Fish feed amount analysis method and system based on machine vision

Abstract

The present invention relates to the technical field of fish feed amount, and in particular to a fish feed amount analysis method and system based on machine vision, including: collecting fish school movement video images through electronic monitoring equipment, and performing distribution density analysis between fish morphology differences to obtain distribution density data between fish morphology differences; performing feeding demand cycle benchmark simulation evaluation based on the distribution density data between fish morphology differences to obtain feeding demand cycle benchmark data; identifying metabolic influencing factors for movement consumption estimation data between fish morphology differences, and performing feeding demand metabolic impact correction to obtain feeding demand metabolic impact correction data; constructing a fish feed amount adjustment model for the feeding demand metabolic impact correction data through a strategy gradient algorithm to obtain a fish feed amount adjustment model, and sending the fish feed amount adjustment model to a cloud platform. The present invention makes the fish feed amount technology more perfect by optimizing the fish feed amount technology.

Description

Technical Field

[0001] The present invention relates to the technical field of fish feed feeding amount, and in particular to a fish feed feeding amount analysis method and system based on machine vision. Background Art

[0002] The fish feed amount analysis method based on machine vision combines image recognition, pattern recognition, deep learning and other technologies, and can evaluate the changes in fish demand in real time in an automated way. This method can not only accurately capture the distribution, activity and feeding behavior of fish, but also identify the feeding amount required for fish at different growth stages, significantly improving the level of refined management of aquaculture. The machine vision system captures the behavior and state of fish through high-resolution camera equipment, extracts characteristic data such as fish movement trajectory and fish morphological differences based on computer vision algorithms, and then combines biological models, temperature and humidity and other environmental factors to predict the actual feeding needs of fish and automatically adjust the feed amount. In addition, machine vision can also analyze the changes in the appetite of fish, adjust the feed type and ratio in time, thereby further improving feed utilization, reducing waste, and optimizing feeding effects. However, the traditional fish feed amount analysis method based on machine vision has the problem of inaccurate analysis of the movement consumption of fish of different sizes and the influence of temperature, resulting in large feeding errors and low feed utilization. Summary of the invention

[0003] Based on this, it is necessary to provide a fish feed feeding amount analysis method and system based on machine vision to solve the above technical problems.

[0004] To achieve the above object, the present invention provides a method for analyzing fish feed amount based on machine vision in a first aspect, the method comprising the following steps:

[0005] Step S1: collecting fish school motion video images through electronic monitoring equipment to obtain fish school motion video images; performing fish school morphology difference analysis on the fish school motion video images to obtain fish school morphology difference data; performing distribution density analysis between fish morphology differences based on the fish school morphology difference data to obtain distribution density data between fish morphology differences;

[0006] Step S2: marking the spatiotemporal motion trajectories of fish with different morphological shapes according to the distribution density data of fish with different morphological shapes, and obtaining the spatiotemporal motion trajectory data of fish with different morphological shapes; estimating the motion consumption of fish individuals based on the spatiotemporal motion trajectory data of fish with different morphological shapes, and obtaining the motion consumption estimation data of fish with different morphological shapes; performing a feeding demand cycle benchmark simulation evaluation based on the motion consumption estimation data of fish with different morphological shapes, and obtaining the feeding demand cycle benchmark data;

[0007] Step S3: acquiring real-time temperature data of the fish pond; identifying metabolic influencing factors of the estimated exercise consumption data between fish morphological differences according to the real-time temperature data of the fish pond to obtain temperature-related metabolic influencing factors; performing feeding demand metabolic impact correction on the feeding demand cycle benchmark data based on the temperature-related metabolic influencing factors to obtain feeding demand metabolic impact correction data;

[0008] Step S4: construct a fish feed amount adjustment model based on the feeding demand metabolic impact correction data through a policy gradient algorithm to obtain a fish feed amount adjustment model, and send the fish feed amount adjustment model to the cloud platform to perform fish feed amount analysis.

[0009] The present invention collects real-time video images of the movement of fish schools through electronic monitoring equipment, and can efficiently obtain dynamic information of fish schools. By analyzing the morphological differences of fish schools on these video images, the morphological characteristics of different fish individuals in the fish school can be identified, and detailed morphological difference data of fish schools can be obtained. These data provide a basis for subsequent distribution density analysis and motion trajectory analysis, which is helpful for accurately evaluating the behavior patterns and morphological characteristic changes of fish schools, and providing a scientific basis for feeding management. After obtaining the morphological difference data of fish schools, the spatiotemporal motion trajectory between the morphological differences of fish is marked, and the motion path of each fish individual in the spatial and temporal dimensions can be accurately tracked. Combined with these trajectory data, the motion consumption between individual fish is further estimated to provide a quantitative basis for evaluating the energy demand of fish schools. Based on the motion consumption data, a benchmark simulation evaluation of the feeding demand cycle is performed, and a reasonable feeding demand cycle can be obtained, which provides scientific support for the healthy growth and resource management of fish. By obtaining the temperature data of the fish pond in real time, the metabolic effect of temperature on the motion consumption of fish schools can be identified, and the association between temperature and fish metabolism can be established. Based on the identification of this temperature-related metabolic influencing factor, the previously estimated baseline data of the feeding demand cycle can be corrected to obtain more accurate feeding demand correction data. This step can better take into account the impact of ambient temperature on fish growth and metabolism, thereby optimizing feeding strategies, improving resource utilization efficiency, and ensuring the nutritional supply of fish schools under different temperature conditions. Using the policy gradient algorithm, a fish feed amount adjustment model is constructed based on the corrected feeding demand data. The model can dynamically adjust the feeding amount of fish feed to ensure the optimal nutritional supply of fish schools and effectively avoid resource waste. Uploading the model to the cloud platform can perform real-time fish feed feeding amount analysis, further optimize the feeding strategy, and adjust and optimize according to the actual operation situation to improve the management efficiency and production benefits of intelligent farming.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: collecting fish school movement video images through electronic monitoring equipment to obtain fish school movement video images;

[0012] Step S12: performing image scale normalization processing on the fish school motion video image to obtain scale normalized data of the fish school motion video image;

[0013] Step S13: performing fish school morphology difference analysis on the scale-normalized data of the fish school motion video image to obtain fish school morphology difference data;

[0014] Step S14: performing distribution density analysis of differences in fish morphology on scale-normalized data of fish school motion video images according to the fish school morphology difference data to obtain distribution density data of differences in fish morphology.

[0015] The present invention collects video images of fish movement through electronic monitoring equipment, which is the basic link for obtaining fish behavior data. Efficient image acquisition can not only record the dynamic changes of fish in the water in real time, but also provide rich raw data for subsequent image processing and analysis. Through these video images, the morphology, distribution and movement of fish can be observed in detail, ensuring the accuracy and reliability of subsequent analysis, and providing data support for precise management. The scale normalization processing of the collected video images of fish movement can unify the different sizes in the image and eliminate the scale deviation caused by shooting angle or distance difference. This process ensures that the image data collected at different times and under different environmental conditions are consistent, thereby providing standardized input data for subsequent morphological difference analysis. This step effectively improves the accuracy and comparability of the analysis and avoids data errors caused by scale differences. Morphological difference analysis of fish movement video images after scale normalization can identify the differences between individuals in the fish, such as fish size, shape, posture and other characteristics. Through this analysis, the key morphological information of individual fish can be extracted, providing a basis for subsequent motion trajectory analysis, behavior prediction, etc. This step helps to gain a deeper understanding of the growth status, health status, and environmental adaptability differences of fish schools, and provides a scientific basis for precise aquaculture management. Based on the fish morphological difference data, the distribution density analysis of the scale-normalized data of the fish movement video images can reveal the spatial distribution of individuals of different morphologies in the fish school. Through this analysis, the density distribution characteristics of fish schools in different areas and times can be obtained, and then the clustering behavior of fish schools and the interaction patterns between individuals can be identified. This data is very important for feeding, disease prevention, and prediction of fish behavior, and helps to optimize the aquaculture environment and improve resource utilization efficiency.

[0016] Preferably, step S2 comprises the following steps:

[0017] Step S21: marking the spatiotemporal motion trajectories of the fish morphology differences on the fish school motion video image according to the distribution density data of the fish morphology differences, and obtaining the spatiotemporal motion trajectory data of the fish morphology differences;

[0018] Step S22: based on the spatiotemporal motion trajectory data of the fish morphology differences, the motion consumption estimation data of the fish morphology differences are performed on the fish morphology difference data to obtain the motion consumption estimation data of the fish morphology differences;

[0019] Step S23: performing consumption cluster regression analysis on the motion consumption estimation data of fish morphology differences to obtain motion consumption cluster regression data;

[0020] Step S24: performing a feeding demand cycle benchmark simulation evaluation on the fish school morphology difference data according to the movement consumption cluster regression data and the distribution density data between the fish morphology differences to obtain the feeding demand cycle benchmark data.

[0021] The present invention marks the spatiotemporal motion trajectory of the fish school motion video image according to the distribution density data between the fish morphological differences, which helps to accurately track the motion path of each fish individual in different time and space. This process can provide dynamic change information of the fish school in the water, making the subsequent behavior analysis more detailed and accurate. Through the marking of the spatiotemporal motion trajectory, the movement pattern, behavior law and response of the fish school to environmental changes can be more clearly understood, providing data support for feeding management and environmental optimization. Based on the spatiotemporal motion trajectory data between the marked fish morphological differences, the movement consumption between the fish individuals is estimated, and the energy consumption of each fish individual under specific movement conditions can be effectively evaluated. The movement consumption of the fish individual is not only related to its body shape and morphological differences, but also affected by factors such as exercise intensity and duration. Through this estimation process, a quantitative basis can be provided for the subsequent feeding needs, thereby ensuring that the fish school obtains sufficient energy supply, optimizing the feeding strategy and improving the breeding efficiency. Consumption cluster regression analysis is performed on the movement consumption estimation data between the fish morphological differences, and the movement consumption patterns of different fish individuals can be further analyzed. Cluster regression analysis can identify the energy consumption patterns of different types of individuals in a fish school, and classify different types of fish individuals according to these patterns, thereby providing a basis for the personalized adjustment of subsequent feeding strategies. Through this analysis, the overall energy demand of fish schools can be accurately predicted, excessive or insufficient feeding can be avoided, and resource utilization efficiency can be improved. Based on the cluster regression data of motion consumption and the distribution density data between fish morphological differences, a baseline simulation evaluation of the feeding demand cycle of fish morphological difference data can be performed to obtain the required feeding amount in each time period. This evaluation takes into account the motion consumption, individual differences and spatial distribution of fish schools, and provides scientific feeding cycle data for the long-term breeding planning of fish schools. Through this evaluation, the feeding strategy can be optimized to ensure that fish schools can obtain appropriate nutrient supply at different growth stages, improve breeding efficiency and fish health.

[0022] Preferably, step S22 includes the following steps:

[0023] Step S221: Calculate the size of each fish based on the fish morphology difference data to obtain morphology, body size and data;

[0024] Step S222: quantifying the fluid resistance of each fish according to the fish morphology difference data based on the morphology and body size data to obtain the fish morphology fluid resistance quantification data;

[0025] Step S223: extracting the swimming mode of each fish from the fish school motion video image based on the spatiotemporal motion trajectory data between the fish morphological differences, and obtaining the fish swimming mode data;

[0026] Step S224: performing resistance compensation between different swimming modes on the quantified data of the fluid resistance of the fish morphology according to the data of the swimming mode of the fish, and obtaining the fluid resistance compensation data of the fish morphology-swimming mode;

[0027] Step S225: Estimate the motion consumption of individual fish based on the spatiotemporal motion trajectory data of different fish morphologies according to the fish morphology-swimming mode fluid resistance compensation data, and obtain motion consumption estimation data of different fish morphologies.

[0028] The present invention calculates the size of each fish based on the fish morphological difference data, and can obtain detailed body information of each individual fish. This process takes into account important parameters such as the length, width, and weight of the fish body, and provides basic data for subsequent exercise consumption estimation. The calculation of body size is a key link in evaluating the growth status and health status of individual fish, which helps to understand the relationship between the fish body and factors such as energy consumption and swimming ability, and lays the foundation for accurate energy demand prediction and feeding management. According to the body size data, the fluid resistance of each fish is quantified, and the water flow resistance encountered by the fish when swimming in the water can be quantified. The size, shape and surface characteristics of the fish body determine the resistance it needs to overcome when swimming in the water. Through this quantification process, an accurate evaluation basis can be provided for the exercise consumption of fish of different sizes, ensuring that personalized energy consumption calculations can be performed based on the fluid resistance of the fish body, thereby optimizing the feeding plan and avoiding unnecessary energy waste. Based on the spatiotemporal motion trajectory data of the fish school, the swimming mode of each fish is extracted, which helps to identify the swimming habits of different individual fish, such as swimming speed, swimming mode (such as straight swimming, circling swimming, etc.). Different swimming patterns will affect the energy consumption of individual fish in water. Therefore, by accurately extracting swimming pattern data, more detailed information can be provided for subsequent motion consumption estimation. This step can improve the accuracy of energy consumption estimation and ensure that a reasonable feeding strategy is formulated according to the actual swimming characteristics of the fish. According to the fish swimming pattern data, the resistance compensation between different swimming patterns of the fish morphology fluid resistance quantification data can take into account the difference in fluid resistance of the fish body under different swimming modes. Different swimming patterns cause different resistances to the fish body in water, so it is necessary to compensate for the swimming pattern to more accurately reflect the impact of different swimming patterns on energy consumption. Through this compensation process, the estimation of the true energy consumption of each individual fish can be improved, ensuring the precise regulation of the feeding amount and avoiding excessive or insufficient feeding. According to the fish morphology-swimming pattern fluid resistance compensation data, the spatiotemporal motion trajectory data between the fish morphology differences are used to estimate the motion consumption of individual fish, so that the energy consumption of each individual fish under specific motion conditions can be more accurately calculated. Combining factors such as body shape, fluid resistance, and swimming pattern, the motion consumption of individual fish can be comprehensively evaluated. This estimate provides a scientific basis for the formulation of reasonable feeding plans, which helps ensure that fish stocks have access to appropriate energy supply, thereby optimizing farming efficiency and improving the healthy growth of fish.

[0029] Preferably, performing resistance compensation between different swimming modes on the quantified data of fluid resistance of fish morphology according to the swimming mode data of fish comprises the following steps:

[0030] According to the fish swimming mode data, key posture nodes between different swimming modes are analyzed to obtain swimming mode posture node analysis data;

[0031] Based on the swimming mode posture node analysis data, fluid dynamic field simulation is performed to obtain posture node associated fluid dynamic field data;

[0032] The mean difference of surface pressure distribution of the attitude node is calculated for the fluid dynamic field data associated with the attitude node to obtain the mean difference data of surface pressure distribution of the node;

[0033] According to the mean difference data of node surface pressure distribution, the resistance compensation between different swimming modes is performed on the quantitative data of fish morphology fluid resistance to obtain the fish morphology-swimming mode fluid resistance compensation data.

[0034] The present invention analyzes the key nodes of the postures between different swimming modes according to the data of the swimming mode of fish, which is helpful to deeply understand the different posture changes of fish when swimming in water. Each swimming mode has a specific posture and movement characteristics. By analyzing the key nodes of the postures, the changing position and angle of each key point of the fish body during the swimming process can be clarified. This process provides accurate geometric and physical basis for subsequent fluid dynamics simulation and energy consumption evaluation, and provides support for optimizing the relationship between the swimming mode of fish and energy consumption. Fluid dynamic field simulation based on the swimming mode posture node analysis data can simulate the fluid mechanics effect on the fish body when swimming in water. Fluid dynamic field simulation can help predict the force conditions of different parts of the fish body in the water flow, and reveal the interaction between the water flow and the surface of the fish body. Through this simulation, more detailed fluid dynamic data can be obtained, which helps to analyze the resistance changes of the fish body under different swimming modes, and provides a scientific basis for subsequent resistance compensation and motion consumption estimation. The mean difference calculation of the surface pressure distribution of the posture node is performed on the posture node associated fluid dynamic field data, which helps to evaluate the difference in fluid pressure on various parts of the fish body surface. The pressure distribution on the surface of the fish directly affects the energy consumption during swimming. By calculating the mean difference of pressure distribution under different postures, the key areas of pressure change under each swimming mode can be identified. This analysis provides a quantitative basis for the resistance compensation of the fish body, ensuring that the actual fluid resistance encountered by the fish in the water can be accurately calculated, and further optimizing the estimation of motion consumption. According to the mean difference data of the node surface pressure distribution, the resistance compensation between different swimming modes is performed on the quantitative data of the fluid resistance of the fish morphology, and the fluid resistance data under each swimming mode can be accurately adjusted. This compensation process combines the actual posture, swimming mode and physical characteristics of the fish body, and obtains a more accurate actual resistance value of the fish body by compensating for the resistance differences between different swimming modes. This compensation result provides accurate parameter support for subsequent motion consumption estimation, which helps to optimize the feeding amount and breeding management of the fish school and improve the overall breeding efficiency.

[0035] Preferably, estimating the motion consumption of the spatiotemporal motion trajectory data between the fish morphology differences according to the fish morphology-swimming mode fluid resistance compensation data comprises the following steps:

[0036] According to the fish morphology-swimming mode fluid resistance compensation data, the spatiotemporal motion trajectory data between the fish morphology differences are extracted for the speed changes between the resistance states of different swimming modes, and the speed change data between the resistance states of different swimming modes are obtained;

[0037] Based on the speed change data between different swimming mode resistance states and the fish morphology-swimming mode fluid resistance compensation data, multi-directional force output balance decomposition is performed to obtain multi-directional force output decomposition data;

[0038] Perform output torque matching on the multi-directional force output decomposition data to obtain multi-directional output torque matching data;

[0039] The motion consumption of the spatiotemporal motion trajectory data of fish with different morphologies is estimated based on the multi-directional force output decomposition data and the multi-directional output torque matching data, so as to obtain the motion consumption estimation data of fish with different morphologies.

[0040] The present invention extracts the speed change between the resistance states of different swimming modes from the spatiotemporal motion trajectory data of the fish morphology differences according to the fluid resistance compensation data of the fish morphology-swimming mode, and can accurately identify the speed change law of the fish body under different swimming modes. Since the fluid resistance of the fish body under different swimming modes is different, this will cause the change of the swimming speed. By extracting these speed change data, the influence of the resistance under different swimming modes on the movement of the fish school can be effectively evaluated, and a reliable speed basis can be provided for the subsequent movement consumption estimation. This data provides accurate input for further mechanical analysis. Based on the speed change data between the resistance states of different swimming modes and the fluid resistance compensation data of the fish morphology-swimming mode, the multi-directional force output balance decomposition can be performed, and the multiple mechanical effects experienced by the fish body under different swimming modes can be quantified, including forward force, resistance, lateral force, etc. This process provides a comprehensive understanding of the mechanical state of each fish in the water through the balance decomposition of force. The multi-directional force decomposition data can accurately describe the complex mechanical relationship under different movement states, provide detailed support for the accurate estimation of movement consumption, and play a vital role in the movement trajectory and energy conversion process. Output torque matching of multi-directional force output decomposition data helps to ensure torque balance under multiple mechanical effects, and adjust the accuracy of fish movement under different swimming modes according to the matching results. By calculating torque matching, the influence of forces in different directions on the rotation and balance of the fish body can be analyzed, so as to obtain more accurate resistance and motion consumption data. This process ensures that the energy use and mechanical state of the fish body are accurately reflected in the complex swimming process, providing accurate mechanical parameters for subsequent energy consumption estimation. According to the multi-directional force output decomposition data and multi-directional output torque matching data, the motion consumption of the spatiotemporal motion trajectory data between the fish morphological differences is estimated, which can comprehensively estimate the energy consumption under different morphologies and swimming modes. By combining the decomposition results of multi-directional forces with the torque matching data, the actual energy consumption of the fish body moving in the water is estimated. This estimation takes into account the morphological differences, swimming modes and complex mechanical effects of the fish body, so as to obtain more accurate motion consumption data. This provides a scientific basis for precise feeding management, energy supply and breeding optimization, ensuring that the fish can obtain appropriate energy support and optimize growth effects.

[0041] Preferably, step S24 includes the following steps:

[0042] Step S241: performing theoretical growth cycle analysis of fish morphology difference data according to the movement consumption cluster regression data, and obtaining theoretical growth cycle of fish morphology difference;

[0043] Step S242: performing a periodic fluctuation analysis of satiation intake between different growth stages of the theoretical growth cycle of fish morphological differences based on the exercise consumption cluster regression data to obtain periodic fluctuation data of satiation intake;

[0044] Step S243: defining the growth base increments between different fish morphological differences for the theoretical growth period of fish morphological differences according to the distribution density data between different fish morphological differences, and obtaining the growth base increment data between different fish morphological differences;

[0045] Step S244: Perform a feeding requirement cycle benchmark simulation evaluation based on the growth base increment data and satiation intake cycle fluctuation data of different fish morphological differences to obtain feeding requirement cycle benchmark data.

[0046] The present invention analyzes the theoretical growth cycle of fish morphology difference data according to the cluster regression data of motion consumption, which can help identify the time course of fish swarms with different forms in the growth process. Each fish morphology has different energy requirements and consumption, so their growth cycles will also be different. Through the analysis of the theoretical growth cycle, it can be revealed how the differences in fish morphology affect their overall growth progress, provide a theoretical basis for subsequent feeding management and optimization, ensure accurate prediction of the growth cycle of different fish morphologies, and thus formulate reasonable breeding strategies. Based on the cluster regression data of motion consumption, the theoretical growth cycle of fish morphology difference is analyzed for periodic fluctuations of satiation intake between different growth stages, which helps to reveal the energy intake pattern of fish swarms at different growth stages. Fish have different growth stages, and their demand for food also fluctuates periodically. By analyzing these fluctuations, scientific feeding plans can be formulated for different growth stages to ensure that fish swarms take in enough nutrition at each stage, thereby optimizing growth rate and avoiding waste. The analysis results provide important data support for optimizing feeding strategies and improving breeding efficiency. Based on the distribution density data between the fish morphological differences, the theoretical growth cycle of fish morphological differences is defined to define the growth base increment between different fish morphological differences, which can clarify the energy increment required by different fish morphologies during their growth. This definition process provides customized increment data for each fish morphology by quantifying the basic energy required by different morphological fish bodies during their growth. This data not only helps to accurately calculate the actual energy requirements of each fish swarm, but also optimizes energy allocation, thereby improving feeding efficiency and economic benefits. Based on the growth base increment data between different fish morphological differences and the satiation intake cycle fluctuation data, the feeding requirements of fish swarms at different growth stages can be accurately predicted. This simulation evaluation provides a comprehensive feeding cycle evaluation method by integrating factors such as morphological differences, energy consumption, food intake cycle and growth base increment. It can not only ensure that fish swarms have enough energy intake at each stage, but also effectively avoid overfeeding or lack of nutrition, thereby achieving more scientific feeding management, improving breeding efficiency and reducing costs.

[0047] Preferably, step S3 comprises:

[0048] Step S31: performing convolution calculation on the feeding demand cycle benchmark data to obtain feeding demand cycle benchmark convolution data;

[0049] Step S32: Obtaining real-time temperature data of the fish pond;

[0050] Step S33: identifying metabolic influencing factors of the estimated exercise consumption data of fish morphological differences according to the real-time temperature data of the fish pond, and obtaining temperature-related metabolic influencing factors;

[0051] Step S34: performing feeding demand metabolic impact correction on the feeding demand cycle benchmark convolution data based on the temperature-related metabolic impact factor to obtain feeding demand metabolic impact correction data.

[0052] The present invention performs convolution calculation on the feeding demand cycle benchmark data, and can effectively integrate the historical feeding demand data with the current periodic pattern. Through the convolution calculation, the feeding demand change trend in different time periods can be better revealed, thereby optimizing the feeding plan. This step captures the actual demand fluctuations of fish at different time nodes through weighted integration of periodic demand, provides data support for subsequent feeding adjustment and optimization, and makes the breeding process more accurate and efficient. The real-time temperature data of the fish pond is obtained to accurately reflect the impact of the external environment on the physiological activities of fish. Water temperature has a significant impact on the metabolic rate, feeding behavior and growth rate of fish, so real-time temperature monitoring is an important basis for regulating breeding management. By obtaining real-time temperature data, direct environmental parameter support can be provided for the subsequent identification of metabolic influencing factors and the correction of feeding strategies, ensuring that feeding management can respond to water temperature changes in a timely manner, thereby optimizing the breeding effect. According to the real-time temperature data of the fish pond, the metabolic influencing factor identification of the estimated data of exercise consumption between fish morphological differences can reveal the specific impact of different water temperatures on fish metabolism and energy consumption. Increased or decreased water temperature will cause changes in the metabolic level of fish, thereby affecting their food intake and energy consumption. By identifying the relationship between temperature and metabolism, the metabolic influencing factors of temperature can be accurately extracted, thus providing a scientific basis for subsequent feeding strategy adjustments and ensuring that the energy needs of fish are reasonably matched. Based on the temperature-related metabolic influencing factors, the metabolic impact of feeding demand cycle benchmark convolution data is corrected, and the actual energy consumption of fish can be dynamically adjusted in combination with changes in ambient temperature. Changes in pond temperature directly affect the metabolic rate of fish, and thus affect their demand for food. By correcting the feeding demand data, the amount and time of feed under different temperature conditions can be more accurately predicted, thereby optimizing the feeding plan and ensuring that fish can obtain appropriate nutritional support in any temperature environment to avoid waste or hunger.

[0053] Preferably, step S33 includes the following steps:

[0054] Step S331: performing temperature stratification analysis on the real-time temperature data of the fish pond to obtain stratified temperature distribution data of the fish pond;

[0055] Step S332: performing movement inhibition activity evaluation on the movement consumption estimation data between fish morphological differences according to the stratified temperature distribution data of the fish pond to obtain temperature-related movement inhibition activity data;

[0056] Step S333: performing metabolic response function fitting processing on the temperature-related motion inhibition activity data to obtain a metabolic inhibition response function;

[0057] Step S334: Identify metabolic influencing factors of the temperature-related motion inhibition activity data according to the metabolic inhibition response function to obtain temperature-related metabolic influencing factors.

[0058] The present invention performs temperature stratification analysis on the real-time temperature data of the fish pond, which can accurately reflect the temperature changes at different levels in the water body. Since the water body temperature often has vertical distribution differences, the temperature stratification analysis helps to reveal the temperature differences at different depths in the pond and their effects on fish activities. By obtaining the stratified temperature distribution data of the fish pond, accurate temperature parameters can be provided for subsequent analysis of the metabolic activity, exercise consumption, and feeding requirements of the fish. This step is the basis for a deep understanding of the influence of water body temperature and can help optimize the management of the breeding environment. According to the stratified temperature distribution data of the fish pond, the exercise inhibition activity evaluation is performed on the exercise consumption estimation data between the fish morphological differences, and the influence of different temperature environments on the movement ability of the fish school can be quantified. The increase or decrease in temperature will affect the activity level of the fish, and then affect its energy consumption. Through this evaluation, the exercise inhibition effect of temperature on different fish morphologies can be identified, providing data support for subsequent feeding management, helping to adjust the exercise and activity of the fish school, and ensuring reasonable energy consumption and metabolic needs in different environments. The temperature-related exercise inhibition activity data is subjected to metabolic response function fitting processing, which can more accurately associate temperature changes with the metabolic activities of the fish. The metabolic rate of fish is closely related to water temperature, and the metabolic response function can quantify the impact of temperature changes on fish metabolism. Through fitting processing, a more scientific metabolic response function can be obtained, and then the metabolic changes of fish at different temperatures can be predicted. This function provides a theoretical basis for subsequent energy demand estimation and feeding plan optimization, making aquaculture management more refined. According to the metabolic inhibition response function, the metabolic influencing factors of temperature-related motor inhibition activity data can be identified, and the specific influencing factors of temperature on fish metabolism can be accurately extracted. By identifying these metabolic influencing factors, the specific mechanism of temperature on fish energy consumption, feeding behavior and growth rate can be further understood. This provides a quantitative basis for adjusting feeding strategies and optimizing environmental control. By reasonably identifying metabolic influencing factors, aquaculture managers can take the best management measures for different temperature conditions, improve aquaculture efficiency, and ensure the healthy growth of fish.

[0059] A second aspect of the present invention provides a fish feed feeding amount analysis system based on machine vision, which is used to execute the fish feed feeding amount analysis method based on machine vision as described above. The fish feed feeding amount analysis system based on machine vision comprises:

[0060] The distribution density analysis module is used to collect fish school movement video images through electronic monitoring equipment to obtain fish school movement video images; perform fish school morphology difference analysis on the fish school movement video images to obtain fish school morphology difference data; perform distribution density analysis between fish morphology differences based on the fish school morphology difference data to obtain distribution density data between fish morphology differences;

[0061] A feeding demand cycle benchmark assessment module is used to mark the spatiotemporal motion trajectories of fish with different morphological shapes according to the distribution density data of fish with different morphological shapes, and obtain the spatiotemporal motion trajectory data of fish with different morphological shapes; estimate the motion consumption of fish individuals based on the spatiotemporal motion trajectory data of fish with different morphological shapes, and obtain the motion consumption estimation data of fish with different morphological shapes; perform a feeding demand cycle benchmark simulation assessment based on the motion consumption estimation data of fish with different morphological shapes, and obtain the feeding demand cycle benchmark data;

[0062] The metabolic impact correction module is used to obtain the real-time temperature data of the fish pond; identify the metabolic impact factor of the estimated exercise consumption data between the fish morphological differences according to the real-time temperature data of the fish pond, and obtain the temperature-related metabolic impact factor; based on the temperature-related metabolic impact factor, correct the feeding demand metabolic impact of the feeding demand cycle benchmark data to obtain the feeding demand metabolic impact correction data;

[0063] The feeding amount adjustment model construction module is used to construct a fish feed feeding amount adjustment model based on the feeding demand metabolic impact correction data through the policy gradient algorithm, obtain the fish feed feeding amount adjustment model, and send the fish feed feeding amount adjustment model to the cloud platform to perform fish feed feeding amount analysis.

[0064] The beneficial effects of the present invention are as follows: the present invention optimizes a traditional method for analyzing the amount of fish feed to be fed based on machine vision, solves the problem that the traditional method for analyzing the amount of fish feed to be fed based on machine vision has inaccurate analysis of the movement consumption of fish of different sizes and the influence of temperature, thereby causing a large feeding error, improves the accuracy of the analysis of the movement consumption of fish of different sizes and the influence of temperature, and reduces the error of the feeding amount. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A schematic diagram of the steps of a method for analyzing fish feed amount based on machine vision;

[0066] Figure 2 for Figure 1Detailed implementation steps of step S2 in the flowchart;

[0067] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. DETAILED DESCRIPTION

[0068] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0069] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0070] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0071] To achieve this, please refer to Figures 1 to 3 , a method for analyzing fish feed amount based on machine vision, the method comprising the following steps:

[0072] Step S1: collecting fish school motion video images through electronic monitoring equipment to obtain fish school motion video images; performing fish school morphology difference analysis on the fish school motion video images to obtain fish school morphology difference data; performing distribution density analysis between fish morphology differences based on the fish school morphology difference data to obtain distribution density data between fish morphology differences;

[0073] Step S2: marking the spatiotemporal motion trajectories of fish with different morphological shapes according to the distribution density data of fish with different morphological shapes, and obtaining the spatiotemporal motion trajectory data of fish with different morphological shapes; estimating the motion consumption of fish individuals based on the spatiotemporal motion trajectory data of fish with different morphological shapes, and obtaining the motion consumption estimation data of fish with different morphological shapes; performing a feeding demand cycle benchmark simulation evaluation based on the motion consumption estimation data of fish with different morphological shapes, and obtaining the feeding demand cycle benchmark data;

[0074] Step S3: acquiring real-time temperature data of the fish pond; identifying metabolic influencing factors of the estimated exercise consumption data between fish morphological differences according to the real-time temperature data of the fish pond to obtain temperature-related metabolic influencing factors; performing feeding demand metabolic impact correction on the feeding demand cycle benchmark data based on the temperature-related metabolic influencing factors to obtain feeding demand metabolic impact correction data;

[0075] Step S4: construct a fish feed amount adjustment model based on the feeding demand metabolic impact correction data through a policy gradient algorithm to obtain a fish feed amount adjustment model, and send the fish feed amount adjustment model to the cloud platform to perform fish feed amount analysis.

[0076] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for analyzing fish feed amount based on machine vision of the present invention. In this example, the method for analyzing fish feed amount based on machine vision includes the following steps:

[0077] Step S1: collecting fish school motion video images through electronic monitoring equipment to obtain fish school motion video images; performing fish school morphology difference analysis on the fish school motion video images to obtain fish school morphology difference data; performing distribution density analysis between fish morphology differences based on the fish school morphology difference data to obtain distribution density data between fish morphology differences;

[0078] In the embodiment of the present invention, the motion video images of the fish school are first collected by electronic monitoring equipment. These monitoring devices include high-resolution cameras and use high-frame-rate camera technology to ensure that the subtle motion changes of the fish school in the water can be captured. The collected video images of the fish school motion need to be pre-processed. First, denoising is performed to remove interference caused by external factors such as water flow and light changes to ensure the clarity of the video image. In order to improve the accuracy of image analysis, the image size needs to be unified and standardized so that the size of each frame image is consistent, which is convenient for subsequent analysis. Next, the morphological differences of the fish school are analyzed using image processing technology. This process uses image processing algorithms such as edge detection and contour extraction to focus on analyzing the body shape differences and arrangement of different individuals in the fish school. For example, the fish school area is separated from the background by image segmentation technology, and the morphological characteristics of different individuals are extracted by color analysis and contour recognition. On this basis, the morphological difference data of the fish school is obtained by calculating the morphological characteristics (such as the aspect ratio of the fish body, volume, etc.) and the difference in motion trajectory. Then, distribution density analysis is performed based on the morphological difference data of the fish school. The distribution density of fish individuals with different morphologies is analyzed by spatial statistical analysis methods (such as KDE: kernel density estimation). This method can calculate the spatial distribution of fish morphological differences based on the position information of different individuals in the fish school. For example, by calculating the concentration and distribution range of fish schools in a specific area, the distribution density data of fish morphological differences can be obtained. These data can reveal the distribution characteristics of fish individuals with different morphologies in the pond, thus providing a basis for subsequent movement trajectory and consumption estimation.

[0079] Step S2: marking the spatiotemporal motion trajectories of fish with different morphological shapes according to the distribution density data of fish with different morphological shapes, and obtaining the spatiotemporal motion trajectory data of fish with different morphological shapes; estimating the motion consumption of fish individuals based on the spatiotemporal motion trajectory data of fish with different morphological shapes, and obtaining the motion consumption estimation data of fish with different morphological shapes; performing a feeding demand cycle benchmark simulation evaluation based on the motion consumption estimation data of fish with different morphological shapes, and obtaining the feeding demand cycle benchmark data;

[0080] In the embodiment of the present invention, the spatiotemporal motion trajectory is first marked according to the distribution density data between the fish morphological differences. In order to track the motion trajectory of each fish, the optical flow method (Optical Flow) or the motion estimation algorithm is used to extract the motion path of each individual fish from the continuous fish school motion video image. The optical flow method can track the motion trajectory of the fish in real time by calculating the pixel changes between adjacent frames, and then accurately calibrate the spatiotemporal position of each fish. The distribution density data of the fish morphological differences will help determine the spatial position of each individual fish in a specific area, thereby optimizing the tracking accuracy of the motion trajectory. Subsequently, based on the spatiotemporal motion trajectory data between the fish morphological differences, the motion consumption between the individual fish is estimated. The motion consumption estimation can calculate the energy consumed by each fish during the swimming process through a fluid dynamics model. Specifically, by modeling the body size (such as length, width, etc.) of each fish, combined with factors such as swimming mode and swimming speed, an energy consumption model (such as a fluid dynamics model based on resistance) is used to estimate its energy consumption under different motion states. According to the body size and movement mode of each fish, the motion consumption is estimated to obtain the motion consumption estimation data between the fish morphological differences. Next, a baseline simulation of the feeding requirement cycle is conducted based on the estimated data of exercise expenditure between fish morphological differences. The baseline assessment uses a model based on energy balance to simulate the feed requirements of fish at different growth stages. For example, by calculating the daily energy consumption of each fish and combining factors such as its body shape and swimming pattern to estimate its energy requirements at different time periods, a baseline data of the feeding requirement cycle is obtained. This process helps to determine the feeding requirements of fish at different growth stages.

[0081] Step S3: acquiring real-time temperature data of the fish pond; identifying metabolic influencing factors of the estimated exercise consumption data between fish morphological differences according to the real-time temperature data of the fish pond to obtain temperature-related metabolic influencing factors; performing feeding demand metabolic impact correction on the feeding demand cycle benchmark data based on the temperature-related metabolic influencing factors to obtain feeding demand metabolic impact correction data;

[0082] In the embodiment of the present invention, it is first necessary to obtain the real-time temperature data of the fish pond. This data is used to monitor the temperature changes at different depths in the fish pond in real time through a temperature sensor (such as a water temperature probe, a temperature data recorder). The temperature data is collected regularly and transmitted to the central control system for processing. Next, the metabolic influencing factor is identified for the estimated data of exercise consumption between the fish morphological differences according to the real-time temperature data of the fish pond. Temperature is a key factor affecting the metabolic rate of fish, and changes in water temperature will directly affect the energy consumption of fish. In this step, a relationship model between temperature and fish metabolic rate is first established through experimental data or existing literature. The model considers the specific impact of water temperature changes on the metabolic rate of fish, and uses the acquired temperature data and the estimated data of exercise consumption between the fish morphological differences to identify the association between temperature and energy consumption, and obtain the temperature-related metabolic influencing factor. Based on the temperature-related metabolic influencing factor, the feeding demand cycle benchmark data is further corrected for the metabolic impact of feeding demand. The feeding demand benchmark data is corrected by using the temperature impact factor as an adjustment coefficient. This correction can be performed by weighted averaging or other adjustment methods to include the impact of temperature factors on fish energy consumption and feeding demand into the calculation, thereby obtaining the corrected data for the metabolic impact of feeding demand.

[0083] Step S4: construct a fish feed amount adjustment model based on the feeding demand metabolic impact correction data through a policy gradient algorithm to obtain a fish feed amount adjustment model, and send the fish feed amount adjustment model to the cloud platform to perform fish feed amount analysis.

[0084] In an embodiment of the present invention, a fish feed amount adjustment model is constructed by using a policy gradient algorithm to correct the metabolic impact of feeding requirements. The policy gradient algorithm is a reinforcement learning algorithm that optimizes the strategy by constructing a reward function. In this step, a reward function is first defined, which is directly related to the accuracy of the fish feed amount and the health status of the fish school. By analyzing the impact of excessive or insufficient feeding on the health of the fish school, a reasonable reward mechanism is set to prompt the model to adjust the feeding amount to achieve the optimal state. When using the policy gradient method, the system continuously optimizes the strategy in a simulated environment based on the correction data of the metabolic impact of feeding requirements. After each adjustment of the fish feed amount, the performance of the fish school (such as weight growth, health status, etc.) will be evaluated, and the feeding strategy will be optimized through the gradient update process, and finally a relatively stable feeding amount adjustment model will be generated. Finally, the optimized fish feed amount adjustment model will be sent to the cloud platform. The cloud platform will be responsible for real-time monitoring and execution of fish feed amount analysis, and dynamically adjust the feeding strategy according to different environmental conditions (such as temperature, fish school status, etc.) to ensure that the fish can obtain the best nutritional supplement under different conditions.

[0085] Step S1 includes the following steps:

[0086] Step S11: collecting fish school movement video images through electronic monitoring equipment to obtain fish school movement video images;

[0087] Step S12: performing image scale normalization processing on the fish school motion video image to obtain scale normalized data of the fish school motion video image;

[0088] Step S13: performing fish school morphology difference analysis on the scale-normalized data of the fish school motion video image to obtain fish school morphology difference data;

[0089] Step S14: performing distribution density analysis of differences in fish morphology on scale-normalized data of fish school motion video images according to the fish school morphology difference data to obtain distribution density data of differences in fish morphology.

[0090] In the embodiment of the present invention, the movement of the fish school is firstly captured by an electronic monitoring device installed around or above the fish pond. The electronic monitoring device used is a high-resolution camera that can provide high-definition images and has a high frame rate acquisition capability to ensure that the movement details of the fish school in the water are captured. These cameras use panoramic or fisheye lenses to ensure full coverage of the area where the fish school is located and prevent image blind spots caused by angle restrictions. The key parameters in the acquisition process include the exposure time, focal length, aperture, etc. of the camera, which are adjusted according to the lighting conditions to ensure the clarity of the image and the recognizability of the details. After the video image is acquired, the original image data is transmitted to the processing center in real time through the transmission system to ensure that the data is stored in a timely and accurate manner. The acquired video image of the fish school movement needs to be scaled to ensure the size consistency of the fish school in the image. Scale normalization is an important step in image preprocessing. It scales all objects in the image to a uniform ratio so that the image size meets the requirements of subsequent analysis. In the specific implementation process, the image resampling technology is used to adjust the image size to a fixed resolution. For example, the resolution of the original video image is standardized to 640x480 pixels to ensure that the aspect ratio of each image is consistent. When normalizing the scale, the density and dynamic changes of the fish school in the image need to be considered, and no distortion or loss of details will occur during the processing. In order to avoid the blurring phenomenon caused by image scaling, the bilinear interpolation method or the nearest neighbor interpolation method can be used to smooth the image, retain the details and eliminate the errors caused by scale changes. First, the fish school morphological difference analysis is performed on the scale-normalized fish school motion video image. This process aims to identify and analyze the differences in individual morphology in the fish school, such as the size, shape, body shape and other characteristics of the fish body. To achieve this goal, a morphological analysis method based on edge detection is adopted. The Canny edge detection algorithm is used to extract the edge information in the image, accurately divide the outline of the individual fish, and remove background interference. Then, through image segmentation technology, the region growing algorithm or the threshold segmentation method is used to automatically identify and calibrate the morphological boundary of each individual fish. For the analysis of morphological differences, statistical methods are used, such as calculating the aspect ratio, volume, shape roundness and other characteristics of each fish body. The generation of morphological difference data not only includes the quantification of fish body size, but also involves the relative position and arrangement of the fish body with other individuals. These difference data provide the basis for subsequent density analysis and motion behavior analysis. The distribution density analysis of the scale-normalized data of the fish school motion video image is carried out based on the morphological difference data of the fish school. First, based on the morphological difference data of each individual fish, a statistical analysis of the spatial distribution is carried out. To this end, the kernel density estimation (KDE) method is used to calculate the spatial distribution density of fish individuals with different morphologies in the fish school.In the specific operation, by mapping the coordinates of the position of each individual fish, a spatial coordinate system is established, the coordinate values ​​of each fish body are used as input, and the Gaussian kernel function is used to smooth these coordinates to obtain the density distribution of fish schools in different areas. This method can effectively identify the distribution characteristics of fish schools at different positions in the pond, thereby analyzing the relationship between the morphological differences of fish and their spatial positions. For example, by performing density assessment on the distribution of individual fish schools at different depths and areas in the pond, the concentrated and dispersed areas of fish morphological differences in space can be obtained, providing a scientific basis for subsequent feeding amount analysis. Through density analysis, it is possible to identify which areas have greater differences in fish morphology and which areas have higher fish density, thereby further optimizing the feeding strategy.

[0091] Step S2 includes the following steps:

[0092] Step S21: marking the spatiotemporal motion trajectories of the fish morphology differences on the fish school motion video image according to the distribution density data of the fish morphology differences, and obtaining the spatiotemporal motion trajectory data of the fish morphology differences;

[0093] Step S22: based on the spatiotemporal motion trajectory data of the fish morphology differences, the motion consumption estimation data of the fish morphology differences are performed on the fish morphology difference data to obtain the motion consumption estimation data of the fish morphology differences;

[0094] Step S23: performing consumption cluster regression analysis on the motion consumption estimation data between fish morphology differences to obtain motion consumption cluster regression data;

[0095] Step S24: performing a feeding demand cycle benchmark simulation evaluation on the fish school morphology difference data according to the movement consumption cluster regression data and the distribution density data between the fish morphology differences to obtain the feeding demand cycle benchmark data.

[0096] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0097] Step S21: marking the spatiotemporal motion trajectories of the fish morphology differences on the fish school motion video image according to the distribution density data of the fish morphology differences, and obtaining the spatiotemporal motion trajectory data of the fish morphology differences;

[0098] In the embodiment of the present invention, firstly, the spatiotemporal motion trajectory of the fish morphological differences is marked for the fish school motion video image according to the distribution density data of the fish morphological differences. The core goal of this process is to accurately track the position change of each individual fish in the water area, and then identify its motion trajectory. To achieve this goal, firstly, the position of each fish body in the fish school and the morphological differences between them are determined according to the distribution density data obtained in step S14. By adopting the optical flow method (OpticalFlow) technology, the continuous fish school motion video image is processed frame by frame to extract the fish body position information in each frame image. The optical flow method is based on the motion estimation of the image pixel level, and the motion trajectory of the fish body in the water area is determined by analyzing the displacement between pixels. Next, the motion trajectory of each individual fish is marked to identify the position and motion direction at different time points. In order to improve the accuracy of the trajectory marking, the target detection and tracking algorithm, such as the Kalman filter, is used to combine the known morphological characteristics and position changes of the fish body to perform trajectory smoothing processing to reduce errors. In this way, after processing, the complete motion trajectory of each individual fish from the starting point to the target point in the water area can be obtained, and its spatiotemporal characteristics that change with time can be marked. These spatiotemporal motion trajectory data provide key basis for subsequent motion expenditure estimation.

[0099] Step S22: based on the spatiotemporal motion trajectory data of the fish morphology differences, the motion consumption estimation data of the fish morphology differences are performed on the fish morphology difference data to obtain the motion consumption estimation data of the fish morphology differences;

[0100] In the embodiment of the present invention, based on the spatiotemporal motion trajectory data between the fish morphological differences, the motion consumption between the fish individuals is estimated for the fish morphological difference data. First, the body length, weight, morphological proportion and other information of each fish individual provided in the fish morphological difference data are used, combined with its spatiotemporal motion trajectory data, and the motion energy consumption of each fish individual is estimated through a dynamic model. This process is modeled by classical physics principles, that is, the motion consumption of the fish body is calculated based on parameters such as the mass, acceleration, speed and water flow resistance of the fish body. First, according to the body length and weight data of the fish, combined with the interaction coefficient between the fish body and the water flow, the water motion resistance of each fish individual is calculated. Then, based on the spatiotemporal motion trajectory data, the average speed and acceleration of each fish individual per unit time are calculated, and the power consumed by the fish individual during the motion process can be estimated, thereby obtaining the motion consumption data. It should be noted that the consumption of the fish individual is not only related to its speed, but also closely related to its morphological characteristics such as body shape and posture. Therefore, in this step, the morphological difference data is combined with the spatiotemporal trajectory data to provide a more accurate calculation basis for consumption estimation.

[0101] Step S23: performing consumption cluster regression analysis on the motion consumption estimation data between fish morphology differences to obtain motion consumption cluster regression data;

[0102] In an embodiment of the present invention, the motion consumption estimation data between fish morphological differences are subjected to consumption cluster regression analysis. The purpose of this step is to reveal the consumption patterns of different types of fish individuals by analyzing the motion consumption data of different fish individuals, so as to provide support for the feeding demand assessment. First, the motion consumption estimation data of all fish individuals are grouped according to the morphological differences of the fish individuals, and the fish individuals are divided into several categories by cluster analysis methods. Commonly used clustering algorithms such as K-means or DBSCAN algorithms can classify fish groups with similar motion consumption characteristics based on the similarity of motion consumption estimation data. After clustering is completed, regression analysis is performed based on the data in each category. Through linear regression or nonlinear regression models, combined with parameters such as the body shape, movement speed, and motion consumption of fish individuals, the motion consumption trends of fish groups of each category are fitted. Through the regression model, the motion consumption law of each type of fish individual under different environmental conditions can be obtained, which provides a quantitative basis for the subsequent feeding amount adjustment. For example, through regression analysis, a curve of energy consumption of fish individuals of different body shapes over time can be obtained, which provides a reference for the optimization of feeding strategies.

[0103] Step S24: performing a feeding demand cycle benchmark simulation evaluation on the fish school morphology difference data according to the movement consumption cluster regression data and the distribution density data between the fish morphology differences to obtain the feeding demand cycle benchmark data.

[0104] In the embodiment of the present invention, the fish morphology difference data is subjected to a baseline simulation evaluation of the feeding demand cycle based on the motion consumption cluster regression data and the distribution density data between the fish morphology differences. The core goal of this process is to evaluate the feeding demand of the fish swarm in different time periods based on the motion consumption data and distribution density data of the fish swarm. First, according to the aforementioned cluster regression results, the energy consumption demand curve of different types of fish individuals is obtained. At the same time, the distribution of the fish swarm in different areas of the water area is determined in combination with the distribution density data between the fish morphology differences obtained in step S14. Based on these data, a baseline simulation model is established to evaluate the feeding demand of the fish swarm in a specific time period. Specifically, first, the total energy demand in different areas and different time periods is calculated based on the fish swarm density distribution and the consumption curves of various types of fish individuals. Then, based on the density and consumption data of the fish swarm in each area, a simulation evaluation model is established to predict the feeding amount required by the fish swarm in different time periods. This simulation process takes into account the impact of environmental factors (such as water temperature, oxygen content, etc.) on the fish swarm, and periodically evaluates the analysis of historical data to ensure the accuracy of the feeding demand cycle simulation. Finally, the baseline data of feeding demand cycle is obtained, which provides a reliable basis for subsequent feeding amount analysis and adjustment.

[0105] Step S22 includes the following steps:

[0106] Step S221: Calculate the size of each fish based on the fish morphology difference data to obtain morphology, body size and data;

[0107] Step S222: quantifying the fluid resistance of each fish according to the fish morphology difference data based on the morphology and body size data to obtain the fish morphology fluid resistance quantification data;

[0108] Step S223: extracting the swimming mode of each fish from the fish school motion video image based on the spatiotemporal motion trajectory data between the fish morphological differences, and obtaining the fish swimming mode data;

[0109] Step S224: performing resistance compensation between different swimming modes on the quantified data of the fluid resistance of the fish morphology according to the data of the swimming mode of the fish, and obtaining the fluid resistance compensation data of the fish morphology-swimming mode;

[0110] Step S225: Estimate the motion consumption of individual fish based on the spatiotemporal motion trajectory data of different fish morphologies according to the fish morphology-swimming mode fluid resistance compensation data, and obtain motion consumption estimation data of different fish morphologies.

[0111] In the embodiment of the present invention, the body size of each fish is calculated for the fish shoal morphological difference data to obtain the morphological body size data. This step accurately extracts the length, width, height and other geometric size data of the fish body by analyzing each fish in the fish shoal motion video image. During the implementation process, the fish shoal motion video image is first segmented and the fish body contour is extracted, and the edge detection technology, such as Canny edge detection or Sobel operator, is used to separate the contour of each fish from the video image. On this basis, the contour of each fish is fitted by a geometric method to calculate its length, width and height. According to the relationship between the image resolution and the size of the fish body, these size data are converted into actual physical dimensions. In addition, considering that the size of the fish body changes at different viewing angles, the fish body is three-dimensionally modeled by three-dimensional reconstruction technology. Through video images from multiple viewing angles, the three-dimensional size data of each fish is obtained by using a geometric reconstruction algorithm such as stereo vision reconstruction (Stereo Vision Reconstruction), which further improves the accuracy of size estimation. Finally, the obtained morphological body size data includes parameters such as length, width, height, volume and the like of each fish, and provides a basis for the subsequent calculation of fluid resistance. Based on the morphological and body size data obtained in step S221, the fluid resistance of each fish is further quantified. When the fish swims in the water, the friction and resistance between the fish and the water have an important impact on its movement consumption. In order to quantify this impact, the fluid resistance of each fish is first calculated based on the size data of the fish using the principle of fluid dynamics. Specifically, the length, width and height data of the fish are combined with the morphological characteristics of the fish in the water, and the resistance is calculated using Stokes' Law and Reynolds Number in fluid mechanics. The fluid resistance coefficient of the fish is calculated based on the surface roughness and volume characteristics of the fish, and then the Reynolds number model is used to correct the fluid resistance based on the movement speed of the fish and the relative speed of the water flow. In the case of low Reynolds numbers, the fluid resistance is mainly dominated by viscous friction, while in the case of high Reynolds numbers, the fluid resistance is more dominated by inertial force. By comprehensively considering the size, movement speed and water flow environment of the fish, the fluid resistance data of each fish swimming in the water is calculated. Finally, the quantitative data of fluid resistance of each fish is obtained, including the calculation results of drag coefficient, fluid friction, inertial resistance and other aspects. These data provide key parameters for the estimation of subsequent swimming mode and motion consumption. Based on the spatiotemporal motion trajectory data between fish morphological differences, the swimming mode of each fish is extracted from the video images of the school of fish movement to obtain the fish swimming mode data. The swimming mode refers to the specific movement pattern of the fish when moving in the water, such as uniform swimming, accelerated swimming, sharp turns, etc. In order to extract the swimming mode of each fish, the movement trajectory of the school of fish is first analyzed in detail to identify the speed, acceleration and direction changes of each fish.By analyzing the continuous frames of the spatiotemporal motion trajectory data of the school of fish, the instantaneous speed and acceleration of each fish at each time point are calculated. On this basis, the kinematic method (such as the uniform acceleration motion model) is used to classify the movement mode of the fish. The swimming mode of the fish can be identified by calculating the speed change rate (acceleration) and the direction change (angle change) of the fish. The swimming mode data of the fish are extracted by the pattern recognition algorithm combined with the characteristics of the fish's motion trajectory. Commonly used pattern recognition methods include rule-based classification methods and data-driven machine learning methods (such as support vector machine SVM or K-means clustering). According to the fish swimming mode data, the resistance compensation between different swimming modes is performed on the quantitative data of the fluid resistance of the fish morphology. Different swimming modes will have different effects on the fluid resistance according to the swimming mode of the fish. For example, when swimming rapidly or turning suddenly, the fish will encounter greater resistance. In order to accurately estimate this difference, it is necessary to compensate for the resistance of each swimming mode. First, the swimming mode of each fish is classified, and different fluid resistance correction coefficients are applied according to different swimming modes. For example, for fish swimming at a constant speed, the fluid resistance compensation coefficient is low; while for fish accelerating rapidly, the compensation coefficient increases accordingly. The fluid resistance compensation factor under each swimming mode is obtained through experimental or simulation data, and these factors are applied to the fluid resistance data obtained in step S222. The average speed, acceleration and steering angle of individual fish under different swimming modes can be calculated, and the fluid resistance under each swimming mode can be corrected using the dynamic correction model in fluid mechanics. In this way, accurate resistance compensation data for each fish under different swimming modes is obtained. Based on the fluid resistance compensation data of fish morphology-swimming mode, the spatiotemporal motion trajectory data between fish morphological differences are used to estimate the motion consumption of individual fish. The purpose of this process is to combine the fluid resistance compensation data and motion trajectory data of individual fish to accurately estimate the motion energy consumption of each individual fish. According to the fluid resistance compensation data of fish morphology-swimming mode obtained in step S224, the motion resistance of the fish is corrected. Combined with the motion trajectory data of individual fish, the total resistance that the fish body needs to overcome when swimming in water is calculated. Then, the instantaneous power consumption of the fish is calculated using the power calculation formula in physics. By integrating the power in each time period, the motion consumption of individual fish is obtained. In the specific implementation, the motion trajectory of each individual fish is first discretized and decomposed into multiple time periods, and the instantaneous speed and corresponding resistance of the fish body are calculated in each time period. According to the speed of the fish body and the fluid resistance, the motion consumption in each time period is calculated, and finally the total motion consumption data of each fish is obtained by summing the motion consumption of all time periods. These data provide an accurate basis for the subsequent feeding amount estimation.

[0112] Compensating the resistance between different swimming modes for the quantitative data of the fluid resistance of the fish morphology according to the data of the swimming mode of the fish includes the following steps:

[0113] According to the fish swimming mode data, key posture nodes between different swimming modes are analyzed to obtain swimming mode posture node analysis data;

[0114] Based on the swimming mode posture node analysis data, fluid dynamic field simulation is performed to obtain posture node associated fluid dynamic field data;

[0115] The mean difference of surface pressure distribution of the attitude node is calculated for the fluid dynamic field data associated with the attitude node to obtain the mean difference data of surface pressure distribution of the node;

[0116] According to the mean difference data of node surface pressure distribution, the resistance compensation between different swimming modes is performed on the quantitative data of fish morphology fluid resistance to obtain the fish morphology-swimming mode fluid resistance compensation data.

[0117] In an embodiment of the present invention, the motion mode of each fish is first extracted from the fish swimming mode data, including uniform swimming, accelerated swimming, turning, etc. In order to perform posture analysis, it is necessary to analyze the swimming posture of each fish frame by frame to extract the position and change of each key node of the fish body. Key nodes refer to important positions on the fish body that can represent the change of the fish body posture, such as the head, tail, dorsal fin, pelvic fin, etc. In specific implementation, the posture of each fish in different swimming modes is geometrically modeled by kinematic methods. The position, speed and relative angle change of each key node are calibrated using vector analysis technology. By extracting the real-time position of the key node in each swimming cycle, the posture change data of each fish is obtained, and finally the swimming mode posture node analysis data is formed. Based on the swimming mode posture node analysis data obtained in step S221, fluid dynamics simulation is performed to calculate the fluid dynamic field of the fish body in different swimming modes. First, using the basic principles of fluid dynamics, the Navier-Stokes equation is applied to model the interaction between water flow and fish body. In the model, the posture data of each fish is used as input to simulate the fluid characteristics such as eddies and turbulence generated by the fish body under different swimming modes. By calculating the pressure change of the fluid flowing through the surface of the fish body, the relationship between the posture node and the fluid dynamic field is obtained. The finite volume method (FVM) is used for discretization solution to obtain the fluid dynamic field data, and the flow characteristics, pressure changes and water velocity distribution of the fluid under different postures are calibrated. Finally, the posture node associated fluid dynamic field data is formed. Based on the posture node associated fluid dynamic field data obtained in step S222, the pressure distribution on the surface of each posture node is calculated. The specific operation is to first mesh the surface of the fish body, and then calculate the fluid pressure on each grid unit based on the simulation results of the fluid dynamic field. On this basis, the mean difference of the pressure distribution on the surface of each posture node is calculated, that is, the pressure difference around each posture node is analyzed by statistical methods to obtain the mean square error value to measure the uniformity of the pressure distribution. By averaging the mean difference of the pressure of each posture node, the mean difference data of the pressure distribution on the node surface is obtained. The resistance compensation is performed on the quantitative data of the fluid resistance of the fish morphology by using the mean difference data of the pressure distribution of the node surface obtained in step S223. The specific operation is to correct the mean difference of the pressure distribution of the posture node under each swimming mode, and calculate the change in the fluid resistance of the fish body under each swimming mode. According to the Bernoulli principle in physics and the relevant formulas of fluid dynamics, combined with the pressure distribution data on the surface of the fish body, the fluid resistance coefficient under each swimming mode is adjusted and applied to the calculation of the fluid resistance of the fish body. Through this compensation process, the actual resistance of the fish body under different swimming modes is corrected, making the resistance compensation data more accurate. Finally, the fluid resistance compensation data related to the fish morphology and swimming mode is obtained, providing accurate data for subsequent motion consumption estimation.

[0118] Estimating the motion consumption of the spatiotemporal motion trajectory data between the fish morphology differences according to the fluid resistance compensation data of the fish morphology-swimming mode includes the following steps:

[0119] According to the fish morphology-swimming mode fluid resistance compensation data, the spatiotemporal motion trajectory data between the fish morphology differences are extracted for the speed changes between the resistance states of different swimming modes, and the speed change data between the resistance states of different swimming modes are obtained;

[0120] Based on the speed change data between different swimming mode resistance states and the fish morphology-swimming mode fluid resistance compensation data, multi-directional force output balance decomposition is performed to obtain multi-directional force output decomposition data;

[0121] Perform output torque matching on the multi-directional force output decomposition data to obtain multi-directional output torque matching data;

[0122] The motion consumption of the spatiotemporal motion trajectory data of fish with different morphologies is estimated based on the multi-directional force output decomposition data and the multi-directional output torque matching data, so as to obtain the motion consumption estimation data of fish with different morphologies.

[0123] In the embodiment of the present invention, firstly, the speed change under different swimming modes is extracted according to the fluid resistance compensation data of the fish morphology-swimming mode, combined with the spatiotemporal motion trajectory data of the fish school. In specific implementation, by analyzing the spatiotemporal motion trajectory of the fish school under different swimming modes (such as uniform speed, acceleration, deceleration, turning, etc.), the speed of the fish body at each time step is calculated, and the speed change amount is obtained by combining the fluid resistance compensation data. Specifically, by comparing the relationship between the speed and fluid resistance of the fish body under different swimming modes, the speed change data corresponding to each swimming mode is calculated. This process requires accurate extraction of data of each point in the motion trajectory, and by calculating the speed change caused by the fluid resistance of the fish body during the movement, the speed change data under the resistance state of different swimming modes is obtained. According to the speed change data between the resistance states of different swimming modes and the fluid resistance compensation data of the fish morphology-swimming mode, the multi-directional force output balance is decomposed by a mechanical method. First, the forces in all directions received by the fish body during the swimming process are analyzed, taking into account the resistance of the water flow, the inertia force of the fish body itself, and the driving force of the fins. The force is decomposed by using the Newton-Euler equation to calculate the relationship between the force in each direction and the motion state of the fish. By decomposing these forces, the force balance in each direction is obtained, so as to obtain the motion trajectory and force condition of the fish in the water flow. Through this balanced decomposition, the multi-directional force output decomposition data under each swimming mode is obtained, covering the main forces on the fish in different directions. The multi-directional force output decomposition data obtained in step S222 is matched with torque. The specific operation is to calculate the point of action of the force in each direction, that is, the "fulcrum" of the force, so as to obtain the torque in each direction. By comparing the multi-directional force output data under different swimming modes, the torque values ​​in each direction are matched. By applying the principle of torque balance, it is ensured that the torque in each direction can be balanced, so that the force output of the fish during movement is in a stable state. During the torque matching process, by adjusting the torque, the fish can maintain a certain stability during the movement. The torque matching data further provides accurate basic data for the estimation of exercise consumption. According to the multi-directional force output decomposition data and multi-directional output torque matching data obtained in steps S223 and S224, the fish movement consumption is estimated. The specific operation is to combine the force conditions of the fish body under different swimming modes with the torque matching data, and use the principle of conservation of dynamic energy to calculate the energy consumed by the fish body during the swimming process. According to the effects of internal and external forces generated by the fish body during the movement, combined with speed changes, fluid resistance and different movement modes, the movement energy consumption of each fish under different swimming modes is calculated. By statistics on these movement consumption data, the estimated movement consumption data between fish morphological differences are obtained. Through this process, the energy consumption of the fish body under different swimming modes can be accurately evaluated, providing reliable data support for the accurate prediction of subsequent feeding needs.

[0124] Step S24 includes the following steps:

[0125] Step S241: performing theoretical growth cycle analysis of fish morphology difference data according to the movement consumption cluster regression data, and obtaining theoretical growth cycle of fish morphology difference;

[0126] Step S242: performing a periodic fluctuation analysis of satiation intake between different growth stages of the theoretical growth cycle of fish morphological differences based on the exercise consumption cluster regression data to obtain periodic fluctuation data of satiation intake;

[0127] Step S243: defining the growth base increments between different fish morphological differences for the theoretical growth period of fish morphological differences according to the distribution density data between different fish morphological differences, and obtaining the growth base increment data between different fish morphological differences;

[0128] Step S244: Perform a feeding requirement cycle benchmark simulation evaluation based on the growth base increment data and satiation intake cycle fluctuation data of different fish morphological differences to obtain feeding requirement cycle benchmark data.

[0129] In an embodiment of the present invention, the theoretical analysis of the growth cycle of fish is performed by combining the cluster regression data of motion consumption with the data of fish morphological differences. In this step, the growth cycle needs to be calculated by measuring the relationship between the energy consumption of the fish (i.e., the motion consumption data of the fish) and its morphological characteristics, combined with the growth dynamics model (such as the exponential growth model or the logistic growth model). The specific method is: using the regression analysis results, the growth curve of each fish body is combined with its corresponding energy consumption to obtain the theoretical growth cycle of the fish body. The morphological differences of the fish body (such as the size and shape of the fish body) will affect its growth rate, thereby determining its growth cycle. Through this analysis, the corresponding theoretical growth cycle can be determined for fish schools of different morphologies. This step ensures that the estimation of the growth cycle accurately reflects the actual growth of the fish school. The theoretical growth cycle data of the fish morphological differences obtained in step S241 is combined with the cluster regression data of motion consumption to perform the periodic fluctuation analysis of the satiation intake of fish. The specific operation is to divide the fish body into several growth stages according to its theoretical growth cycle, and calculate the periodic fluctuation of the satiation intake of fish in each stage by analyzing the fluctuation of energy demand in each growth stage. This process is based on the cyclical changes in fish feeding behavior, such as the high intake demand of fish in the early growth stage and the stable demand in the later growth stage. By capturing the changing law of fish feeding behavior, the feeding fluctuation data of fish in different growth stages are obtained, and the cyclical change trend of satiation intake is evaluated. These fluctuation data provide a reliable cyclical change model for subsequent feeding amount prediction. In this process, the growth base increment of fish of different forms in their respective growth cycles is defined according to the distribution density data between fish morphological differences. First, by statistically analyzing the distribution density of fish of different forms in the fish school, the base increment (such as the increase in weight or body length) required by fish of each morphological type during the growth process is determined. On this basis, the growth rate of each fish morphology at different growth stages is calculated through the growth increment model, such as the weighted average method or the cumulative increment method, so as to obtain the growth base increment data of each fish morphology. These data reflect the growth rate of fish of different forms in a specific growth cycle and provide basic data support for the subsequent calculation of feeding demand cycle. In this step, the growth base increment data in step S243 is combined with the satiation intake cycle fluctuation data in step S242, and a baseline simulation evaluation of the feeding demand cycle is performed by establishing a periodic simulation model of feeding demand. First, the weight gain requirements of the fish at each stage are determined based on the growth base increment data of the morphological differences of each fish. Combined with the feeding fluctuation data, the energy intake requirements of each fish at different stages are predicted, and the overall feeding requirements of the fish are estimated through the simulation model. In this process, mathematical optimization methods (such as linear programming, least squares method, etc.) are required to fine-tune the feeding requirements of different fish morphologies at each stage to ensure accurate prediction of the feeding cycle.The final feeding demand cycle benchmark data reflects the average feeding demand of fish in different time periods, providing a quantitative basis for the calculation and control of actual feeding amount.

[0130] Step S3 includes:

[0131] Step S31: performing convolution calculation on the feeding demand cycle benchmark data to obtain feeding demand cycle benchmark convolution data;

[0132] Step S32: Obtaining real-time temperature data of the fish pond;

[0133] Step S33: identifying metabolic influencing factors of the estimated exercise consumption data of fish morphological differences according to the real-time temperature data of the fish pond, and obtaining temperature-related metabolic influencing factors;

[0134] Step S34: performing feeding demand metabolic impact correction on the feeding demand cycle benchmark convolution data based on the temperature-related metabolic impact factor to obtain feeding demand metabolic impact correction data.

[0135] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0136] Step S31: performing convolution calculation on the feeding demand cycle benchmark data to obtain feeding demand cycle benchmark convolution data;

[0137] In an embodiment of the present invention, a convolution calculation is performed on the existing feeding demand cycle benchmark data to smooth the feeding demand data and extract the characteristics of its periodic changes. The specific operation is: first determine the size of the convolution kernel, which is set according to the time resolution of the feeding demand data and the frequency of the periodic changes of the data. The convolution operation generates smoothed feeding demand cycle benchmark convolution data by gradually superimposing the feeding demand data with the convolution kernel. In this calculation, the convolution kernel selects a Gaussian filter kernel or a weighted average kernel, the purpose of which is to reduce the interference of outliers on the data and highlight the periodic change trend in the data. After the calculation is completed, the obtained feeding demand cycle benchmark convolution data reflects the comprehensive feeding demand trend of the fish school in different time periods, has a smooth curve, and is convenient for subsequent adjustment and correction.

[0138] Step S32: Obtaining real-time temperature data of the fish pond;

[0139] In an embodiment of the present invention, a temperature sensor (such as a water temperature sensor or a thermocouple) is used to monitor the temperature change of the water in the fish pond, and the water temperature data is recorded regularly. The temperature sensors are deployed at multiple locations in the fish pond to ensure that the temperature data of different areas of the fish pond are obtained. At regular intervals (for example, every minute or every hour), the temperature data is automatically updated and stored in the data storage system. To ensure the accuracy of the data, the sensor needs to be calibrated regularly, and the temperature data of all data collection points will be denoised to avoid the error caused by the temperature sensor.

[0140] Step S33: identifying metabolic influencing factors of the estimated exercise consumption data of fish morphological differences according to the real-time temperature data of the fish pond, and obtaining temperature-related metabolic influencing factors;

[0141] In an embodiment of the present invention, the effect of temperature on fish metabolism is identified by combining real-time temperature data with estimated data of fish exercise consumption. First, correlation analysis is performed using the temperature data in the fish pond (step S32) and the estimated data of exercise consumption between fish morphological differences. Through the metabolic model, the effect of temperature on fish metabolism is included in the calculation, and Allen's Rule in biology is usually used, that is, every time the temperature rises or falls within a certain range, the metabolic rate of the fish will change accordingly. The specific calculation method is: by analyzing the exercise consumption of the fish at different temperatures, a correction coefficient of temperature change on the metabolism of the fish is obtained. This coefficient is the temperature-related metabolic influence factor. The metabolic influence factor can be estimated by a linear regression model or an empirical formula, and the factor is dynamically adjusted as the temperature changes, reflecting the effect of temperature on the exercise consumption of the fish.

[0142] Step S34: performing feeding demand metabolic impact correction on the feeding demand cycle benchmark convolution data based on the temperature-related metabolic impact factor to obtain feeding demand metabolic impact correction data.

[0143] In the embodiment of the present invention, the law of the influence of water temperature on the metabolism of fish is established based on the acquired real-time temperature data of the fish pond (step S32) and the estimated data of exercise consumption between the fish morphological differences (step S33). In this process, the relationship between the actual temperature data and the metabolic activity of the fish is determined by the physiological principles obtained by experiments or literature. Studies have shown that the metabolic rate of fish increases with the increase of water temperature, and the metabolic rate of fish slows down when the temperature is low. Based on this physiological characteristic, the changes in the metabolic demand of fish in different temperature ranges are determined. Using the relationship between the above temperature and metabolic rate, a set of temperature-metabolism influencing factors is generated. The metabolic factor corresponding to each temperature range reflects the changes in the food intake of fish at this temperature. For example, when the water temperature is low, the metabolism of fish slows down and the demand for food intake decreases; when the water temperature is high, the metabolism of fish accelerates and the demand for food increases. The temperature factor is verified by historical data or experimental data, and is associated with factors such as the exercise consumption and body shape of the fish to form a reliable set of metabolic influencing factors. After the metabolic influencing factor is obtained, it is corrected based on the existing feeding demand cycle benchmark convolution data (step S31). Whenever new water temperature data is obtained, the metabolic impact factor that matches the current temperature is searched in real time and applied to the feeding demand data. The correction process is essentially to adjust the original feeding demand according to the current water temperature conditions, so that the adjusted feeding demand is more in line with the energy consumption demand of the actual fish at that temperature. For example, when the water temperature is low, the feeding amount is reduced; when the water temperature is high, the feeding amount is increased to ensure that the fish can maintain normal growth and activity levels. Feeding demand correction is a dynamic adjustment process. Every time new temperature data is obtained, the system automatically searches for the metabolic impact factor corresponding to the current water temperature and adjusts the feeding demand cycle baseline convolution data according to this factor. The system obtains real-time temperature data at regular intervals, combines the historical temperature-metabolic factor relationship, and adjusts the feeding amount in real time, thereby ensuring that the feeding amount of the fish under different environmental conditions is more accurate. After the correction of the temperature-related metabolic impact factor, the feeding demand cycle data will be adjusted to the new feeding demand data. These data are no longer based only on the basic needs of the fish, but also take into account the impact of water temperature on the metabolic activity of the fish, so they are closer to the actual feeding needs. Ultimately, these revised feeding demand data will serve as the basis for actual feeding operations and help formulate reasonable feeding strategies.

[0144] Step S33 includes the following steps:

[0145] Step S331: performing temperature stratification analysis on the real-time temperature data of the fish pond to obtain stratified temperature distribution data of the fish pond;

[0146] Step S332: performing movement inhibition activity evaluation on the movement consumption estimation data between fish morphological differences according to the stratified temperature distribution data of the fish pond to obtain temperature-related movement inhibition activity data;

[0147] Step S333: performing metabolic response function fitting processing on the temperature-related motion inhibition activity data to obtain a metabolic inhibition response function;

[0148] Step S334: Identify metabolic influencing factors of the temperature-related motion inhibition activity data according to the metabolic inhibition response function to obtain temperature-related metabolic influencing factors.

[0149] In an embodiment of the present invention, firstly, the real-time temperature data of the fish pond is obtained, and the data source is the temperature sensor of each layer of the water body in the fish pond or a real-time monitoring system. By analyzing the temperature data at different water depths, the temperature of the fish pond is divided into multiple levels according to the water depth. For example, the fish pond can be divided into surface water temperature, bottom water temperature and middle water temperature. In actual operation, multiple temperature levels can be set according to the water depth range, and each temperature layer represents the temperature state of the water body at different depths. The stratified temperature distribution data of the fish pond is the temperature state data of each layer of water body, and these data are used to describe the temperature change law at different water depths in the fish pond. In order to improve the accuracy of temperature stratification, a temperature sensor is used to feed back the data to the system in real time, and the data is updated according to a certain time interval to ensure the timeliness of the temperature distribution. The stratified temperature distribution data obtained in step S331 is used to further analyze the estimated data of the movement consumption of the fish school. In this analysis process, the estimated data of the movement consumption between the fish morphological differences is used as the basic data, and combined with the stratified temperature data, the inhibitory effect of temperature on the movement activity of the fish school is derived. Studies have shown that changes in water temperature will directly affect the activity level of fish. At lower temperatures, the metabolism and exercise capacity of fish are reduced; at higher temperatures, the metabolism of fish is accelerated, which will lead to excessive physical exhaustion. Through this analysis, the temperature-related movement inhibition activity data obtained reflects the impact of different temperature levels on the activity of fish, and thus affects the level of their energy consumption. Next, the metabolic response function is fitted for the temperature-related movement inhibition activity data. According to biological research, the metabolic activity of fish is significantly affected by temperature changes, especially when the temperature changes are large, the movement capacity and metabolic rate of fish will fluctuate significantly. Through fitting processing, using known biological principles and experimental data, a metabolic response function is constructed, which can accurately reflect the impact of temperature on the metabolism and exercise consumption of fish. In the fitting process, a suitable regression model (such as nonlinear regression, least squares fitting, etc.) is selected to ensure that the metabolic response function can accurately describe the relationship between the exercise activity and energy consumption of fish under different temperature conditions. The fitting result will form a mathematical model between temperature and movement inhibition activity, providing a theoretical basis for the subsequent identification of metabolic influencing factors. Finally, according to the metabolic inhibition response function obtained in step S333, the temperature-related metabolic influencing factors are identified. In this process, the metabolic response function is used to estimate the motor inhibition activity at different temperatures, and the factors affecting the metabolism of fish schools are identified. Specifically, based on the metabolic response function, combined with the exercise consumption data of fish schools under different temperature environments, the main factors affecting the metabolism of fish schools are identified. For example, within a certain temperature range, the metabolism of fish schools is significantly affected, so within this temperature range, the metabolic impact factor is larger. Conversely, when the temperature fluctuation is small, the change in the metabolic impact factor is relatively gentle.Through this identification process, the temperature-related metabolic influencing factors finally obtained will provide the comprehensive impact of temperature on fish metabolism and provide basic data for subsequent feeding requirements and energy consumption analysis.

[0150] The present invention also provides a fish feed feeding amount analysis system based on machine vision, which is used to execute the fish feed feeding amount analysis method based on machine vision as described above. The fish feed feeding amount analysis system based on machine vision comprises:

[0151] The distribution density analysis module is used to collect fish school movement video images through electronic monitoring equipment to obtain fish school movement video images; perform fish school morphology difference analysis on the fish school movement video images to obtain fish school morphology difference data; perform distribution density analysis between fish morphology differences based on the fish school morphology difference data to obtain distribution density data between fish morphology differences;

[0152] A feeding demand cycle benchmark assessment module is used to mark the spatiotemporal motion trajectories of fish with different morphological shapes according to the distribution density data of fish with different morphological shapes, and obtain the spatiotemporal motion trajectory data of fish with different morphological shapes; estimate the motion consumption of fish individuals based on the spatiotemporal motion trajectory data of fish with different morphological shapes, and obtain the motion consumption estimation data of fish with different morphological shapes; perform a feeding demand cycle benchmark simulation assessment based on the motion consumption estimation data of fish with different morphological shapes, and obtain the feeding demand cycle benchmark data;

[0153] The metabolic impact correction module is used to obtain the real-time temperature data of the fish pond; identify the metabolic impact factor of the estimated exercise consumption data between the fish morphological differences according to the real-time temperature data of the fish pond, and obtain the temperature-related metabolic impact factor; based on the temperature-related metabolic impact factor, correct the feeding demand metabolic impact of the feeding demand cycle benchmark data to obtain the feeding demand metabolic impact correction data;

[0154] The feeding amount adjustment model construction module is used to construct a fish feed feeding amount adjustment model based on the feeding demand metabolic impact correction data through the policy gradient algorithm, obtain the fish feed feeding amount adjustment model, and send the fish feed feeding amount adjustment model to the cloud platform to perform fish feed feeding amount analysis.

[0155] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0156] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for analyzing fish feed amount based on machine vision, characterized in that: The following steps are involved: Step S1: collecting fish school motion video images through electronic monitoring equipment to obtain fish school motion video images; performing fish school morphology difference analysis on the fish school motion video images to obtain fish school morphology difference data; performing distribution density analysis between fish morphology differences based on the fish school morphology difference data to obtain distribution density data between fish morphology differences; Step S2: marking the spatiotemporal motion trajectory of the fish morphology differences according to the distribution density data of the fish morphology differences, and obtaining the spatiotemporal motion trajectory data of the fish morphology differences; Based on the spatiotemporal motion trajectory data of fish with different morphological shapes, the motion consumption of fish individuals is estimated, and the motion consumption estimation data of fish with different morphological shapes are obtained; Based on the estimated data of movement consumption between fish morphological differences, a feeding demand cycle benchmark simulation evaluation is performed to obtain the feeding demand cycle benchmark data; Step S3: acquiring real-time temperature data of the fish pond; identifying metabolic influencing factors of the estimated exercise consumption data of fish morphological differences according to the real-time temperature data of the fish pond, and obtaining temperature-related metabolic influencing factors; Based on the temperature-related metabolic impact factor, the feeding demand period baseline data is corrected for the metabolic impact of the feeding demand to obtain the feeding demand metabolic impact correction data; Step S4: construct a fish feed amount adjustment model based on the feeding demand metabolic impact correction data through a policy gradient algorithm to obtain a fish feed amount adjustment model, and send the fish feed amount adjustment model to the cloud platform to perform fish feed amount analysis.

2. The method for analyzing fish feed amount based on machine vision according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting fish school movement video images through electronic monitoring equipment to obtain fish school movement video images; Step S12: performing image scale normalization processing on the fish school motion video image to obtain scale normalized data of the fish school motion video image; Step S13: performing fish school morphology difference analysis on the scale-normalized data of the fish school motion video image to obtain fish school morphology difference data; Step S14: performing distribution density analysis of differences in fish morphology on scale-normalized data of fish school motion video images according to the fish school morphology difference data to obtain distribution density data of differences in fish morphology.

3. The method for analyzing fish feed amount based on machine vision according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: marking the spatiotemporal motion trajectories of the fish morphology differences on the fish school motion video image according to the distribution density data of the fish morphology differences, and obtaining the spatiotemporal motion trajectory data of the fish morphology differences; Step S22: based on the spatiotemporal motion trajectory data of the fish morphology differences, the motion consumption estimation data of the fish morphology differences are performed on the fish morphology difference data to obtain the motion consumption estimation data of the fish morphology differences; Step S23: performing consumption cluster regression analysis on the motion consumption estimation data between fish morphology differences to obtain motion consumption cluster regression data; Step S24: performing a feeding demand cycle benchmark simulation evaluation on the fish school morphology difference data according to the movement consumption cluster regression data and the distribution density data between the fish morphology differences to obtain the feeding demand cycle benchmark data.

4. The method for analyzing fish feed amount based on machine vision according to claim 3, characterized in that: Step S22 includes the following steps: Step S221: Calculate the size of each fish based on the fish morphology difference data to obtain morphology, body size and data; Step S222: quantifying the fluid resistance of each fish according to the fish morphology difference data based on the morphology and body size data to obtain the fish morphology fluid resistance quantification data; Step S223: extracting the swimming mode of each fish from the fish school motion video image based on the spatiotemporal motion trajectory data between the fish morphological differences, and obtaining the fish swimming mode data; Step S224: performing resistance compensation between different swimming modes on the quantified data of the fluid resistance of the fish morphology according to the data of the swimming mode of the fish, and obtaining the fluid resistance compensation data of the fish morphology-swimming mode; Step S225: Estimate the motion consumption of individual fish based on the spatiotemporal motion trajectory data of different fish morphologies according to the fish morphology-swimming mode fluid resistance compensation data, and obtain motion consumption estimation data of different fish morphologies.

5. The method for analyzing fish feed amount based on machine vision according to claim 4, characterized in that: Compensating the resistance between different swimming modes for the quantitative data of the fluid resistance of the fish morphology according to the data of the swimming mode of the fish includes the following steps: According to the fish swimming mode data, key posture nodes between different swimming modes are analyzed to obtain swimming mode posture node analysis data; Based on the swimming mode posture node analysis data, fluid dynamic field simulation is performed to obtain posture node associated fluid dynamic field data; The mean difference of surface pressure distribution of the attitude node is calculated for the fluid dynamic field data associated with the attitude node to obtain the mean difference data of surface pressure distribution of the node; According to the mean difference data of node surface pressure distribution, the resistance compensation between different swimming modes is performed on the quantitative data of fish morphology fluid resistance to obtain the fish morphology-swimming mode fluid resistance compensation data.

6. The method for analyzing fish feed amount based on machine vision according to claim 4, characterized in that: Estimating the motion consumption of the spatiotemporal motion trajectory data between the fish morphology differences according to the fluid resistance compensation data of the fish morphology-swimming mode includes the following steps: According to the fish morphology-swimming mode fluid resistance compensation data, the spatiotemporal motion trajectory data between the fish morphology differences are used to extract the speed change between the resistance states of different swimming modes, and obtain the speed change data between the resistance states of different swimming modes; Based on the speed change data between different swimming mode resistance states and the fish morphology-swimming mode fluid resistance compensation data, multi-directional force output balance decomposition is performed to obtain multi-directional force output decomposition data; Perform output torque matching on the multi-directional force output decomposition data to obtain multi-directional output torque matching data; The motion consumption of the spatiotemporal motion trajectory data of fish with different morphologies is estimated based on the multi-directional force output decomposition data and the multi-directional output torque matching data, so as to obtain the motion consumption estimation data of fish with different morphologies.

7. The method for analyzing fish feed amount based on machine vision according to claim 3, characterized in that: Step S24 includes the following steps: Step S241: performing theoretical growth cycle analysis of fish morphology difference data according to the movement consumption cluster regression data, and obtaining theoretical growth cycle of fish morphology difference; Step S242: performing a periodic fluctuation analysis of satiation intake between different growth stages of the theoretical growth cycle of fish morphological differences based on the exercise consumption cluster regression data to obtain periodic fluctuation data of satiation intake; Step S243: defining the growth base increments between different fish morphological differences for the theoretical growth period of fish morphological differences according to the distribution density data between different fish morphological differences, and obtaining the growth base increment data between different fish morphological differences; Step S244: Perform a feeding requirement cycle benchmark simulation evaluation based on the growth base increment data and satiation intake cycle fluctuation data of different fish morphological differences to obtain feeding requirement cycle benchmark data.

8. The method for analyzing fish feed amount based on machine vision according to claim 1, characterized in that: Step S3 includes: Step S31: performing convolution calculation on the feeding demand cycle benchmark data to obtain feeding demand cycle benchmark convolution data; Step S32: Obtaining real-time temperature data of the fish pond; Step S33: identifying metabolic influencing factors of the estimated exercise consumption data of fish morphological differences according to the real-time temperature data of the fish pond, and obtaining temperature-related metabolic influencing factors; Step S34: performing feeding demand metabolic impact correction on the feeding demand cycle benchmark convolution data based on the temperature-related metabolic impact factor to obtain feeding demand metabolic impact correction data.

9. The method for analyzing fish feed amount based on machine vision according to claim 8, characterized in that: Step S33 includes the following steps: Step S331: performing temperature stratification analysis on the real-time temperature data of the fish pond to obtain stratified temperature distribution data of the fish pond; Step S332: performing movement inhibition activity evaluation on the movement consumption estimation data between fish morphological differences according to the stratified temperature distribution data of the fish pond to obtain temperature-related movement inhibition activity data; Step S333: performing metabolic response function fitting processing on the temperature-related motion inhibition activity data to obtain a metabolic inhibition response function; Step S334: Identify metabolic influencing factors of the temperature-related motion inhibition activity data according to the metabolic inhibition response function to obtain temperature-related metabolic influencing factors.

10. A fish feed feeding amount analysis system based on machine vision, characterized in that: Used to execute the fish feed feeding amount analysis method based on machine vision as claimed in claim 1, the fish feed feeding amount analysis system based on machine vision comprises: The distribution density analysis module is used to collect fish school movement video images through electronic monitoring equipment to obtain fish school movement video images; perform fish school morphology difference analysis on the fish school movement video images to obtain fish school morphology difference data; perform distribution density analysis between fish morphology differences based on the fish school morphology difference data to obtain distribution density data between fish morphology differences; A feeding demand cycle benchmark assessment module is used to mark the spatiotemporal motion trajectories of fish with different morphological shapes according to the distribution density data of fish with different morphological shapes, and obtain the spatiotemporal motion trajectory data of fish with different morphological shapes; estimate the motion consumption of fish individuals based on the spatiotemporal motion trajectory data of fish with different morphological shapes, and obtain the motion consumption estimation data of fish with different morphological shapes; perform a feeding demand cycle benchmark simulation assessment based on the motion consumption estimation data of fish with different morphological shapes, and obtain the feeding demand cycle benchmark data; The metabolic impact correction module is used to obtain the real-time temperature data of the fish pond; identify the metabolic impact factor of the estimated exercise consumption data between the fish morphological differences according to the real-time temperature data of the fish pond, and obtain the temperature-related metabolic impact factor; based on the temperature-related metabolic impact factor, correct the feeding demand metabolic impact of the feeding demand cycle benchmark data to obtain the feeding demand metabolic impact correction data; The feeding amount adjustment model construction module is used to construct a fish feed feeding amount adjustment model based on the feeding demand metabolic impact correction data through the policy gradient algorithm, obtain the fish feed feeding amount adjustment model, and send the fish feed feeding amount adjustment model to the cloud platform to perform fish feed feeding amount analysis.

Images