Multi-dimensional data shrimp group health automatic assessment method

Abstract

The invention relates to the field of image detection, in particular to a multi-dimensional data shrimp group health automatic evaluation method, which comprises the following steps of: arranging a main monitor and an auxiliary monitoring cluster above a shrimp group to be monitored, shooting the shrimp group to be monitored, forming a corresponding health evaluation image, numbering according to a shooting time sequence, and pre-learning to obtain a health evaluation result. The health learning model learns the main assessment pre-data, generates a bubble number change diagram, compares the bubble number change diagram with a number threshold value, generates a corresponding health assessment result according to a comparison result, and learns the auxiliary assessment pre-data to generate an auxiliary judgment diagram when the health assessment result is unqualified; and performing weighted average calculation on the bubble number change diagram and the auxiliary judgment diagram according to a preset proportion, comparing an evaluation average value with a number threshold value, generating a final health evaluation result, and automatically evaluating the health condition of the shrimp group by utilizing machine learning, thereby improving the evaluation efficiency and accuracy.

Description

Technical Field

[0001] The invention relates to the field of image detection, and in particular to a method for automatically evaluating shrimp health using multi-dimensional data. Background Art

[0002] With the advancement of science and technology, deep learning technology is used to perform morphological measurement and health assessment of shrimp, which helps optimize feeding, predict harvest yields, determine the traits required for selective breeding, the grading process, and monitor the health of shrimp populations. This technology can quickly and accurately obtain phenotypic data of aquaculture species through digital imaging and training prediction models. The development and application of these technologies provide a technical basis and implementation path for the automatic assessment method of shrimp population health based on multi-dimensional data, making shrimp population health assessment more scientific, accurate and efficient.

[0003] Chinese patent application publication number: CN103761565A discloses an underwater fish, shrimp and crab seedling quantity estimation and behavior monitoring device and method based on computer vision, which features include a two-way winch and a motor that drive the video acquisition device to sample all-round and different water layers in the nursery pond, a float for carrying the motor, and a video acquisition device for collecting seedling videos. The steps are as follows: placing the video acquisition device in water, and driving the two-way winch and the motor to collect seedling videos in the nursery pond, and using image processing technology to estimate the number and density of seedlings in the entire nursery pond using the collected seedling images in different water layers and different directions; turning on the light attractant light source, and using image processing technology to calculate the target's swimming speed and group center coordinates from the captured video to reflect the seedling's movement ability and light attractant ability. The advantages are simple structure, wide monitoring range, easy use, and high accuracy, and are suitable for automated aquaculture seedling biological quantity estimation and behavior monitoring.

[0004] Chinese patent application publication number: CN119096931A discloses an intelligent oxygen enrichment device for aquaculture based on neural network, which uses low-energy consumption and high-efficiency technology to significantly increase the dissolved oxygen content in the water and promote the healthy growth of aquatic animals such as shrimp, crabs and fish. The device is suitable for small to medium-sized aquaculture pond environments, and the blower frequency is adjusted by the intelligent control system to adaptively adjust the dissolved oxygen level in the water. The system integrates a fuzzy neural network model with strong self-learning ability, which can predict the dissolved oxygen content based on the real-time data collected by the water quality monitoring module, thereby achieving precise environmental control. In addition, the modular design of the device makes installation and maintenance more convenient, greatly reducing the technical threshold for farmers. The device also integrates water quality monitoring and communication modules, supports real-time environmental monitoring and remote data transmission, and realizes remote monitoring and management functions. This highly integrated intelligent solution provides strong technical support for modern and intelligent aquaculture.

[0005] However, the above method has the following problem: it is not possible to intuitively evaluate the health status of the monitored target by monitoring a certain variable. Summary of the invention

[0006] To this end, the present invention provides a method for automatically evaluating shrimp health based on multidimensional data, so as to overcome the problem in the prior art that it is impossible to intuitively evaluate the health status of the monitored target by monitoring a certain variable.

[0007] To achieve the above object, the present invention provides a method for automatically evaluating shrimp health based on multidimensional data, comprising: A main monitor is set just above the center of the shrimp group to be monitored obtained by sampling in the water, and a number of auxiliary monitors are symmetrically set at equal intervals on both sides of the main monitor to form an auxiliary monitoring cluster; The main monitor and the auxiliary monitor take photos of the shrimp group to be monitored to form corresponding health assessment images, wherein: The health assessment image includes a main assessment image taken by the main monitor and an auxiliary assessment image taken by the auxiliary monitor; Numbering the health assessment images in order of shooting time; Pre-learning the health assessment image and generating corresponding pre-assessment data; The health learning model learns the main assessment pre-data to generate a bubble quantity change graph, compares the bubble quantity change graph with a quantity threshold, and generates a corresponding health assessment result according to the comparison result; When the health assessment result is unqualified, the health learning model learns the auxiliary assessment pre-data to generate an auxiliary judgment graph, and calculates the weighted average of the bubble quantity change graph and the auxiliary judgment graph according to a preset ratio to generate a corresponding assessment mean; Comparing the evaluation mean with the quantity threshold and generating a final health evaluation result; Wherein, the pre-evaluation data includes main pre-evaluation data corresponding to the main evaluation image and auxiliary pre-evaluation data corresponding to the auxiliary evaluation image; The health learning model is generated through training based on a training set formed by a number of pre-evaluation data.

[0008] Further, the steps of setting the monitor include: The main monitor is arranged directly above the shrimp group to be monitored; With the main monitor as the center, a number of auxiliary monitors are arranged at preset intervals on both sides; The main monitor and the auxiliary monitor are adjusted to a preset resolution, wherein: The preset resolution is related to the density of the shrimp group to be monitored.

[0009] Furthermore, the step of taking a health assessment image includes: When the server is connected to the main monitor and the auxiliary monitor, it sends a shooting instruction; The main monitor and the auxiliary monitor photograph the shrimp group to be monitored to obtain corresponding health assessment images; Numbering the health assessment images in order of their corresponding shooting time; Wherein, the main monitor and the auxiliary monitor are provided with cameras with adjustable angles; The auxiliary monitor is aimed at the shrimp group to be monitored.

[0010] Furthermore, the steps of pre-learning the health assessment image include: The pre-learner performs pixel filtering on the health assessment image and forms corresponding assessment filtering information; Selecting a number of singular points of the evaluation and filtering information; Standardizing the evaluation filtering information and generating corresponding evaluation preliminary data; Wherein, a pixel threshold is set in the pre-learner; The singular point includes the pixel size, bubble size and / or bubble density of the evaluation filter information; The standardization process is to segment the evaluation filter information according to a standard learning rate; The standard learning rate is a learning rate that can be recognized by the health learning model, and is inversely proportional to the pixels of the health assessment image. The higher the pixels, the lower the corresponding standard learning rate. For single learning, the corresponding standard learning rate is a single learning rate.

[0011] Furthermore, the step of performing pixel filtering on the health assessment image by the pre-learner includes: The pre-learner acquires pixel information of the health assessment image; comparing the pixel information with the pixel threshold; When the pixel information is greater than the pixel threshold, the health assessment image is retained; when the pixel information is less than the pixel threshold, the health assessment image is filtered.

[0012] Furthermore, the steps of the health learning model learning the primary assessment pre-data include: Adjusting the learning rate of the health learning model to the standard learning rate; transferring the master assessment pre-data into the health learning model; The health learning model learns the main assessment pre-data and obtains a corresponding bubble quantity change graph.

[0013] Furthermore, the value of the bubble quantity variation diagram is compared with the quantity threshold value. Wherein, when the value of the bubble quantity change graph is greater than the quantity threshold, the health assessment result is marked as qualified and output; When the value of the bubble quantity variation graph is less than the quantity threshold, the health assessment result is marked as unqualified, and the auxiliary assessment pre-data is transferred to the health learning model for learning.

[0014] Furthermore, the health learning model generates a corresponding auxiliary judgment graph, calculates the weighted average of the bubble quantity change graph and the auxiliary judgment graph at a preset ratio at the same time, and outputs the weighted average as the evaluation mean.

[0015] Further, the evaluation mean is compared with the quantity threshold, When the evaluation mean is greater than the quantity threshold, the final result of the health evaluation is marked as qualified and output.

[0016] Furthermore, when the evaluation mean is less than the quantity threshold, the final result of the health evaluation is marked as unqualified, the server obtains the bubble quantity change graph and calculates the corresponding stocking density, and adjusts the stocking density according to the appropriate density of the corresponding variety of the shrimp population to be monitored.

[0017] Compared with the prior art, the present invention sets a main monitor and an auxiliary monitoring cluster above the shrimp group to be monitored, shoots the shrimp group to be monitored, forms a corresponding health assessment image, numbers them in order of shooting time and performs pre-learning, generates corresponding assessment pre-data, and the health learning model learns the main assessment pre-data, generates a bubble quantity change graph and compares it with the quantity threshold, generates a corresponding health assessment result according to the comparison result, and when the health assessment result is unqualified, learns the auxiliary assessment pre-data to generate an auxiliary judgment graph, calculates the weighted average of the bubble quantity change graph and the auxiliary judgment graph according to a preset ratio, compares the assessment mean with the quantity threshold, generates the final result of the health assessment, and uses machine learning to automatically assess the health status of the shrimp group, thereby improving the efficiency and accuracy of the assessment.

[0018] Furthermore, by taking the main monitor as the center, it can be ensured that the main monitor can capture the overall condition of the shrimp population from the vertical direction, and a number of auxiliary monitors are set at preset intervals on both sides of it. This layout forms a monitoring cluster, which can capture the local condition of the shrimp population from different angles, providing multi-angle data support for health assessment, improving the automation level of shrimp population health assessment, reducing human intervention, and improving the efficiency and accuracy of the assessment.

[0019] Furthermore, the monitored shrimp population is photographed by the main monitor and the auxiliary monitor to obtain the corresponding health assessment images. The health assessment images of the shrimp population can be collected from multiple angles and at different time points, providing comprehensive data support for subsequent image analysis and health assessment. This multi-angle and multi-time point monitoring method helps to improve the accuracy and reliability of the assessment.

[0020] Furthermore, by filtering the pixels of the health assessment image and forming corresponding assessment filter information, selecting several specific points of the assessment filter information, standardizing the assessment filter information, and generating corresponding assessment pre-data, the pre-learner can extract useful information from the health assessment image and convert it into data suitable for further analysis and learning by the health learning model. This method helps to improve the automation and accuracy of shrimp health assessment.

[0021] Furthermore, the pre-learner performs pixel filtering on the health assessment images, and improves the quality of the image data through pixel-level filtering, ensuring that only images containing important information are used for subsequent health assessments. This approach helps reduce the complexity of data processing while improving the accuracy and efficiency of health assessments. In this way, the pre-learner can effectively filter out valuable information from a large amount of image data, providing clearer and more accurate input data for the health learning model.

[0022] Furthermore, by learning the main assessment pre-data through the health learning model, key information about the health status of the shrimp population can be automatically extracted from a large amount of image data, providing a scientific basis for aquaculture management.

[0023] Furthermore, by comparing the numerical value and weighted average of the bubble quantity change graph with the quantity threshold, the method of comprehensively analyzing the primary evaluation pre-data and the auxiliary evaluation pre-data can more comprehensively evaluate the health status of the shrimp population and make a more accurate judgment. This method helps to detect potential health problems in a timely manner and take corresponding measures to improve the shrimp breeding environment and health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flowchart of a method for automatically evaluating shrimp health based on multi-dimensional data of the present invention; Figure 2 A flow chart for setting up a monitor for the present invention; Figure 3 A flow chart for taking a health assessment picture for the present invention; Figure 4 The present invention is a flowchart of pre-learning health assessment images. DETAILED DESCRIPTION

[0025] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0026] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0027] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is only for description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0028] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0029] See also Figure 1 As shown, it is a flow chart of a method for automatically evaluating shrimp health using multidimensional data of the present invention, comprising: Step S1, a main monitor is set just above the center of the shrimp group to be monitored obtained by sampling in the water, and a number of auxiliary monitors are symmetrically set at equal intervals on both sides of the main monitor to form an auxiliary monitoring cluster; Step S2: The main monitor and the auxiliary monitor take photos of the shrimp group to be monitored to form a corresponding health assessment image, wherein: The health assessment images include a main assessment image taken by a main monitor and an auxiliary assessment image taken by an auxiliary monitor; Step S3, numbering the health assessment images in the order of shooting time; Step S4, pre-learning the health assessment image and generating corresponding pre-assessment data; Step S5, the health learning model learns the main assessment pre-data to generate a bubble quantity change graph, compares the bubble quantity change graph with the quantity threshold, and generates a corresponding health assessment result according to the comparison result; Step S6: When the health assessment result is unqualified, the health learning model learns the auxiliary assessment pre-data to generate an auxiliary judgment diagram, and calculates the weighted average of the bubble quantity change diagram and the auxiliary judgment diagram according to a preset ratio to generate a corresponding assessment mean; Step S7, comparing the evaluation mean with the quantity threshold, and generating a final health evaluation result; The pre-evaluation data includes the main pre-evaluation data corresponding to the main evaluation image and the auxiliary pre-evaluation data corresponding to the auxiliary evaluation image; The health learning model is trained and generated based on a training set formed by several evaluation pre-data.

[0030] In specific implementation, water sampling includes sampling a predetermined number of shrimps in the breeding area and placing them in a transparent monitoring container. The container is equipped with necessary oxygen supply equipment to facilitate subsequent comprehensive health assessment of the shrimp population to be monitored.

[0031] By setting a main monitor and an auxiliary monitoring cluster above the shrimp group to be monitored, the shrimp group to be monitored is photographed to form a corresponding health assessment image, which is numbered and pre-learned in order of the shooting time to generate corresponding assessment pre-data. The health learning model learns the main assessment pre-data, generates a bubble quantity change graph and compares it with the quantity threshold, and generates a corresponding health assessment result based on the comparison result. When the health assessment result is unqualified, the auxiliary assessment pre-data is learned to generate an auxiliary judgment graph, and the weighted average of the bubble quantity change graph and the auxiliary judgment graph is calculated according to a preset ratio. The assessment mean is compared with the quantity threshold to generate a final health assessment result. Machine learning is used to automatically evaluate the health status of the shrimp group, thereby improving the efficiency and accuracy of the evaluation.

[0032] See also Figure 2 As shown, it is a flow chart of setting a monitor in the present invention, including: Step S11, placing the main monitor directly above the shrimp group to be monitored; Step S12, taking the main monitor as the center, a plurality of auxiliary monitors are arranged at predetermined intervals on both sides; Step S13, adjusting the main monitor and the auxiliary monitor to a preset resolution, wherein: The preset resolution is related to the density of the shrimp population to be monitored.

[0033] In the specific implementation, the following steps are included: Positioning of the main monitor: Set the main monitor directly above the shrimp to be monitored. This setting ensures that the main monitor can capture the overall condition of the shrimp from a vertical direction, which is crucial for assessing the overall health of the shrimp.

[0034] Auxiliary monitor layout: With the main monitor as the center, several auxiliary monitors are set at preset intervals on both sides. This layout forms a monitoring cluster that can capture the local conditions of the shrimp population from different angles and provide multi-angle data support for health assessment.

[0035] Monitor resolution adjustment: Adjust the primary and secondary monitors to the preset resolution. The preset resolution refers to the image capture accuracy of the monitor set according to the density of the shrimp to be monitored. This means that the resolution of the monitor is not fixed, but dynamically adjusted according to the density of the shrimp to ensure that the captured image can cover enough shrimp individuals and provide enough details for subsequent health assessment.

[0036] The advantages of this setup method are: Comprehensiveness: Through the coordinated use of the main monitor and the auxiliary monitor, the shrimp population can be fully monitored from multiple angles to reduce blind spots.

[0037] Adaptability: The adjustment of preset resolution enables the monitor to adapt to shrimp populations of different densities, and both sparse and dense shrimp populations can be effectively monitored.

[0038] Accuracy: High-resolution image capture helps to more accurately assess the health of shrimp populations, including individual characteristics such as size, morphology, and activity.

[0039] By taking the main monitor as the center, it can be ensured that the main monitor can capture the overall condition of the shrimp population from the vertical direction. Several auxiliary monitors are set at preset intervals on both sides of the main monitor. This layout forms a monitoring cluster, which can capture the local condition of the shrimp population from different angles, providing multi-angle data support for health assessment, improving the automation level of shrimp population health assessment, reducing human intervention, and improving the efficiency and accuracy of the assessment.

[0040] See also Figure 3 As shown, it is a flow chart of taking a health assessment chart of the present invention, including: Step S21, when the server is connected to the main monitor and the auxiliary monitor, a shooting instruction is sent; Step S22, the main monitor and the auxiliary monitor take photos of the shrimp group to be monitored to obtain corresponding health assessment images; Step S23, numbering the health assessment images in order of their corresponding shooting time; Wherein, the main monitor and the auxiliary monitor are provided with cameras with adjustable angles; The auxiliary monitor is angled toward the shrimp group to be monitored.

[0041] In the specific implementation, the following steps are included: Server connection and shooting command sending: When the server is connected to the main monitor and the auxiliary monitor, the shooting command is sent. This step ensures the synchronization of the shooting process and the timely transmission of the command, providing real-time data support for subsequent health assessment.

[0042] Health assessment image acquisition: The main monitor and auxiliary monitor take photos of the shrimp to be monitored and obtain corresponding health assessment images. These images include main assessment images and auxiliary assessment images, which provide visual data for assessing the health status of the shrimp.

[0043] Image numbering: Number the health assessment images in the order of their shooting time. This step helps maintain the order and traceability of the image data, facilitating subsequent data analysis and processing.

[0044] Adjustable Angle Camera Setting: The main monitor and auxiliary monitor are equipped with adjustable angle cameras. This design allows the monitor to adjust the shooting angle according to actual needs to obtain the best monitoring effect, ensuring that the auxiliary monitor can capture the image of the shrimp from the best angle and improve the accuracy of health assessment.

[0045] The monitored shrimp population is photographed by the main monitor and auxiliary monitor to obtain the corresponding health assessment images. The health assessment images of the shrimp population can be collected from multiple angles and different time points, providing comprehensive data support for subsequent image analysis and health assessment. This multi-angle and multi-time point monitoring method helps to improve the accuracy and reliability of the assessment.

[0046] See also Figure 4 As shown, it is a flow chart of the present invention for pre-learning health assessment images, including: Step St1, the pre-learner performs pixel filtering on the health assessment image and forms corresponding assessment filtering information; Step St2, selecting a number of singular points for evaluating the filtering information; Step St3, standardize the evaluation filtering information and generate corresponding evaluation preliminary data; Wherein, a pixel threshold is set in the pre-learner; Singularities include evaluating pixel size, bubble size, and / or bubble density for filtering information; The normalization process is to split the evaluation filter information according to the standard learning rate; The standard learning rate is a learning rate that can be recognized by the health learning model, which is inversely proportional to the pixels of the health assessment image. The higher the pixels, the lower the corresponding standard learning rate. For single learning, the corresponding standard learning rate is a single learning rate.

[0047] In the specific implementation, the following steps are included: Pixel filtering: The pre-learner performs pixel filtering on the health assessment images: The pre-learner processes the images through the set pixel threshold and filters out pixels below the threshold to reduce noise and unnecessary information, improving the efficiency and accuracy of subsequent analysis.

[0048] Forming evaluation filtering information: The filtered image data is called evaluation filtering information, which contains valid pixel data that has been preliminarily screened, laying the foundation for subsequent feature extraction and analysis.

[0049] Singularity selection: Select several singularities for evaluating the filtering information: In the evaluation filtering information, select singularities according to specific algorithms or rules. These singularities may include but are not limited to pixel size, bubble size and bubble density. These singularities are key features for evaluating the health of shrimp populations.

[0050] Standardization: Standardize the evaluation filter information: Process the evaluation filter information according to certain standards to generate evaluation pre-data. The purpose of standardization is to make the data meet the input requirements of the health learning model and ensure that the model can correctly understand and process the data.

[0051] Generate pre-evaluation data: The data after standardization is called pre-evaluation data, which will be used for training and learning of the health learning model.

[0052] Standard learning rate setting: The standard learning rate is a learning rate that can be recognized by the health learning model: This learning rate is set by the pre-learner based on the pixels of the health assessment image and is inversely proportional to the pixels of the image. In other words, the higher the pixels of the image, the lower the corresponding standard learning rate.

[0053] Relationship between pixels and learning rate: This relationship ensures that the model can learn image features more finely when processing high-pixel images, and can learn faster when processing low-pixel images to adapt to different image quality and resolution.

[0054] Application of single learning rate: For single learning, the corresponding standard learning rate is a single learning rate, which means that each learning is independent, and the setting of the learning rate takes into account the specific needs and conditions of single learning.

[0055] By filtering the pixels of the health assessment image and forming the corresponding assessment filter information, selecting several specific points of the assessment filter information, standardizing the assessment filter information, and generating the corresponding assessment pre-data, the pre-learner can extract useful information from the health assessment image and convert it into data suitable for further analysis and learning by the health learning model. This method helps to improve the automation and accuracy of shrimp health assessment.

[0056] Specifically, the steps of pixel filtering of the health assessment image by the pre-learner include: The pre-learner obtains pixel information of the health assessment image; comparing the pixel information to a pixel threshold; When the pixel information is greater than the pixel threshold, the health assessment image is retained, and when the pixel information is less than the pixel threshold, the health assessment image is filtered.

[0057] In the specific implementation, the following steps are included: Obtaining pixel information: The pre-learner obtains the pixel information of the health assessment image. This is the first step in the pre-learning process. The pre-learner needs to extract pixel information from the health assessment image. This information is the basis for subsequent processing and analysis.

[0058] Comparison of pixel information with threshold: Comparing pixel information with pixel threshold, the pre-learner compares the acquired pixel information with the pre-set pixel threshold. The pixel threshold is a key parameter used to determine which pixel information is important and which can be ignored.

[0059] Filtering Decisions: Keep health assessment images when pixel information is greater than pixel threshold: If the pixel information exceeds the set threshold, it means that the information contained in these pixels is important for health assessment, so the pre-learner will keep these images.

[0060] When the pixel information is less than the pixel threshold, the health assessment image is filtered: On the contrary, if the pixel information is lower than the threshold, these pixels may contain noise or unimportant information, and the pre-learner will filter out these images to reduce the amount of data for subsequent processing and improve processing efficiency.

[0061] The pre-learner performs pixel filtering on the health assessment images, improving the quality of the image data through pixel-level filtering, ensuring that only images containing important information are used for subsequent health assessments. This approach helps reduce the complexity of data processing while improving the accuracy and efficiency of health assessments. In this way, the pre-learner can effectively filter out valuable information from a large amount of image data, providing clearer and more accurate input data for the health learning model.

[0062] Specifically, the steps for the health learning model to learn the primary assessment pre-data include: Adjust the learning rate of the healthy learning model to the standard learning rate; Passing master assessment pre-data into the health learning model; The health learning model learns the main evaluation pre-data and obtains the corresponding bubble quantity change graph.

[0063] In a specific implementation, the learning rate of the health learning model is adjusted to the standard learning rate: before starting learning, the learning rate of the model needs to be set to a specific value, which is called the standard learning rate. This standard learning rate is pre-set based on the pixel information of the health assessment image and is inversely proportional to the pixels of the image, that is, the higher the image pixels, the lower the corresponding standard learning rate. Such a setting helps the model learn detailed information from high-pixel images more effectively. Preferably, when the standard learning rate is set to 160 / second-180 / second, the health learning model has the best learning effect on the preprocessed data, and the standard learning rate is usually set to 170 / second.

[0064] By learning the main assessment pre-data through the health learning model, key information about the health status of shrimp populations can be automatically extracted from a large amount of image data, providing a scientific basis for aquaculture management.

[0065] Specifically, the value of the bubble quantity change graph is compared with the quantity threshold. When the value of the bubble quantity change graph is greater than the quantity threshold, the health assessment result is marked as qualified and output; When the value of the bubble quantity change graph is less than the quantity threshold, the health assessment result is marked as unqualified, and the auxiliary assessment pre-data is passed into the health learning model for learning.

[0066] In some possible implementations, the size of the health assessment image is set to 1mx1m, and the quantity threshold is 200. When the value corresponding to the bubble quantity change graph of the health assessment image A is 240, the value is greater than the quantity threshold, and the health assessment result is marked as qualified and output; when the value corresponding to the bubble quantity change graph of the health assessment image B is 140, the value is less than the quantity threshold, and the health assessment result corresponding to the health assessment image B is marked as unqualified.

[0067] Specifically, the health learning model generates a corresponding auxiliary judgment graph, calculates the weighted average of the bubble quantity change graph and the auxiliary judgment graph at the same time under a preset ratio, and outputs the weighted average as the evaluation mean.

[0068] Specifically, the evaluation mean is compared with a quantity threshold, Among them, when the evaluation mean is greater than the quantity threshold, the final result of the health evaluation is marked as qualified and output.

[0069] Specifically, when the evaluation mean is less than the quantity threshold, the final result of the health assessment is marked as unqualified, the server obtains the bubble quantity change graph and calculates the corresponding stocking density, and adjusts the stocking density according to the appropriate density of the corresponding species of the shrimp population to be monitored.

[0070] In some possible implementations, the size of the health assessment image is set to 1mx1m, the number threshold is 200, the value corresponding to the A health assessment image in the bubble number change graph is 150, the value corresponding to the auxiliary judgment graph is 400, and the preset ratio is 3:1. The evaluation mean corresponding to the A health assessment image is calculated to be 212.5, which is greater than the number threshold, and the final health assessment result is marked as qualified and output; the value corresponding to the B health assessment image in the bubble number change graph is 100, the value corresponding to the auxiliary judgment graph is 200, and the preset ratio is 3:1. The evaluation mean corresponding to the B health assessment image is calculated to be 125, which is less than the number threshold. The server obtains the bubble number change graph and calculates the stocking density corresponding to the B health assessment image, and adjusts the stocking density according to the suitable density of the corresponding variety of the shrimp population to be monitored.

[0071] In the specific implementation, the following steps are included: Generate auxiliary judgment diagram: The health learning model generates the corresponding auxiliary judgment graph: When the value of the bubble quantity change graph is less than the quantity threshold, that is, the preliminary evaluation result is unqualified, the health learning model will further analyze the auxiliary evaluation pre-data and generate an auxiliary judgment graph. The auxiliary judgment graph contains additional features or indicators, such as shrimp activity frequency, feeding behavior, etc., to provide a more comprehensive health status assessment.

[0072] Calculate the weighted average: Calculate the weighted average of the bubble quantity change graph and auxiliary judgment graph at the same time in a preset ratio: Calculate the weighted average of the bubble quantity change graph and auxiliary judgment graph at the same time in a preset ratio. This weighted average takes into account the importance and reliability of the two types of data to obtain a more accurate evaluation mean.

[0073] Output evaluation mean: The weighted average is output as the evaluation mean: this evaluation mean is a comprehensive reflection of the bubble quantity change graph and the auxiliary judgment graph, providing a more comprehensive indicator of the shrimp population health status.

[0074] Final result determination and output: Compare the mean assessment to the population threshold: The mean assessment is again compared to the population threshold to determine the final health assessment of the shrimp population.

[0075] When the evaluation mean is greater than the quantity threshold: If the evaluation mean exceeds the quantity threshold, this indicates that the overall health status of the shrimp population is qualified, and this result is marked as qualified and output.

[0076] When the assessment mean is less than the quantity threshold: If the assessment mean is still below the quantity threshold, this indicates that the health status of the shrimp population does not meet the standard and the result is marked as unqualified.

[0077] Adjust stocking density: The server obtains the bubble quantity change graph and calculates the corresponding stocking density: If the evaluation result is unqualified, the server will calculate the current stocking density based on the bubble quantity change graph.

[0078] Adjust the stocking density according to the appropriate density of the species corresponding to the shrimp group to be monitored: The server will adjust the current stocking density according to the appropriate density standard of the species corresponding to the shrimp group. This includes increasing or decreasing the number of shrimp to optimize the breeding environment and improve the health level of the shrimp group.

[0079] By comparing the numerical value and weighted average of the bubble quantity change graph with the quantity threshold, this method of comprehensive analysis of the primary evaluation pre-data and the auxiliary evaluation pre-data can more comprehensively evaluate the health status of the shrimp population and make more accurate judgments. This method helps to detect potential health problems in a timely manner and take corresponding measures to improve the shrimp farming environment and health status.

[0080] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for automatically assessing shrimp health based on multidimensional data, characterized in that: include: A main monitor is set just above the center of the shrimp group to be monitored obtained by sampling in the water, and a number of auxiliary monitors are symmetrically set at equal intervals on both sides of the main monitor to form an auxiliary monitoring cluster; The main monitor and the auxiliary monitor take photos of the shrimp group to be monitored to form corresponding health assessment images, wherein: The health assessment image includes a main assessment image taken by the main monitor and an auxiliary assessment image taken by the auxiliary monitor; Numbering the health assessment images in order of shooting time; Pre-learning the health assessment image and generating corresponding pre-assessment data; The health learning model learns the main assessment pre-data to generate a bubble quantity change graph, compares the bubble quantity change graph with a quantity threshold, and generates a corresponding health assessment result according to the comparison result; When the health assessment result is unqualified, the health learning model learns the auxiliary assessment pre-data to generate an auxiliary judgment diagram, and calculates the weighted average of the bubble quantity change diagram and the auxiliary judgment diagram according to a preset ratio to generate a corresponding assessment mean; Comparing the evaluation mean with the quantity threshold and generating a final health evaluation result; Wherein, the pre-evaluation data includes main pre-evaluation data corresponding to the main evaluation image and auxiliary pre-evaluation data corresponding to the auxiliary evaluation image; The health learning model is generated through training based on a training set formed by a number of pre-evaluation data.

2. The method for automatic shrimp health assessment based on multidimensional data according to claim 1, characterized in that: The steps to setting up the monitor include: The main monitor is arranged directly above the shrimp group to be monitored; With the main monitor as the center, a number of auxiliary monitors are arranged at preset intervals on both sides; The main monitor and the auxiliary monitor are adjusted to a preset resolution, wherein: The preset resolution is related to the density of the shrimp group to be monitored.

3. The method for automatic shrimp health assessment based on multidimensional data according to claim 2, characterized in that: The steps for capturing health assessment images include: When the server is connected to the main monitor and the auxiliary monitor, it sends a shooting instruction; The main monitor and the auxiliary monitor photograph the shrimp group to be monitored to obtain corresponding health assessment images; Numbering the health assessment images in order of their corresponding shooting time; Wherein, the main monitor and the auxiliary monitor are provided with cameras with adjustable angles; The auxiliary monitor is aimed at the shrimp group to be monitored.

4. The method for automatic shrimp health assessment based on multidimensional data according to claim 3 is characterized in that: The steps for pre-learning the health assessment images include: The pre-learner performs pixel filtering on the health assessment image and forms corresponding assessment filtering information; Selecting a number of singular points of the evaluation and filtering information; Standardizing the evaluation filtering information and generating corresponding evaluation preliminary data; Wherein, a pixel threshold is set in the pre-learner; The singular point includes the pixel size, bubble size and / or bubble density of the evaluation filter information; The standardization process is to segment the evaluation filter information according to a standard learning rate; The standard learning rate is a learning rate that can be recognized by the health learning model, and is inversely proportional to the pixels of the health assessment image. The higher the pixels, the lower the corresponding standard learning rate. For single learning, the corresponding standard learning rate is a single learning rate.

5. The method for automatic shrimp health assessment based on multidimensional data according to claim 4 is characterized in that: The steps of pixel filtering of health assessment images by the pre-learner include: The pre-learner acquires pixel information of the health assessment image; comparing the pixel information with the pixel threshold; When the pixel information is greater than the pixel threshold, the health assessment image is retained; when the pixel information is less than the pixel threshold, the health assessment image is filtered.

6. The method for automatic shrimp health assessment based on multidimensional data according to claim 5, characterized in that: The steps for the health learning model to learn the primary assessment pre-data include: Adjusting the learning rate of the health learning model to the standard learning rate; transferring the master assessment pre-data into the health learning model; The health learning model learns the main assessment pre-data and obtains a corresponding bubble quantity change graph.

7. The method for automatically assessing shrimp health based on multidimensional data according to claim 6, characterized in that: Compare the value of the bubble quantity change graph with the quantity threshold, Wherein, when the value of the bubble quantity change graph is greater than the quantity threshold, the health assessment result is marked as qualified and output; When the value of the bubble quantity variation graph is less than the quantity threshold, the health assessment result is marked as unqualified, and the auxiliary assessment pre-data is transferred to the health learning model for learning.

8. The method for automatically assessing shrimp health based on multidimensional data according to claim 7, characterized in that: The health learning model generates a corresponding auxiliary judgment graph, calculates the weighted average of the bubble quantity change graph and the auxiliary judgment graph at a preset ratio at the same time, and outputs the weighted average as the evaluation mean.

9. The method for automatic shrimp health assessment based on multidimensional data according to claim 8, characterized in that: comparing the evaluation mean with the quantity threshold, When the evaluation mean is greater than the quantity threshold, the final result of the health evaluation is marked as qualified and output.

10. The method for automatically assessing shrimp health based on multidimensional data according to claim 9, characterized in that: When the evaluation mean is less than the quantity threshold, the final result of the health evaluation is marked as unqualified, the server obtains the bubble quantity change graph and calculates the corresponding stocking density, and adjusts the stocking density according to the appropriate density of the corresponding variety of the shrimp population to be monitored.

Images