] Growth characteristic parameter set is extracted from the three-dimensional digital twin model; wherein the growth characteristic parameter set includes body length, body height, body weight estimation value and body shape ratio.

[0018] The growth characteristic parameter set is input into a preset hierarchical model for weighted calculation to generate a growth state classification index.

[0019] Preferably, S3 specifically includes:

[0020] The multi-source physiological data is normalized;

[0021] The normalized multi-source physiological data is weighted calculated according to the preset sensitivity weight coefficient to generate a physiological stress index.

[0022] Preferably, S4 specifically includes:

[0023] When the physiological stress index is not higher than the safe stress threshold, the stress state level is determined as the safe zone, and an execution strategy for maintaining the current optimal path is generated;

[0024] When the physiological stress index is higher than the safe stress threshold and not higher than the critical stress threshold, the stress state level is determined as the warning zone, and a correction strategy for dynamically adjusting and reducing the movement speed of the manipulator is generated;

[0025] When the physiological stress index is higher than the critical stress threshold, the stress state level is determined as the danger zone, and an evasive strategy for aborting the task and retreating to a safe distance is generated.

[0026] Preferably, the modification strategy of the early warning zone is specifically:

[0027] The physiological stress index, the safe stress threshold and the critical stress threshold are input into a preset speed modification function to calculate a modified mechanical arm movement speed; wherein the modified mechanical arm movement speed is smoothly and nonlinearly attenuated with the increase of the physiological stress index.

[0028] Preferably, after the avoidance strategy of the danger zone is executed, it further includes:

[0029] The physiological stress index is continuously monitored, and when the physiological stress index falls below the safe stress threshold, the fishing and pool operation is resumed.

[0030] A hippocampus intelligent grading and pool breeding system based on appearance features, comprising:

[0031] A data acquisition module for acquiring sparse visual data and multi-source physiological data of the hippocampus;

[0032] A modeling and evaluation module for generating a three-dimensional digital twin model based on the sparse visual data, and calculating a growth state grading index based on the three-dimensional digital twin model;

[0033] A state monitoring module for calculating a physiological stress index based on the multi-source physiological data;

[0034] A decision control module for comparing the physiological stress index with a preset safe stress threshold and a critical stress threshold to generate a grading adaptive modification strategy;

[0035] An operation execution module for receiving the grading adaptive modification strategy and controlling a flexible mechanical arm to perform fishing and pool operation.

[0036] Compared with the prior art, the present application has the following beneficial effects:

[0037] 1. The present application synchronously acquires the appearance and physiological data of the hippocampus in a non-contact manner, solving the technical problem that traditional physical intervention causes target stress reaction, which further cannot evaluate the true appearance and may cause damage, ensuring the authenticity and effectiveness of the original data;

[0038] 2. The present application generates a high-fidelity three-dimensional digital twin model based on sparse visual data, and decomposes abstract growth concepts into calculable objective indexes, changing the evaluation of the growth state of the hippocampus from relying on subjective experience to an automatic, standardized and quantitative process, significantly improving the objectivity, consistency and fineness of the grading results;

[0039] 3、The application constructs a closed-loop adaptive control system dominated by biological signals, dynamically corrects the fishing strategy by real-time monitoring of the physiological stress index, changes the traditional one-way fishing behavior into a two-way adaptive interaction, maximally reduces the physical intervention intensity, and realizes safe and non-destructive acquisition of the living target;

[0040] 4、The application deeply integrates the biological characteristics and physiological state of hippocampus, effectively solves the technical contradiction between high-precision feature recognition and maintaining the natural state of the target through non-intervention modeling evaluation and adaptive fishing dominated by biological signals, and greatly improves the automated pooling efficiency and breeding benefit of precious living aquatic products. BRIEF DESCRIPTION OF DRAWINGS

[0041] The application will be further explained below in combination with the drawings and embodiments:

[0042] Figure 1 is a flowchart of the method of the application;

[0043] Figure 2 is a flowchart of the system of the application. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical scheme and advantages of the application clearer, the technical scheme of the application will be described in detail below. Obviously, the described embodiments are only some of the embodiments of the application, not all the embodiments; based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0045] Embodiment 1:

[0046] Please refer to Figure 1 A hippocampus intelligent grading and pool breeding method based on appearance characteristics, comprising the following steps:

[0047] S1, sparse visual data and multi-source physiological data of hippocampus are collected;

[0048] S2, a three-dimensional digital twin model is generated based on the sparse visual data; a growth state grading index is calculated based on the three-dimensional digital twin model;

[0049] S3, a physiological stress index is calculated based on the multi-source physiological data;

[0050] S4, the real-time calculated physiological stress index is compared with the preset safe stress threshold and critical stress threshold to generate a grading adaptive correction strategy;

[0051] S5, according to the grading adaptive correction strategy, a flexible mechanical arm is controlled to perform fishing and pool operation.

[0052] The embodiment provides a hippocampus intelligent grading and pool culture method based on appearance characteristics; the method aims to solve the technical problems that in the grading and fishing process of live hippocampus, stress reaction is caused due to physical intervention, so that the real appearance of the hippocampus cannot be accurately evaluated and damage may be caused; the method realizes nondestructive, accurate and pool culture of the hippocampus by constructing a complete technical process from non-intervention data acquisition, high-fidelity modeling evaluation to biological signal dominant closed-loop adaptive control;

[0053] The specific process of the method comprises the following steps:

[0054] S1, sparse visual data and multi-source physiological data of hippocampus are collected: in order to obtain the appearance and physiological information of hippocampus without disturbing the natural state of hippocampus, non-contact data acquisition is performed; the sparse visual data refers to a data set composed of target low-resolution two-dimensional images quickly captured from multiple preset azimuth angles by a long-distance visual sensor under the natural light condition of the culture environment, and the purpose is to provide basic input for subsequent three-dimensional modeling; the multi-source physiological data refers to dynamic data of multiple dimensions reflecting the physiological stress state of hippocampus, and the purpose is to quantify the tension degree of hippocampus to provide decision basis for subsequent adaptive fishing;

[0055] S2, a three-dimensional digital twin model is generated based on the sparse visual data; a growth state grading index is calculated based on the three-dimensional digital twin model: the core of this step is to accurately quantify and evaluate the growth state of hippocampus; the specific implementation path is that a pre-trained deep generation model is used to process the sparse visual data set collected in step S1, and a high-precision, full-view three-dimensional digital twin model of hippocampus under zero stress state is rendered in real time; the model is a digital reproduction of the real three-dimensional shape of hippocampus, and provides a high-fidelity data object for subsequent morphological analysis; the system automatically extracts and quantifies a series of preset growth characteristic parameters based on the model, and calculates the final growth state grading index through a nonlinear evaluation model; the score is a single, quantifiable comprehensive index, which is used to objectively evaluate the growth stage and size of hippocampus;

[0056] S3, a physiological stress index is calculated based on the multi-source physiological data: in order to realize real-time sensing of the stress state of hippocampus, the multi-source physiological parameters collected in S1 are input into a comprehensive evaluation model to dynamically calculate a physiological stress index reflecting the current tension degree of the target; the physiological stress index is a quantitative index, which converts the invisible internal stress level of hippocampus into a measurable dynamic signal, which is used as the core feedback variable of the subsequent closed-loop control system;

[0057] S4, compare the real-time calculated physiological stress index with the preset safe stress threshold and critical stress threshold to generate a hierarchical adaptive correction strategy: this step is the decision-making center connecting perception and action; the system compares the real-time physiological stress index calculated in step S3 with the two preset thresholds continuously to determine the stress state grade of the hippocampus; the safe stress threshold refers to the index value corresponding to the inflection point at which the slope of the hippocampus stress response curve first increases significantly, representing the critical point at which the hippocampus begins to produce initial nervous emotions; the critical stress threshold refers to the index value corresponding to the time when the hippocampus begins to exhibit behaviors such as violent escape or rigidity, representing the time when the hippocampus is in a highly nervous state; according to the comparison result, the system generates a corresponding hierarchical adaptive correction strategy, which aims to dynamically adjust the execution mode of the fishing and sorting operation according to the real-time feedback of the hippocampus;

[0058] S5, control the flexible manipulator to perform the fishing and sorting operation according to the hierarchical adaptive correction strategy: as the final execution link of the technical process, this step converts the strategy generated in step S4 into specific physical actions; the flexible manipulator is a kind of soft and bionic actuator used to complete fishing without damaging the hippocampus; it can accurately control its motion trajectory and speed according to the received strategy, such as normal execution, slow speed, or stop and retreat, to realize adaptive interaction with the hippocampus;

[0059] The present application constructs an intelligent decision-making and control closed loop that deeply integrates the biological characteristics and physiological state of the hippocampus through the above steps; it realizes objective and accurate grading of the growth state of the hippocampus through non-intervention modeling and evaluation; in the fishing process, it converts the traditional one-way fishing behavior into a two-way adaptive interaction guided by biological signals by real-time monitoring of physiological stress signals, solves the technical contradiction between high-precision feature recognition and maintaining the natural state of the target, realizes non-destructive grading and efficient acquisition of the living target, and improves the quality of automated processing of valuable aquatic products and the breeding efficiency;

[0060] Embodiment 2:

[0061] S1 specifically includes:

[0062] Acquire low-resolution two-dimensional images of the target from multiple preset azimuth angles through a long-distance visual sensor to construct a sparse visual data set;

[0063] Acquire multi-source physiological data through a non-contact visual sensor; the multi-source physiological data includes gill movement frequency, limb movement frequency and amplitude, and instantaneous body color change rate.

[0064] This embodiment is based on embodiment 1 and aims at the data acquisition method of step S1.

[0065] In a preferred embodiment, the sparse visual data is collected by deploying a long-range visual sensor in the breeding environment; the sensor is configured to take rapid snapshots of the target from multiple preset azimuth angles to obtain low-resolution two-dimensional images; the long-range design aims to minimize the visual interference of the sensor on the sea horse; multiple azimuth angles ensure that the collected data is sparse but still covers key morphological information, providing sufficient geometric constraints for three-dimensional reconstruction; and low resolution ensures collection speed, enabling instantaneous snapshots and avoiding image blurring caused by sea horse movement; these images are collectively integrated to construct a sparse visual data set;

[0066] At the same time, the multi-source physiological data is collected by a non-contact visual sensor, which continuously monitors the sea horse through video frames during the entire process of approaching and capturing by the mechanical arm, and by analyzing the subtle changes in the video frames, the system can extract core parameters reflecting the physiological stress state of the sea horse; in this embodiment, the multi-source physiological data specifically includes: gill movement frequency, limb movement frequency and amplitude, and instantaneous body color change rate caused by tension; wherein the gill movement frequency is obtained by analyzing the small color periodic changes on the surface; the limb movement frequency and amplitude are calculated by optical flow method or target tracking algorithm; and the instantaneous body color change rate is obtained by analyzing the color value change rate of a specific area on the surface;

[0067] Through the above-mentioned specific collection method, the present application can simultaneously and in parallel acquire two kinds of heterogeneous data for appearance modeling and physiological state monitoring without any physical contact and stress interference; this dual-data flow non-invasive collection scheme provides high-quality, distortion-free raw data input for subsequent high-fidelity growth state evaluation and high-sensitivity stress closed-loop control, ensuring the effectiveness and reliability of the entire technical solution.

[0068] To cope with challenges such as water turbidity, light changes and occlusions in actual breeding environments, the data collection module of the system can further integrate image enhancement and denoising algorithms, such as using the dark channel prior algorithm for underwater image restoration; at the same time, by fusing data collected by multiple sensors at different positions and using target tracking algorithms to take continuous snapshots of the same target within a short time, data of poor quality caused by occlusion or poor posture can be effectively removed, ensuring the quality of data input to the subsequent model, thereby enhancing the robustness of the entire system in complex real-world scenarios.

[0069] Embodiment 3:

[0070] S2 specifically includes:

[0071] The sparse visual data set is input into a pre-trained deep generative model to generate a three-dimensional digital twin model;

[0072] extract a growth characteristic parameter set from the three-dimensional digital twin model; wherein the growth characteristic parameter set includes body length, body height, body weight estimated value, and body proportion;

[0073] input the growth characteristic parameter set into a preset grading model, perform weighted calculation, and generate a growth state grading index.

[0074] This embodiment is based on embodiment 1, and is directed to the growth state evaluation process of step S2;

[0075] For further illustration, the generation process of the three-dimensional digital twin model is as follows: input the sparse visual data set collected in the foregoing step into a pre-trained deep generation model, thereby generating a high-fidelity three-dimensional digital twin model; this process can be described by the following formula:

[0076]

[0077] wherein, : represents the final generated high-fidelity three-dimensional digital twin model, which is used as a data object for subsequent growth state evaluation, and the data type is three-dimensional point cloud or mesh model;

[0078] : represents a pre-trained deep generation model function, which is used to predict and render a complete three-dimensional model according to sparse two-dimensional input; in this embodiment, the model can adopt a neural radiance field or a generative adversarial network architecture known in the art;

[0079] : represents a sparse two-dimensional image data set as the input of the model, which is collected in the previous step;

[0080] : represents the internal network weight parameters of the generated model that have been converged and solidified; these parameters are determined by offline training on a professional data set containing thousands of groups of omnidirectional high-precision 3D scanning data and corresponding sparse angle 2D images, and are the training results of machine learning;

[0081] This method utilizes the strong prior knowledge of the deep learning model, solves the technical contradiction that traditional three-dimensional reconstruction technology requires multi-angle and long-time scanning, which causes strong stress interference to the living body, and realizes observation without intervention.

[0082] The system automatically extracts a set of growth characteristic parameters from the generated three-dimensional digital twin model; in this embodiment, the set of growth characteristic parameters specifically includes body length, body height, body weight estimated value, and body shape ratio; the body length is obtained by calculating the maximum distance along the main axis between the head vertex and the tail end on the three-dimensional digital twin model; the body height is obtained by calculating the maximum width perpendicular to the main axis of the model's torso part, such as the back to the abdomen; the body weight estimated value is obtained by calculating the grid volume of the three-dimensional model and multiplying it by a preset species density coefficient; the body shape ratio is obtained by calculating the ratio between the above-mentioned body length and body height;

[0083] Based on the above parameter set, in order to ensure the stability and repeatability of the evaluation results, the extracted set of growth characteristic parameters is strictly standardized before being input into the nonlinear evaluation model; this processing aims to eliminate the influence of differences in dimensions, units, or numerical ranges between different growth characteristics original values ; Specifically, the system will convert the original quantitative values to unified, dimensionless standardized values through Z-score standardization or Min-Max normalization method according to the historical statistical data of each characteristic ;

[0084] The standardized values are input into a preset nonlinear evaluation model for weighted calculation, thereby generating the final growth status grading index; this calculation process can be defined by the following growth status grading index function:

[0085]

[0086] wherein, : represents the final growth status grading index, which is a dimensionless scalar, and its role is to provide a direct basis for intelligent grading;

[0087] : represents the total number of growth characteristic parameters;

[0088] : represents the dimensionless weight coefficient of the i-th growth characteristic, and the sum of these weight coefficients is 1; these weight coefficients are not subjectively set, but are determined through statistical analysis of expert scoring data using objective weighting methods such as analytic hierarchy process or entropy weight method after inviting multiple experienced sea horse breeding experts and traders to conduct double-blind scoring on the standard sample set, in order to ensure the objectivity and industry recognition of the weights;

[0089] : represents the standardized value of the i-th growth characteristic extracted from the model M, which is a unified dimensionless value;

[0090] : represents a standardized value a function that performs a non-linear mapping, specifically a standard Sigmoid function: The function can smoothly map the normalized feature values to the interval [0, 1] to achieve the best degree of discrimination.

[0091] In this way based on deep learning modeling and multi-attribute decision theory evaluation, the evaluation of hippocampal growth state will be changed from the traditional way relying on subjective experience to a fully automated, standardized and highly accurate quantitative process; the deep generative model ensures that the model on which the evaluation is based is the most realistic and natural reflection of the hippocampus, and the growth state grading index function decomposes the abstract growth concept into calculable objective indicators, improving the objectivity, consistency and fineness of the grading results.

[0092] Embodiment 4:

[0093] S3 specifically includes:

[0094] normalizing the multi-source physiological data;

[0095] performing weighted calculation on the normalized multi-source physiological data according to the preset sensitivity weight coefficient to generate a physiological stress index.

[0096] This embodiment is based on embodiment 1 and aims at the calculation method of the physiological stress index in step S3.

[0097] In a preferred embodiment, in order to effectively fuse physiological data of different sources and different dimensions, the multi-source physiological data needs to be normalized. Specifically, for the collected raw data such as gill movement frequency, frequency and amplitude of escape behavior, and body color change rate, the system will perform linear normalization according to the dynamic range thereof in historical statistical data, and map it to the dimensionless interval [0, 1]. For example, the normalized value of the gill movement frequency change rate , the normalized value of the frequency and amplitude of escape behavior , and the normalized value of the body color change rate are all obtained in this way; wherein the escape behavior index is a comprehensive value obtained by weighted calculation on the normalized escape frequency and amplitude, so as to fully reflect the intensity of the stress behavior;

[0098] After normalization, the system will perform weighted calculation on the normalized multi-source physiological data according to the preset sensitivity weight coefficient to generate the final physiological stress index, and this calculation process is realized through the following physiological stress index function:

[0099]

[0100] wherein, : represents the final generated dimensionless physiological stress index, the higher the value, the stronger the tension of the hippocampus;

[0101] : respectively represent the aforementioned normalized gill beat frequency change rate, escape behavior index and body color change rate, the values of which are calculated from real-time monitoring data;

[0102] : are respectively the dimensionless sensitivity weight coefficients of the above three physiological parameters, and the sum of the three is 1; these weight coefficients are not equally set, but are obtained through controlled experiments; in the experiment, by gradually applying different intensities of external stimuli to the sample and synchronously monitoring the change intensity of each physiological indicator, the indicator determined as the earliest and most sensitive stress response indicator will be given the highest weight coefficient, to ensure that the system has the strongest early warning ability;

[0103] Through normalization processing and weighted calculation, the present application creates a quantitative indicator that can reflect the internal comprehensive stress level of the hippocampus in real time and sensitively ; this method not only integrates information from multiple physiological dimensions, but more importantly, by introducing the sensitivity weight calibrated through experiments, the index can focus on the most critical early warning signal, thereby providing a high-quality, high signal-to-noise ratio core feedback signal for the subsequent closed-loop adaptive control system, ensuring that the fishing behavior can be accurately guided by the physiological state of the hippocampus.

[0104] It should be noted that the linear weighting model used in this embodiment is a simplified calculation of complex physiological stress response. This model has good characterization ability and calculation efficiency in the monotonically increasing interval of stress level from zero to high. For specific varieties that may exhibit non-monotonic responses such as rigidity, a quadratic term can be introduced into the model or combined with other state judgment logic to improve the adaptability of the model.

[0105] Example 5:

[0106] S4 specifically includes:

[0107] When the physiological stress index is not higher than the safe stress threshold, the stress state level is determined to be in the safe zone, and an execution strategy is generated to maintain the current optimal path;

[0108] When the physiological stress index is higher than the safe stress threshold and not higher than the critical stress threshold, the stress state level is determined to be in the warning zone, and a correction strategy is generated to dynamically reduce the movement speed of the mechanical arm;

[0109] When the physiological stress index is higher than the critical stress threshold, the stress state level is determined to be in the danger zone, and an evasion strategy is generated to abort the task and retreat to a safe distance.

[0110] The correction strategy of the early warning zone is specifically:

[0111] The physiological stress index, the safe stress threshold and the critical stress threshold are input into a preset speed correction function to calculate the corrected mechanical arm movement speed; wherein the corrected mechanical arm movement speed is smoothly and nonlinearly attenuated with the increase of the physiological stress index.

[0112] After the avoidance strategy of the danger zone is executed, it further includes:

[0113] The physiological stress index is continuously monitored, and when the physiological stress index falls below the safe stress threshold, the fishing and sorting operation is resumed.

[0114] Based on the embodiment 1, the hierarchical adaptive correction strategy in step S4 is a complete data-driven decision and control process. It converts the physiological stress index into precise control instructions for the mechanical arm;

[0115] The strategy compares the physiological stress index with the safe stress threshold and the critical stress threshold The specific values of these two thresholds are not fixed and need to be experimentally calibrated for specific sea horse breeds before use; the calibration process includes: in a controlled environment, gradually approach the sample sea horse with the mechanical arm at different speeds, while continuously monitoring the changes in its physiological stress index Through analysis of multiple sets of experimental data, determine the index level at which the sea horse generally begins to show initial nervousness, and determine the index level at which it begins to show violent escape or rigidity behavior, and determine Thus, the thresholds are preset; the stress state of the sea horse is divided into three levels, and the corresponding control logic is executed:

[0116] First level state: when the physiological stress index is not higher than the safe stress threshold, i.e. , the system determines the stress state level as the safe zone; in this state, the system considers that the activity of the mechanical arm does not cause significant disturbance to the sea horse; therefore, the decision control module generates an execution strategy to maintain the current optimal path, and instructs the flexible mechanical arm of the operation execution module to execute the fishing and sorting operation according to the optimal path and speed pre-planned based on the three-dimensional digital twin model;

[0117] Second level state: when the physiological stress index is higher than the safe stress threshold but not higher than the critical stress threshold, i.e. , the system determines that the stress state level is in the warning zone; this indicates that the hippocampus has begun to perceive a threat and produce a preliminary nervous emotion; at this time, the system generates a correction strategy for dynamically adjusting the speed of the manipulator motion to actively slow down the disturbance to the target;

[0118] The correction strategy in the warning zone is as follows: the real-time physiological stress index , the safety stress threshold , and the critical stress threshold are input into a preset speed correction function to calculate the corrected speed of the manipulator motion ; the function is designed as follows:

[0119]

[0120] wherein, : represents the output motion speed of the manipulator after correction, which has the same dimension as ;

[0121] : represents the original speed of the current planning, which is given by the optimal path planning algorithm;

[0122] : are the real-time stress index, safety threshold, and critical threshold, respectively, which are all dimensionless parameters;

[0123] : is a dimensionless adjustment factor greater than 1, which is used to adjust the nonlinearity of the speed decay curve; its value is determined according to the variety sensitivity experimental data, wherein the sensitivity is quantified as the average slope of the stress index response curve under standard incremental stimulation, and the greater the slope, the greater the value of ;

[0124] The corrected speed of the manipulator motion will be smoothly and nonlinearly decayed from to 0 as the physiological stress index increases; this intelligent approximation behavior can reduce the intervention intensity in a gradual and safe manner, effectively pacifying the target and avoiding further escalation of the stress state;

[0125] Third level: when the physiological stress index is higher than the critical stress threshold, i.e. , the system determines that the stress state level is in the danger zone; this indicates that the hippocampus is in a high stress state, and any further approach may cause it to struggle violently or suffer physiological damage; at this time, the system generates an evasive strategy to abort the task and retreat to a safe distance; the operation execution module immediately instructs the manipulator to abort the current capture task and actively moves backward to a safe distance that does not cause stress to the hippocampus, and then suspends the operation;

[0126] After executing the evasive strategy in the danger zone, the system does not stop permanently, but enters a waiting mode; in this mode, the state monitoring module continuously monitors the physiological stress index, and when the index falls below the safe stress threshold , the system determines that the stress state of the hippocampus has returned to normal, and resumes execution of the fishing and pooling operation;

[0127] Through this set of hierarchical adaptive correction and recovery logic, the application constructs a truly biological signal dominant closed-loop control system; the system not only can respond to the stress level of the hippocampus in proportion, smoothly, but also can execute decisive termination and evasion operation in extreme cases, and intelligently resume the task after the danger is removed; this fine, multi-level control strategy ensures that the fishing and pooling operation is always within the physiological bearing range of the hippocampus, maximally reduces physical intervention, and realizes safe and non-destructive acquisition of the living target.

[0128] The physiological stress index function and the speed correction function are optimized in the scheme; for the calculation of the physiological stress index , in order to avoid the failure of normalization due to the fluctuation of the historical data range, the system introduces a dynamic safety margin ; when any original physiological data exceeds the historical maximum value, the system will adaptively expand the normalization dynamic range of the feature, to ensure that its normalized value is always in the effective interval [0, 1];

[0129] In the speed correction function, when the difference between the critical stress threshold and the safe stress threshold is less than a preset minimum value , to prevent the denominator from tending to zero and causing calculation instability, the system will directly execute the danger zone evasive strategy; this strategy can be defined as: if , then directly execute ;

[0130] This correction ensures that even in the rare case where the threshold settings are extremely close, the system can make a safe and decisive response, avoiding uncontrollable movement of the robotic arm due to calculation error;

[0131] In addition, in order to deal with the temporary loss or too low quality of sensor data, the decision control module also has data validity checking logic built in; if valid physiological data cannot be received within the preset time window, the system will enter the warning zone by default and execute the slow approach strategy instead of maintaining high-speed motion to ensure safety; if the data loss lasts for more than a longer threshold, the danger zone avoidance strategy is executed until the data returns to normal, thereby ensuring the fault safety capability of the system under abnormal sensor conditions.

[0132] Embodiment 6:

[0133] Please refer to Figure 2 A hippocampus intelligent grading and separate pond breeding system based on appearance features, comprising:

[0134] A data acquisition module for acquiring sparse visual data and multi-source physiological data of the hippocampus;

[0135] A modeling and evaluation module for generating a three-dimensional digital twin model based on the sparse visual data and calculating a growth state grading index based on the three-dimensional digital twin model;

[0136] A state monitoring module for calculating a physiological stress index based on the multi-source physiological data;

[0137] A decision control module for comparing the physiological stress index with a preset safe stress threshold and a critical stress threshold to generate a grading adaptive correction strategy;

[0138] An operation execution module for receiving the grading adaptive correction strategy and controlling the flexible mechanical arm to perform the catching and separate pond operation.

[0139] The embodiment provides a hippocampus intelligent grading and separate pond breeding system based on appearance features, which is a collection of hardware and software designed to implement the foregoing method, and comprises:

[0140] A data acquisition module, which is used to non-contact acquisition of raw data of the hippocampus; in the embodiment, the module is composed of a long-distance visual sensor and a non-contact visual sensor array deployed in the breeding environment; it is responsible for performing the S1 step, i.e., for acquiring sparse visual data and multi-source physiological data of the hippocampus and transmitting the data to the subsequent module in real time;

[0141] A modeling and evaluation module, which is used to quantitatively evaluate the growth state of the hippocampus; the module is usually a server or an edge computing device integrated with a high-performance computing unit such as a GPU, and a pre-trained deep generation model and a nonlinear evaluation model are fixed inside; it is responsible for performing the S2 step, i.e., for generating a three-dimensional digital twin model based on the sparse visual data and calculating a growth state grading index based on the model;

[0142] a state monitoring module, which aims to quantify the physiological stress level of hippocampus in real time; this module can be an independent signal processing unit or a software algorithm integrated in the main controller, which embeds the physiological stress index function; it is responsible for performing S3, i.e., for calculating the physiological stress index in real time based on multi-source physiological data;

[0143] a decision control module, which aims to generate intelligent control strategies according to real-time states; this module is the central control unit of the system, usually implemented by a central processing unit (CPU) or a microcontroller (MCU), and its core is the hierarchical adaptive correction logic; it is responsible for performing S4, i.e., for comparing the physiological stress index calculated by the state monitoring module with the preset safety and critical stress thresholds, and generating the corresponding hierarchical adaptive correction strategy;

[0144] a job execution module, which aims to perform the final physical operation; this module is mainly composed of a high-precision flexible manipulator and its motion controller; it is responsible for performing S5, i.e., for receiving the hierarchical adaptive correction strategy generated by the decision control module, and accurately controlling the flexible manipulator to perform the fishing and pooling operation, including a series of actions such as normal travel, dynamic deceleration, suspension and retreat;

[0145] Through the organic combination and collaborative work of the above modules, the system integrates data acquisition, modeling evaluation, state monitoring, intelligent decision and physical execution into one, forming a complete and automated intelligent grading and pooling aquaculture system; each module has its own function, and the data flow and control flow are clear and efficient, thus converting the complex method process into a stable, reliable and deployable industrial solution, providing a solid technical support for realizing the high-quality and standardized processing of precious live aquatic products.

[0146] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. A hippocampus intelligent grading and ponding method based on appearance characteristics, characterized in that, The method comprises the following steps: S1, sparse visual data and multi-source physiological data of hippocampus are collected; S2, a three-dimensional digital twin model is generated based on the sparse visual data; a growth state classification index is calculated based on the three-dimensional digital twin model; S3, a physiological stress index is calculated based on the multi-source physiological data; S4, the real-time calculated physiological stress index is compared with the preset safe stress threshold and critical stress threshold to generate a classification adaptive correction strategy; S5, the flexible mechanical arm is controlled to perform the catching and pool sorting operation according to the classification adaptive correction strategy.

2. The hippocampus intelligent grading and ponding method based on appearance characteristics according to claim 1, characterized in that, S1 specifically comprises: Low-resolution two-dimensional images of the target are collected from multiple preset azimuth angles by a long-distance visual sensor to construct a sparse visual data set; Multi-source physiological data are collected by a non-contact visual sensor; wherein the multi-source physiological data include gill movement frequency, limb movement frequency and amplitude, and instantaneous body color change rate.

3. The hippocampus intelligent grading and grouping method based on appearance characteristics according to claim 1, characterized in that, S2 specifically comprises: The sparse visual data set is input into a pre-trained deep generation model to generate a three-dimensional digital twin model; Growth characteristic parameter sets are extracted from the three-dimensional digital twin model; wherein the growth characteristic parameter sets include body length, body height, body weight estimation value and body shape ratio; The growth characteristic parameter sets are input into a preset classification model for weighted calculation to generate a growth state classification index.

4. The hippocampus intelligent grading and grouping method based on appearance characteristics according to claim 1, characterized in that, S3 specifically comprises: The multi-source physiological data are normalized; The normalized multi-source physiological data are weighted calculated according to preset sensitivity weight coefficients to generate a physiological stress index.

5. The hippocampus intelligent grading and grouping method based on appearance characteristics according to claim 1, characterized in that, S4 specifically comprises: When the physiological stress index is not higher than the safe stress threshold, the stress state level is determined as the safe zone, and an execution strategy of maintaining the current optimal path is generated; When the physiological stress index is higher than the safe stress threshold and not higher than the critical stress threshold, the stress state level is determined as the warning zone, and a correction strategy of dynamically adjusting and reducing the mechanical arm movement speed is generated; When the physiological stress index is higher than the critical stress threshold, the stress state level is determined as the danger zone, and an evasive strategy of aborting the task and retreating to a safe distance is generated.

6. The hippocampus intelligent grading and ponding method based on appearance characteristics according to claim 5, characterized in that, The correction strategy of the warning zone specifically comprises: The physiological stress index, the safe stress threshold and the critical stress threshold are input into a preset speed correction function to calculate a corrected mechanical arm movement speed; wherein the corrected mechanical arm movement speed is smoothly and nonlinearly attenuated with the increase of the physiological stress index.

7. The hippocampus intelligent grading and ponding method based on appearance characteristics according to claim 5, characterized in that, After the execution of the evasive strategy of the danger zone, it further comprises: The physiological stress index is continuously monitored, and when the physiological stress index falls below the safe stress threshold, the catching and pool sorting operation is resumed.

8. A hippocampus intelligent grading and separate pool breeding system based on appearance characteristics, based on the hippocampus intelligent grading and separate pool breeding method based on appearance characteristics of any one of claims 1-7, characterized in that, It comprises: A data collection module for collecting sparse visual data and multi-source physiological data of hippocampus; A modeling and evaluation module for generating a three-dimensional digital twin model based on the sparse visual data, and calculating a growth state classification index based on the three-dimensional digital twin model; A state monitoring module for calculating a physiological stress index based on multi-source physiological data; A decision control module for comparing the physiological stress index with the preset safe stress threshold and critical stress threshold to generate a classification adaptive correction strategy; An operation execution module for receiving the classification adaptive correction strategy and controlling the flexible mechanical arm to perform the catching and pool sorting operation.

Images