Facility aquaculture multi-source information online monitoring and early warning system and method

Abstract

The invention provides a facility aquaculture multi-source information online monitoring and early warning system and method.The system comprises a multifunctional monitoring device, a central controller is arranged in the multifunctional monitoring device, and the central controller comprehensively controls a water quality detection module, an environment detection module and a fish body vision monitoring module; aquaculture water quality and environment parameter data are accurately obtained through the data information processing module and the water quality / environment module controller, detection and evaluation of fish growth characteristics, behavior characteristics and health states are achieved, and the data are uploaded to the cloud server through the communication module; the cloud server has the functions of cloud monitoring, historical data query and threshold exceeding alarm. Multi-source information can be automatically and continuously monitored, intelligent analysis and evaluation and real-time early warning are carried out, an intelligent decision strategy is formed and fed back to regulation and control equipment, the problems that facility aquaculture online monitoring information is not comprehensive, the intelligent degree is low, and multiple kinds of information cannot be dynamically monitored in real time are solved, and development of aquaculture information monitoring equipment is promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-source information monitoring and early warning of facility aquaculture environment and aquatic product bodies, and in particular relates to an online monitoring and early warning system and method for multi-source information of facility aquaculture. Background Art

[0002] Aquaculture is a production activity that artificially breeds, cultivates and harvests aquatic plants and animals. The growth of aquatic organisms is closely related to their breeding environment. For example, temperature, pH, dissolved oxygen, ammonia nitrogen content, etc. will affect the growth of different fish species. Therefore, effective monitoring and management of the breeding environment is crucial for aquaculture. In the process of facility aquaculture, higher requirements are placed on the water quality of aquaculture water to ensure the growth rate and quality of aquaculture. The water quality of aquaculture water is usually monitored in terms of pH, turbidity, ammonia nitrogen concentration, and dissolved oxygen. Especially in high-density intensive farms, due to the high breeding density, the water quality deteriorates rapidly, which can easily cause irreparable and large losses. Therefore, it is necessary to monitor the water quality frequently or even in real time, and adjust the water quality in real time according to the monitoring results to reduce the breeding losses caused by the deterioration of water quality. At the same time, due to high-density breeding, the number of organisms is huge and prone to multiple diseases. In order to effectively monitor and warn of diseases, it is necessary to continuously monitor the aquaculture organisms themselves and form disease analysis and early warning technology so as to quickly and effectively feedback disease information and remind aquaculture managers to replenish medicines and prevent diseases in time. In addition, in order to promote unmanned and intelligent factory farming models, it is also necessary to obtain the growth characteristics and behavioral characteristics of farmed organisms in a timely manner for abnormal behavior analysis and to form intelligent decision-making strategies for water quality control equipment and precision feeding equipment.

[0003] The prior art (CN 115720872A) discloses an intelligent detection and control device for aquaculture environment, which collects water quality information through 6 water quality sensors, namely, dissolved oxygen sensor, pH sensor, temperature sensor, nitrite sensor, ammonia nitrogen sensor and liquid level sensor, transmits information through Alibaba Cloud, and then performs monitoring and early warning. However, the current prior art only monitors water quality, ignores the impact of facility environment, has a single function, and has a low integration degree between water quality index monitoring and environmental index monitoring, and cannot meet the real-time dynamic monitoring needs of multi-source information. At the same time, water quality monitoring is limited to electrochemical sensors, which are easy to wear and tear, and there is no way to obtain information about the cultured organisms themselves, so it cannot be effectively monitored, and the physiological growth characteristics and behavioral characteristics of the cultured organisms cannot be monitored in real time and dynamically, lacks health evaluation of the cultured organisms, and cannot well establish a correlation model between water quality, facility environment and cultured organisms. In addition, there are problems such as low automation and intelligence, high cost, less monitoring information, and poor information feedback effect.

[0004] Based on this, the present invention designs a novel online monitoring and early warning system and method for multi-source information of facility aquaculture. Summary of the invention

[0005] In view of the shortcomings in the prior art, the present invention provides a system and method for online monitoring and early warning of multi-source information in facility aquaculture, which can automatically and continuously monitor multi-source information, conduct intelligent analysis and evaluation and realize early warning, form intelligent decision-making strategies to be used for feedback to control equipment, improve monitoring efficiency, reduce the tediousness of manual monitoring, solve the problems of incomplete online monitoring information of facility aquaculture, low degree of automation and intelligence, and inability to monitor multiple information (water quality-facility environment-cultivated organisms) in real time and dynamically, and promote the development of automated and intelligent equipment for aquaculture information monitoring.

[0006] The present invention achieves the above technical objectives through the following technical means.

[0007] A multi-source information online monitoring and early warning system for facility aquaculture includes a multifunctional monitoring device, wherein a central controller, a water quality detection module, an environmental detection module, a fish visual monitoring module, and a power supply module are arranged inside the multifunctional monitoring device; the water quality detection module and the environmental detection module are both connected to the data information processing module by signal, the data information processing module is connected to the water quality / environmental module controller by signal, the water quality / environmental module controller is connected to the central controller by signal, the water quality detection data and the environmental detection data are transmitted to the central controller, the central controller uploads the data to the cloud server through the 5G communication module, the cloud server is provided with a cloud configuration design module and a remote monitoring and early warning module, the remote monitoring and early warning module realizes water quality pollution degree evaluation and abnormal early warning; the central controller is also deployed with a fish visual monitoring module execution algorithm, based on the data collected by the fish visual monitoring module, the fish growth characteristic measurement, behavior characteristic detection, disease assessment and status assessment are realized.

[0008] Furthermore, the multifunctional monitoring equipment includes an equipment housing, support wheels and rotating support feet are installed at the bottom of the equipment housing, and the upper part of the interior of the equipment housing is divided into a front device room and a rear control room by a partition. The total phosphorus digestion detection module in the water quality detection module is arranged inside the front device room, and the total phosphorus digestion detection module includes a digestion and photoelectric detection integrated module, a peristaltic pump, a peristaltic pump stepper motor, a five-row valve, a five-row valve control circuit, a metering module, and a metering module control circuit; the rear control room is provided with a data information processing module, a water quality / environmental module controller, a central controller, a 5G communication module and a power supply module.

[0009] Furthermore, the water quality sensor module, water quality sampling unit and solvent storage module in the water quality detection module are all arranged at the lower part of the equipment shell; the water quality sampling unit includes a water quality flow path pipeline, which is used to transmit the water sample to be tested to the water quality sensor module, and a Y-type filter is provided on the flow path pipeline. An external pump is connected to the water quality flow path pipeline to collect the water sample to be tested from the breeding pond to the water quality detection module; the solvent storage module is used to place the chemical reagent bottles required for the detection and the waste liquid collection bottle for the detection; the water quality sensor module includes a high-precision sensor for detecting various water quality parameters, and the sensor is inserted in the flow path pipeline structure.

[0010] Furthermore, the environment detection module is installed outside the device housing and includes an air temperature and humidity sensor, a CO2 concentration sensor, a wind speed sensor, and a light intensity sensor.

[0011] Furthermore, a human-computer interaction module is installed on the upper front part of the equipment shell of the multifunctional monitoring equipment, which is used for on-site equipment operation and program setting, and is also used to display water quality parameter information, environmental parameter information, fish growth characteristics, fish behavior characteristics, fish disease assessment, and fish status assessment data information.

[0012] A method for online monitoring and early warning of multi-source information of facility aquaculture based on the above-mentioned online monitoring and early warning system of multi-source information of facility aquaculture, wherein the remote monitoring and early warning module adopts the pollution index method to evaluate the degree of water pollution, selects water pH, ammonia nitrogen, COD, turbidity, and total phosphorus as water pollution evaluation indicators, obtains the standard index by the ratio of the measured value of each water quality indicator to its evaluation standard, and then calculates the weight of the water quality indicator by the improved entropy method, linearly adds the standard index and the weight to obtain a comprehensive pollution index, which is used to evaluate the pollution of the comprehensive water quality; the central controller adopts edge computing to realize fish disease assessment, growth characteristic measurement, and behavior characteristic detection, based on Fish growth characteristic data are used to evaluate fish status. The central controller uses an improved target detection deep learning model, namely the SK-YOLOv5s model, to detect underwater fish diseases. To address the problem of blurred and color-distorted fish images caused by low underwater light intensity, the SK-YOLOv5s model integrates the visual attention mechanism SKNet with the target detection deep learning model YOLOv5s, and uses the segmentation network UNet to pre-process the image to obtain clear fish disease images. SKNet is then integrated into the Backbone end of YOLOv5s to form a feature extraction network that focuses on pixel-level information, thereby enhancing the ability to recognize blurred fish diseases.

[0013] Furthermore, the specific method of the remote monitoring and early warning module to evaluate the degree of water pollution using the pollution index method is as follows:

[0014] Step 1: Set the single pollution index as:

[0015]

[0016] Where, f(P) is the pollution index of water pollutants; C pH , S pH They are the measured value and limit standard of pH respectively; are the measured value and limit standard of ammonia nitrogen, C COD , S COD are the measured value and limit standard of COD respectively; C TURB , S TURB are the measured value and limit standard of turbidity respectively; C TP , S TP They are the measured value and limit standard of total phosphorus respectively;

[0017] Step 2: The entropy weight calculation of water quality indicators is consistent with the time series. The water quality data contains 5 water quality indicators, represented by the matrix [x ij ] m×n To represent, first standardize the data to get the standardized matrix [y ij ] m×n , at this time the entropy value of the water quality index is:

[0018]

[0019] Where f(H) represents the entropy value, f(p) represents the probability of the element in the i-th row and j-th column of the normalized matrix, f(k) = -1 / lnm, y i represents the i-th row element in the standardized matrix, i=1,2,...,m, m represents the length of the water quality data sequence of the constructed matrix, j=1,2,3,4,5;

[0020] The entropy weight calculation formula of water quality index is:

[0021]

[0022] In the formula, f(ω * ) is the improved entropy weight; n is the number of water quality indicators, n is 5, f(H pH ) is the entropy value of the indicator pH, is the entropy value of the indicator ammonia nitrogen, f(H COD ) is the entropy value of COD, f(H TURB ) is the entropy value of the index turbidity, f(H TP ) is the entropy value of the indicator total phosphorus;

[0023] Step 3: The water pollution degree uses the improved entropy value method to calculate the comprehensive pollution index. The calculation formula is:

[0024]

[0025] In the formula, CWQI is the comprehensive pollution index; according to the comprehensive pollution index of water quality, the water quality status is divided into five levels, and its pollution degree is evaluated as follows:

[0026] Level Ⅰ: CWQI ≤ 0.7, the pollution degree is relatively clean; Level Ⅱ: 0.7 < CWQI ≤ 1.0, the pollution degree is fairly clean; Level Ⅲ: 1.0 < CWQI ≤ 2.0, the pollution degree is slightly polluted; Level Ⅳ: 2.0 < CWQI ≤ 3.0, the pollution degree is moderately polluted; Level Ⅴ: CWQI > 3.0, the pollution degree is severely polluted; among them, the pollution levels of Level Ⅰ - Ⅱ are within the standard limits, and the pollution levels of Level Ⅲ - Ⅴ exceed the warning limits. For the over-limit values, the remote monitoring and early warning module gives an alarm notification and feeds back the abnormal index data to relevant personnel.

[0027] Furthermore, the specific method for the central controller to evaluate the fish state is as follows:

[0028] Step 1: Set the fish weight calculation formula as:

[0029] f(WE) = f(L) * ω 11 + f(WI) * ω 12 + f(HD) * ω 13 + f(TW) * ω 14 + f(TL) * ω 15 + b1

[0030] where ω 11 is the weight coefficient of body length, ω 12 is the weight coefficient of body width, ω 13 is the weight coefficient of head diameter, ω 14 is the weight coefficient of caudal peduncle width, ω 15 is the weight coefficient of caudal peduncle length, and b1 is a constant; f(WE), f(L), f(WI), f(HD), f(TW), f(TL) represent weight, body length, body width, head diameter, caudal peduncle width, and caudal peduncle length respectively;

[0031] Step 2: The fish state evaluation formula is:

[0032] f(D, ID, N, H) = max(1, f(FI, HD, RO)) * W1 + f(WE) * W2 +

[0033] f(SC, AC, S) * W3 + f(OF, NF, HS, MS, LS) * W4 + B

[0034] Among them, W1 is the weight coefficient of disease characteristics, W2 is the weight coefficient of body weight characteristics, W3 is the weight coefficient of color and斑纹 characteristics, W4 is the weight coefficient of behavior characteristics, and B is a constant; f(D, ID, N, H) represents the fish status index, f(FI, HD, RO) represents the disease status index, f(SC, AC, S) represents the fish color and斑纹 characteristics index, and f(OF, NF, HS, MS, LS) represents the fish behavior characteristics index;

[0035] f(FI, HD, RO) = max(f(FI), f(HD), f(RO))

[0036] f(SC, AC, S) = f(SC) * ω 21 + f(AC) * ω 22 + f(S) * ω 23 + b2

[0037]

[0038] Among them, f(FI), f(HD), and f(RO) respectively represent the degree of fish body injury, the degree of hemorrhagic disease, and the degree of rotten body, and max(f(FI), f(HD), f(RO)) represents the maximum disease degree; f(SC), f(AC), and f(S) respectively represent the back color, belly color, and斑纹 depth index of the fish, and ω 21 , ω 22 , ω 23 respectively represent the weight coefficients of the back color, belly color, and斑纹 depth index of the fish, and b2 is a constant; f(OF, NF) and f(HS, MS, LS) respectively represent the fish feeding status index and swimming intensity index; max(f(OF), f(NF)) and max(f(HS), f(MS), f(LS)) respectively represent whether the fish is in the feeding state and whether it is in the maximum swimming intensity;

[0039] According to the fish status index f(D, ID, N, H), the fish status is divided into four grades or categories, and its health level is evaluated:

[0040] Grade Ⅰ (H): 0 < f(D, ID, N, H) ≤ 2, the fish is healthy; Grade Ⅱ (N): 2 < f(D, ID, N, H) ≤ 5, the fish is normal; Grade Ⅲ (ID): 5 < f(D, ID, N, H) ≤ 10, the fish is about to die; Grade Ⅳ (D): f(D, ID, N, H) > 10, the fish has died;

[0041] Among them, the status levels of Grade Ⅰ - Ⅱ are normal states, and the status levels of Grade Ⅲ - Ⅴ are abnormal states. For abnormal states, the remote monitoring and early warning module gives an alarm notification and feedbacks the degree of abnormality to relevant personnel.

[0042] The present invention has the following beneficial effects:

[0043] (1) The multi-source information online monitoring and early warning system and method for aquaculture facilities of the present invention have complete functions and solve the problem of low integration of aquaculture water quality index monitoring, environmental index monitoring and aquaculture organism health monitoring. At the same time, it also enriches the traditional contact sensor water quality detection method and forms an automatic monitoring module for total phosphorus based on ammonium molybdate spectrophotometry.

[0044] (2) The multi-source information online monitoring and early warning system and method for facility aquaculture of the present invention takes into account the monitoring and analysis of the growth characteristics and behavioral characteristics of the cultured organisms themselves, and realizes fish growth characteristic measurement (body length L, body width WI, head diameter HD, tail peduncle width TW, tail peduncle length TL, back color SC, abdominal color AC, stripes S, body weight WE), behavioral characteristic detection (feeding OF, non-feeding NF, high-intensity swimming HS, medium-intensity swimming MS, low-intensity swimming LS), disease assessment (fish body injury FI, hemorrhagic disease HD, body rot RO) and status assessment (death D, impending death ID, normal N, healthy H), etc. through visual algorithms.

[0045] (3) The present invention integrates the housing of the aquaculture multifunctional monitoring equipment, the central controller, the water quality detection module, the environmental detection module, the fish visual monitoring module, the signal processing module, the 5G communication module, the cloud server and the power supply module into one, forming an automatic and intelligent facility aquaculture multi-source information online monitoring and early warning technology, which solves the problems of low automation level, high cost, less monitoring information and poor information feedback effect of traditional aquaculture monitoring.

[0046] (4) The entire process of the present invention can realize the function of alarming when each monitoring information exceeds the threshold value, which can not only provide intelligent decision-making strategies for automated equipment, but also intelligently analyze the health status of the cultured organisms.

[0047] (5) The technical solution of the present invention greatly reduces the manpower input, improves the level of multi-source information monitoring of aquaculture, and promotes the intelligent development of multi-source information monitoring equipment for facility aquaculture.

[0048] (6) The present invention can view monitoring information on site or online on a cloud platform. The system startup response time is less than 2 seconds, the system stabilization response time is less than 5 seconds, and the single multi-source information acquisition time is less than 1 minute (except for total phosphorus). The water quality total phosphorus detection can realize automatic sampling, reagent addition, high temperature digestion, color reaction, content calculation, self-cleaning and other functions through program control, and realize the calculation of total phosphorus content. The single measurement time is 35 minutes. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1This is a schematic diagram of the framework of the online monitoring and early warning system of the present invention;

[0050] Figure 2 This is a schematic diagram of the overall structure of the online monitoring and early warning system of the present invention;

[0051] Figure 3 This is a schematic diagram of the arrangement of the water quality detection module of the present invention;

[0052] Figure 4 This is a schematic diagram of the arrangement of components in the rear control room of the present invention;

[0053] Figure 5 Schematic diagram for quantifying fish growth characteristics;

[0054] Figure 6 This is a schematic diagram of the SK-YOLOv5s network structure;

[0055] Figure 7 This is a schematic diagram of fish disease detection methods;

[0056] In the figure: 1-multifunctional monitoring equipment; 101-equipment housing; 102-support wheel; 103-rotating support foot; 104-partition; 2-central controller; 3-water quality detection module; 31-water quality sensor module; 311-turbidity sensor; 312-ammonia nitrogen sensor; 313-pH sensor; 314-COD sensor; 32-total phosphorus digestion detection module; 321-digestion and photoelectric detection integrated module; 322-peristaltic pump; 323-peristaltic pump stepper motor; 324-five-row valve; 325-five-row valve control circuit; 326-metering module; 327-metering module control circuit; 33-water quality sampling unit; 331-flow pipeline; 3 32-Y-type filter; 333-external pump; 34-solvent storage module; 341-chemical reagent bottle; 342-waste liquid collection and detection bottle; 4-environmental detection module; 401-air temperature and humidity sensor; 402-CO2 concentration sensor; 403-wind speed sensor; 404-light intensity sensor; 5-fish body visual monitoring module; 501-underwater high-definition camera; 502-signal conditioning circuit; 6-5G communication module; 61-antenna; 7-human-computer interaction module; 8-data information processing module; 9-water quality / environment module controller; 10-cloud configuration design module; 11-remote monitoring and early warning module; 12-power supply module; 13-cloud server. DETAILED DESCRIPTION

[0057] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

[0058] like Figure 1 , 2As shown in Figures 3 and 4, the multi-source information online monitoring and early warning system for aquaculture facilities of the present invention comprises a multifunctional monitoring device 1, wherein a central controller 2 is provided inside the multifunctional monitoring device 1, and the central controller 2 comprehensively controls a water quality detection module 3, an environmental detection module 4 and a fish visual monitoring module 5, and accurately obtains aquaculture water quality and environmental parameter data through a data information processing module 8 and a water quality / environmental module controller 9, and uploads the data to a cloud server 13 through a 5G communication module 6; the cloud server 13 is provided with a cloud configuration design module 10 and a remote monitoring and early warning module 11, and has functions such as cloud monitoring, historical data query, and over-threshold alarm. The fish visual monitoring module 5 is used to assist in realizing the detection and evaluation of fish physiological growth characteristics, behavioral characteristics, and health status, and the execution algorithm of the fish visual monitoring module 5 is deployed in the central controller 2, and the edge computing method is adopted, which can speed up data processing and transmission speed, reduce delay, and reduce visual data traffic. The present invention adopts a modular design, and each module is connected and coordinated through an electric wire signal line, and the power supply module 12 supplies power to the multifunctional monitoring device 1.

[0059] like Figure 2 , 3 As shown in Figure 4, the multifunctional monitoring device 1 includes a device housing 101, a support wheel 102, a rotating support foot 103, and a partition 104. The support wheel 102 and the rotating support foot 103 are installed at the bottom of the device housing 101; the upper part of the device housing 101 is divided into a front device room and a rear control room by a partition 104. The total phosphorus digestion detection module 32 in the water quality detection module 3 is arranged inside the front device room. The total phosphorus digestion detection module 32 includes a digestion and photoelectric detection integrated module 321, a peristaltic pump 322, a peristaltic pump stepper motor 323, a five-row valve 324, a five-row valve control circuit 325, a metering module 326, and a metering module control circuit 327; the rear control room is provided with a data information processing module 8, a water quality / environment module controller 9, a central controller 2, a 5G communication module 6, and a power supply module 12. The partition 104 at the upper part of the device housing 101 separates the detection module from the control center, which can effectively prevent moisture from causing circuit imbalance or short circuit.

[0060] like Figure 3As shown, the total phosphorus digestion detection module 32 can be applied to the detection of total phosphorus in aquaculture water quality, and can continuously detect, analyze, read in real time, and upload data; the total phosphorus digestion detection module 32 has the advantages of less reagents and waste liquid required for detection, suitable for long-term and high-frequency use, and also has the advantages of simple structure, easy maintenance, long maintenance cycle, accurate test results, good stability, etc. The total phosphorus digestion detection module 32 adopts ammonium molybdate spectrophotometry for detection, and realizes automatic sampling, reagent addition, high-temperature digestion, color reaction, content calculation, self-cleaning and other functions through program control to achieve accurate calculation of total phosphorus content. The total phosphorus measurement range is 0-10 mg / L, the resolution is 0.01 mg / L, the shortest measurement cycle is 35 min each time, and the accuracy, range drift, zero drift, etc. are all ≤±10%; the sequential injection method of the peristaltic pump 322 and the five-row valve 324 is adopted, and there is no pump tube consumption and no pump tube blockage in the short term, and the quantitative accuracy is high. The detection principle of the total phosphorus digestion detection module 32 is as follows: after high-temperature digestion at 125°C for 30 minutes, the water sample is used under neutral conditions, and the chemical reagents required for detection are added through the peristaltic pump 322. The heteropolyacid solution formed by ammonium molybdate + ammonium metavanadate + deionized water + hydrochloric acid reacts in the digestion and photoelectric detection integrated module 321. In the middle of one side of the digestion and photoelectric detection integrated module 321 is a light source UV LED with an emission wavelength of approximately 370nm, and in the middle of the other side is a photodiode, which is used to receive the ultraviolet light intensity transmitted through the cuvette. The light source UV LED, the photodiode and the photoelectric control program constitute a photoelectric module.

[0061] like Figure 2 , 3 As shown, the water quality sensor module 31, the water quality sampling unit 33, and the solvent storage module 34 in the water quality detection module 3 are all arranged at the lower part of the inside of the equipment housing 101; wherein the water quality sampling unit 33 includes a water quality flow path pipeline 331, which is used to transmit the water sample to be tested to the water quality sensor module 31, and a Y-type filter 332 is arranged on the flow path pipeline 331, and an external pump 333 is connected to the water quality flow path pipeline 331 through a plastic steel wire water pipe, and its function is to collect the water sample to be tested from the breeding pond to the water quality detection module 3; the solvent storage module 34 is provided with a water quality sensor module 31, and the water quality sensor module 31 is provided with a water quality sensor module 31; the water quality sampling unit 33 includes a water quality flow path pipeline 331, and the water quality sensor module 31 is provided with a Y-type filter 332, and the external pump 333 is connected to the water quality flow path pipeline 331 through a plastic steel wire water pipe, and its function is to collect the water sample to be tested from the breeding pond to the water quality detection module 3; the solvent storage module 34 is provided with a water quality sensor module 31, and the water quality sensor module 31 is provided with a water quality sensor module 31; the water quality sensor module 3 Module 34 is used to place the chemical reagent bottle 341 required for detection and the waste liquid collection bottle 342; the water quality sensor module 31 includes a high-precision sensor for detecting various water quality parameters, and the sensor is inserted into the flow pipe 331 structure, wherein the turbidity sensor 311 is placed before the Y-type filter 332 on the flow pipe 331, the ammonia nitrogen sensor 312, the pH sensor 313, and the chemical oxygen demand COD sensor 314 are placed after the Y-type filter 332, and the dissolved oxygen DO sensor is directly placed in the breeding pond. The water quality sensor module 31 reserves an expandable MODBUS RTU RS485 interface, which can meet the access of other water quality detection sensors such as salinity sensors.

[0062] like Figure 1 , 2 As shown, the environment detection module 4 is installed on the outside of the equipment housing 101 and is composed of an environmental sensor group, including an air temperature and humidity sensor 401, a CO2 concentration sensor 402, a wind speed sensor 403, and a light intensity sensor 404; the data collected by the environment detection module 4 is received and processed by the data information processing module 8, and transmitted to the water quality / environment module controller 9 in real time to realize visual data collection, and then sent to the 5G communication module 6 through the central controller 2 to complete the cloud upload of the data.

[0063] The information between modules is output by digital signals of MODBUS RTU, and the transmission communication interface is RS485 interface. Figure 2 As shown, a human-machine interaction module 7 is installed on the upper front part of the equipment housing 101, which is used for on-site equipment operation and program setting, etc. The human-machine interaction module 7 adopts a programmable and communication protocol using an RS485 serial port screen as an LCD display screen. The human-machine interaction module 7 is used to display water quality parameter information, environmental parameter information, fish growth characteristics, behavioral characteristics, status characteristics and other data information.

[0064] like Figure 1 As shown, the central controller 2 uses the Jetson Nano motherboard, which is an embedded development board based on the GPU processor. The ARM end is transplanted and loaded with the Ubuntu 18.04LTS system, using the AI ​​framework and algorithm based on PyTorch, and also includes the GPU operating environment such as OpenCV and the corresponding calling interface. The central controller 2 is mainly used to deploy the execution algorithm of the fish visual monitoring module 5, which is used for edge computing to improve the real-time processing capability of video / pictures. At the same time, it is also the joint control center of the various detection modules, human-computer interaction module 7, and 5G communication module 6 of the multi-functional monitoring device 1; the fish visual monitoring module 5 includes an underwater high-definition camera 501 and a signal conditioning circuit 502. The water quality / environment module controller 9 uses the STM32F407ZET6 motherboard, which uses the Cortex-M4 core and has 192KB of on-chip SRAM; the water quality / environment module controller 9 is used for the linkage control of signal acquisition and processing of various sensors of the water quality detection module 3 and the environmental detection module 4.

[0065] like Figure 1 , 2As shown, the 5G communication module 6 adopts a 5G DTU wireless communication terminal. The 5G DTU wireless communication terminal has serial data conversion IP data. The DTU is connected to the central controller 2 at the front end through an RS485 interface, and the data is transparently transmitted and encrypted to the background center through 5G cellular network wireless communication. The 5G communication module 6 is installed at the middle lower position of the rear control room partition 104 above the inside of the device housing 101. It transmits signals through an antenna 61, and the antenna 61 is installed on the top of the device housing 101. The 5G DTU wireless communication terminal supports TCP / IP and standard MQTT protocols, and is uploaded to the cloud server after setting the communication protocol. The 5G DTU wireless communication terminal is fully compatible with the 5G network of mainstream domestic operators, providing high-bandwidth, low-latency, and highly reliable wireless network connections for IoT devices and sensors.

[0066] like Figure 1 , 2 As shown, the cloud server 13 uses the Alibaba Cloud IoT platform, which can provide secure and reliable connection and communication capabilities for the multifunctional monitoring device 1, connect to the device downward, support device data collection on the cloud; provide cloud API upward, and command data is sent to the device through API calls to achieve remote control. The IoT platform provides device management, rule engine, data analysis, etc., provides access solutions for different network devices such as 4G, 5G, NB-IoT, WiFi, and provides device-side SDKs for multiple protocols such as MQTT and HTTP / S.

[0067] like Figure 1 , 2 As shown, the cloud configuration design module 10 and the remote monitoring and early warning module 11 are used to realize remote monitoring and early warning of various information of the system, which is a visual embodiment of system integration. The cloud configuration design module 10 performs system cloud monitoring large screen, historical data query, and over-threshold alarm design, and remotely controls the external pump 333, peristaltic pump 322, five-row valve 324 and various sensors in the multifunctional monitoring device 1. The remote monitoring and early warning module 11 realizes the alarm of abnormal values ​​of single indicators of water quality and environmental parameters such as aquaculture water quality and temperature, dissolved oxygen, ammonia nitrogen, pH, COD, total phosphorus, turbidity, salinity, air temperature, humidity, CO2 concentration, wind speed, and light intensity. By setting the upper and lower limits of each indicator, an alarm notification is issued for the over-limit value, and the abnormal indicator data is fed back through the WeChat applet or official account.

[0068] like Figure 1 , 2As shown in Figure 3, the remote monitoring and early warning module 11 can also evaluate the degree of water pollution in order to ensure healthy water quality and realize green breeding; the degree of water pollution is evaluated by the pollution index method, and water pH, ammonia nitrogen, COD, turbidity, and total phosphorus are selected as water pollution evaluation indicators. The standard index is obtained by the ratio of the measured value of each water quality indicator to its evaluation standard, and the weight of the water quality indicator is calculated by the improved entropy method. The standard index and the weight are linearly added to obtain a comprehensive pollution index, which is used to evaluate the pollution of the comprehensive water quality. The specific method is as follows:

[0069] Step 1: Set the single pollution index calculation formula as:

[0070]

[0071] Among them, f(P) is the pollution index of water pollutants, C is the measured value of water pollutants, S is the limit standard of water pollutants, C pH is the measured value of pH, S pH It is the limit standard of pH. is the measured value of ammonia nitrogen, is the limit standard of ammonia nitrogen, C COD is the measured value of COD, S COD is the limit standard of COD, C TURB is the measured value of turbidity, S TURB is the limit standard for turbidity, C TP is the measured value of total phosphorus, S TP It is the limit standard for total phosphorus.

[0072] Step 2: The entropy weight calculation of water quality indicators is consistent with the time series. The water quality data contains 5 water quality indicators (n = 5), with the matrix [x ij ] m×n To represent, first standardize the data to get the standardized matrix [y ij ] m×n , the entropy value calculation formula of water quality index is:

[0073]

[0074] Where f(H) represents the entropy value, f(p) represents the probability of the element in the i-th row and j-th column of the normalized matrix, f(k) = -1 / lnm, y i represents the i-th row element in the standardized matrix (i=1,2,...,m), m represents the length of the water quality data sequence of the constructed matrix, j=1,2,3,4,5; the entropy weight calculation formula of water quality index is:

[0075]

[0076] In the formula, f(ω) is the entropy weight value; the entropy value method is improved as follows: the traditional entropy value method assigns greater weights to stable evaluation indicators. However, in the analysis of water quality index monitoring data, it is expected that water quality indicators with drastic changes will obtain greater entropy weights. Nevertheless, the traditional entropy value method will make the entropy weights of water quality indicators with large differences in changes smaller, which is contrary to the expectation. Therefore, it is necessary to improve the entropy value method. The improved entropy weight calculation formula of the present invention is:

[0077]

[0078] In the formula, f(ω * ) is the improved entropy weight value; n is the number of water quality indicators, where n is 5 here, f(H pH ) is the entropy value of the indicator pH, is the entropy value of the indicator ammonia nitrogen, f(H COD ) is the entropy value of the indicator COD, f(H TURB ) is the entropy value of the indicator turbidity, f(H TP ) is the entropy value of the indicator total phosphorus; to prevent the situation where the weight value of a certain indicator becomes zero due to too small a difference, the calculation result is optimized, and a basic weight is assigned to the water quality indicators.

[0079] Step 3: The water pollution degree uses the improved entropy value method weight to calculate the comprehensive pollution index, and the calculation formula is:

[0080]

[0081] In the formula, CWQI is the comprehensive pollution index; according to the water quality comprehensive pollution index, the water quality status is divided into five levels, and its pollution degree is evaluated.

[0082] Level Ⅰ: CWQI ≤ 0.7, the pollution degree is relatively clean; Level Ⅱ: 0.7 < CWQI ≤ 1.0, the pollution degree is fairly clean; Level Ⅲ: 1.0 < CWQI ≤ 2.0, the pollution degree is slightly polluted; Level Ⅳ: 2.0 < CWQI ≤ 3.0, the pollution degree is moderately polluted; Level Ⅴ: CWQI > 3.0, the pollution degree is severely polluted. Among them, the pollution levels of Level Ⅰ to Ⅱ are within the standard limits, and the pollution levels of Level Ⅲ to Ⅴ exceed the warning limits. For the over-limit values, the remote monitoring and early warning module 11 issues an alarm notification and feedbacks the abnormal index data through the WeChat mini-program or official account.

[0083] Such as Figure 1 、 2As shown in Figures 4 and 5, the fish visual monitoring module 5 is composed of an underwater high-definition camera 501 and a signal conditioning circuit 502. The module is used to collect various data of fish by taking pictures and videos. In the early stage, the collected data are made into a data set and trained on the server through the YOLOv5 series model to generate the best executable file (.pt file for reasoning on the edge calculator) and deployed in the central controller 2. Through the data collection of the fish visual monitoring module 5, the central controller 2 can realize the measurement of fish growth characteristics (body length L, body width WI, head diameter HD, tail peduncle width TW, tail peduncle length TL, back color SC, belly color AC, stripes S, weight WE), behavior characteristics (feeding OF, non-feeding NF, high-intensity swimming HS, medium-intensity swimming MS, low-intensity swimming LS), disease (fish body injury FI, hemorrhagic disease HD, rotten body RO) evaluation and status (death D, dying ID, normal N, healthy H) evaluation, etc.

[0084] like Figure 1 , 5 As shown in Figures 6 and 7, the central controller 2 uses edge computing to identify fish diseases. In the present invention, an improved target detection deep learning model, namely the SK-YOLOv5s model, is used for fish disease assessment to realize underwater fish disease detection. To address the problem of blurred and color-distorted fish images caused by low underwater light intensity, the SK-YOLOv5s model is a farmed fish disease detection method that integrates a visual attention mechanism (selective kernel networks, SKNet) and a target detection deep learning model (you only look once, YOLOv5s) network. The method uses a segmentation network UNet to preprocess the image to obtain a clear fish disease image, and then integrates SKNet into the Backbone end of YOLOv5s to form a feature extraction network that focuses on pixel-level information, thereby enhancing the recognition capability of fuzzy fish diseases.

[0085] Reference Figure 1 , 5 , 6, 7, the central controller 2 uses the SK-YOLOv5s model to implement a method for detecting fish diseases in farmed fish, as follows:

[0086] Step 1: Data collection;

[0087] Step 1.1: Equipment selection: The fish visual monitoring module 5 uses an underwater high-definition camera 501 with high resolution and high frame rate, so as to capture clear images of fast-moving fish schools with a wide viewing angle in a low-light environment;

[0088] Step 1.2: Shooting settings: Place the underwater high-definition camera 501 in the aquaculture pond and fix it, determine the shooting angle, and ensure that the lens has a good field of view; adjust the exposure time and aperture to adapt to the underwater lighting conditions, and add supplementary lights appropriately at night;

[0089] Step 1.3: Data storage: Save the acquired images in chronological order to local storage.

[0090] Step 2: The fish body visual monitoring module 5 performs image preprocessing;

[0091] Step 2.1: Image denoising: Use the segmentation network UNet to perform image segmentation and denoising to reduce the noise generated during underwater shooting; optimize the denoising algorithm by comparing the original image with the denoised image;

[0092] Step 2.2: Image enhancement: Use histogram equalization technology to improve the contrast of the image; use data enhancement techniques such as rotation, flipping, and scaling to expand the data set to improve the robustness of the model;

[0093] Step 2.3: Data labeling: Label the processed images with diseases to provide supervision information for model training.

[0094] Step 3: Feature extraction and model integration;

[0095] Step 3.1: Model selection: Select YOLOv5s as the base model and integrate SKNet to enhance feature extraction capabilities;

[0096] Step 3.2: Network training: Use the labeled data to train the SK-YOLOv5s model to ensure that the disease types (fish injury FI, hemorrhagic disease HD, body rot RO) can be effectively identified under different lighting and background conditions; use the transfer learning method and use the pre-trained model to accelerate the training process;

[0097] Step 3.3: Model evaluation: Evaluate the model performance through cross-validation to ensure that the accuracy and recall meet the expected standards;

[0098] Step 4: Model deployment;

[0099] Step 4.1: Environment configuration: Install the necessary deep learning framework (PyTorch) on the central controller 2; use the Jetson Nano motherboard, which is an embedded development board based on the GPU processor, to ensure that the computing power of the controller meets the model operation requirements, and perform hardware upgrades when necessary;

[0100] Step 4.2: Model loading: Upload the trained SK-YOLOv5s model file (.pt file) to the central controller 2 and perform a loading test to ensure that the model can run normally;

[0101] Step 4.3: System integration: Connect the fish visual monitoring module 5 to the central controller 2, set the data transmission protocol, and realize real-time data stream transmission.

[0102] Step 5: Fish disease detection;

[0103] Step 5.1: Real-time detection: The fish images collected and preprocessed by the fish visual monitoring module 5 are input into the SK-YOLOv5s model for target detection and disease identification;

[0104] Step 5.2: Evaluate the confidence of the detection results and set a threshold to filter out false positive results;

[0105] Step 5.3: Result recording: Archive and transfer the results and related images of each test online to cloud storage for subsequent analysis and tracking;

[0106] Step 6: Result feedback and evaluation;

[0107] Step 6.1: Abnormal alarm: according to the detection results, an automatic alarm mechanism is set up, and an alarm notification is made through the remote monitoring and early warning module 11 to promptly notify the breeding personnel;

[0108] Step 6.2: Generate a report to record the type and extent of the diseases found and provide data for subsequent analysis and evaluation;

[0109] Step 6.3: Data visualization: Use the monitoring screen of the human-computer interaction module 7 and the cloud configuration design module 10 to realize data visualization and storage, display the test results, and facilitate decision-making and analysis.

[0110] Step 7: Continuous monitoring and model updating;

[0111] Step 7.1: Health monitoring: Conduct fish disease inspections in real time, record each test result, and establish health records; analyze disease trends and adjust breeding management strategies;

[0112] Step 7.2: Model optimization: Regularly collect new data and update the model to adapt to new diseases and environmental changes;

[0113] Step 7.3: Perform online learning or incremental learning to maintain the accuracy and timeliness of the model.

[0114] Through the above steps, effective identification and monitoring of fish diseases can be achieved, and the level of intelligent aquaculture management can be improved.

[0115] The central controller 2 uses edge computing to measure fish growth characteristics and detect fish behavior characteristics; specifically, for fish growth characteristic measurement, an improved target detection deep learning model, namely the SK-YOLOv5s-A model, is used, and for fish behavior characteristic detection, an improved target detection deep learning model, namely the SK-YOLOv5s-B model, is used; unlike the SK-YOLOv5s model, the SK-YOLOv5s-A model and the SK-YOLOv5s-B model are retrained with different data sets, the features of target recognition are replaced, different data sets are collected for training, and optimization of new detection tasks is formed; the SK-YOLOv5s-A model and the SK-YOLOv5s-B model are a fusion of a visual attention mechanism (selective kernel networks, SKNet) and a target detection deep learning model (you only look Once, YOLOv5s) network is used to detect the growth and behavioral characteristics of farmed fish. This method uses the segmentation network UNet to preprocess the image to obtain a clear fish feature image, and then integrates SKNet into the Backbone end of YOLOv5s to form a feature extraction network that focuses on pixel-level information, thereby enhancing the recognition, quantification and detection capabilities of fuzzy fish body and fish school features.

[0116] The central controller 2 uses the SK-YOLOv5s-A model and the SK-YOLOv5s-B model to implement a method for measuring fish growth characteristics and detecting behavioral characteristics of farmed fish. The specific steps are consistent with the method for the central controller 2 to use the SK-YOLOv5s model to detect fish diseases in farmed fish. The only difference is in step 2.3 and step 3.2, as follows:

[0117] For growth feature measurement, step 2.3 is specifically as follows: data labeling, annotating the processed images with growth features to provide supervision information for model training; for behavior feature detection, step 2.3 is specifically as follows: data labeling, annotating the processed images with behavior features to provide supervision information for model training.

[0118] For growth feature measurement, step 3.2 is specifically as follows: Network training: Use labeled data to train the SK-YOLOv5s-A model to ensure that the growth characteristics of fish can be effectively quantified and detected under different lighting and background conditions; Use transfer learning methods and use pre-trained models to accelerate the training process; For behavioral feature detection, step 3.2 is specifically as follows: Network training: Use labeled data to train the SK-YOLOv5s-B model to ensure that the behavioral characteristics of fish can be effectively quantified and identified under different lighting and background conditions; Use transfer learning methods and use pre-trained models to accelerate the training process.

[0119] Compared with the steps for disease status detection, the method for detecting growth characteristics and behavioral characteristics of farmed fish can be adaptively modified in steps 5, 6, and 7 to meet the requirements of different detection purposes.

[0120] The specific method of the fish status assessment is as follows:

[0121] Step 1: Set the fish weight calculation formula to:

[0122] f(WE)=f(L)*ω 11 +f(WI)*ω 12 +f(HD)*ω 13 +f(TW)*ω 14 +f(TL)*ω 15 +b1

[0123] Among them, ω 11 is the weight coefficient of body length, ω 12 is the weight coefficient of body width, ω 13 is the weight coefficient of the head diameter, ω 14 is the weight coefficient of the caudal peduncle width, ω 15 is the weight coefficient of the caudal peduncle length, b1 is a constant; f(WE), f(L), f(WI), f(HD), f(TW), and f(TL) represent body weight, body length, body width, head diameter, caudal peduncle width, and caudal peduncle length, respectively.

[0124] Step 2: The fish status assessment formula is:

[0125] f(D,ID,N,H)=max(1,f(FI,HD,RO))*W1+f(WE)*W2+

[0126] f(SC,AC,S)*W3+f(OF,NF,HS,MS,LS)*W4+B

[0127] Among them, W1 is the weight coefficient of disease characteristics, W2 is the weight coefficient of body weight characteristics, W3 is the weight coefficient of color and pattern characteristics, W4 is the weight coefficient of behavioral characteristics, and B is a constant; f(D, ID, N, H) represents the fish state index, f(FI, HD, RO) represents the disease state index, f(SC, AC, S) represents the fish color and pattern characteristic index, and f(OF, NF, HS, MS, LS) represents the fish behavioral characteristic index;

[0128] f(FI,HD,RO)=max(f(FI),f(HD),f(RO))

[0129] f(SC,AC,S)=f(SC)*ω 21 +f(AC)*ω 22 +f(S)*ω23 +b2

[0130]

[0131] Among them, f(FI), f(HD), and f(RO) respectively represent the degree of fish body damage, the degree of hemorrhagic disease, and the degree of body rot, and max(f(FI), f(HD), f(RO)) represents the maximum disease degree (disease state index); f(SC), f(AC), and f(S) respectively represent the back color, belly color, and stripe depth index of the fish, ω 21 、ω 22 、ω 23 respectively represent the weight coefficients of the back color, belly color, and stripe depth index of the fish, b2 is a constant; f(OF,NF) and f(HS,MS,LS) respectively represent the fish feeding state index and swimming intensity index; max(f(OF), f(NF)) and max(f(HS), f(MS), f(LS)) respectively indicate whether the fish is in the feeding state and whether it is in the maximum swimming intensity. When the fish is in the feeding state, max(f(OF), f(NF)) = 0, and when it is not in the feeding state, max(f(OF), f(NF)) = 1. When the fish is in the maximum swimming intensity, max(f(HS), f(MS), f(LS)) = 0, and when it is not in the maximum swimming intensity, max(f(HS), f(MS), f(LS)) = 1.

[0132] According to the fish state index f(D, ID, N, H), the fish state is divided into four grades or categories, and its health level is evaluated:

[0133] Grade Ⅰ (H): 0 < f(D, ID, N, H) ≤ 2, the fish is healthy; Grade Ⅱ (N): 2 < f(D, ID, N, H) ≤ 5, the fish is normal; Grade Ⅲ (ID): 5 < f(D, ID, N, H) ≤ 10, the fish is about to die; Grade Ⅳ (D): f(D, ID, N, H) > 10, the fish is already dead.

[0134] Among them, the state levels of Grade Ⅰ to Ⅱ are normal states, and the state levels of Grade Ⅲ to Ⅴ are abnormal states. For abnormal states, the remote monitoring and early warning module 11 issues an alarm notification and feedbacks the degree of abnormality through a WeChat mini-program or official account.

[0135] Appendix Figure 6In the figure, the Focus module represents the convolutional neural network layer used for feature extraction, which can downsample the input feature map to reduce the amount of calculation and parameters; the CBL module represents the combination of convolution operation (Conv), normalization operation (BN) and activation function (Leaky-ReLU), which is used for feature extraction, accelerated training and enhanced model generalization ability; the SPP module represents spatial pyramid pooling, which is used to achieve an output of an adaptive size; the SCSP1_1 module represents an improved CSPNet (a network structure used to enhance the learning ability of convolutional neural networks) that integrates the SKnet module, which is composed of CBL, Res The CSPNet consists of unit, SKnet, convolutional layer, Concat, BN and Leaky-ReLU, which are used to enhance the feature extraction capability; the SCSP1_3 module indicates that there are three SCSP1_1 modules; the SCSP2_1 module indicates that it is also an improved CSPNet network structure, which is composed of convolutional layer, 2 CBL modules and Concat; the Upsam-pling module indicates an upsampling operation; the Concat module indicates a feature map connection method, which is used to connect feature maps from different layers in the channel dimension to generate richer feature maps; the Conv layer indicates a convolution operation, which usually includes a convolution kernel, a bias term and an activation function.

[0136] In practical applications, the present invention can realize on-site visual monitoring and cloud-based remote monitoring and early warning, and the specific operations can be appropriately modified according to actual conditions. In the present invention, relational terms such as first and then are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. The embodiments described are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.

Claims

1. A multi-source information online monitoring and early warning system for facility aquaculture, characterized in that: The invention comprises a multifunctional monitoring device (1), wherein a central controller (2), a water quality detection module (3), an environment detection module (4), a fish visual monitoring module (5), and a power supply module (12) are arranged inside the multifunctional monitoring device (1); the water quality detection module (3) and the environment detection module (4) are both connected to a data information processing module (8) by signal, the data information processing module (8) is connected to a water quality / environment module controller (9) by signal, the water quality / environment module controller (9) is connected to a central controller (2) by signal, and water quality detection data and environmental detection data are transmitted to the central controller (2); the central controller (2) uploads the data to a cloud server (13) through a 5G communication module (6); the cloud server (13) is provided with a cloud configuration design module (10) and a remote monitoring and early warning module (11); the remote monitoring and early warning module (11) realizes water quality pollution degree evaluation and abnormal early warning; the central controller (2) is also deployed with a fish visual monitoring module (5) to execute an algorithm, and fish growth characteristic measurement, behavior characteristic detection, disease evaluation, and status evaluation are realized based on data collected by the fish visual monitoring module (5).

2. The multi-source information online monitoring and early warning system for facility aquaculture according to claim 1 is characterized in that: The multifunctional monitoring device (1) comprises a device housing (101), a support wheel (102) and a rotating support foot (103) are installed at the bottom of the device housing (101), and the upper part of the interior of the device housing (101) is divided into a front device room and a rear control room by a partition (104), and the total phosphorus digestion detection module (32) in the water quality detection module (3) is arranged inside the front device room, and the total phosphorus digestion detection module (32) comprises a digestion and photoelectric detection integrated module (321), a peristaltic pump (322), a peristaltic pump stepper motor (323), a five-row valve (324), a five-row valve control circuit (325), a metering module (326), and a metering module control circuit (327); and the rear control room is provided with a data information processing module (8), a water quality / environmental module controller (9), a central controller (2), a 5G communication module (6), and a power supply module (12).

3. The multi-source information online monitoring and early warning system for facility aquaculture according to claim 2 is characterized in that: The water quality sensor module (31), the water quality sampling unit (33), and the solvent storage module (34) in the water quality detection module (3) are all arranged at the lower part of the interior of the equipment housing (101); the water quality sampling unit (33) comprises a water quality flow path pipeline (331) for transmitting the water sample to be tested to the water quality sensor module (31); a Y-type filter (332) is arranged on the flow path pipeline (331); an external pump (333) is connected to the water quality flow path pipeline (331) to collect the water sample to be tested from the breeding pond to the water quality detection module (3); the solvent storage module (34) is used to place a chemical reagent bottle (341) and a waste liquid collection and detection bottle (342); the water quality sensor module (31) comprises a high-precision sensor for detecting various water quality parameters, and the sensor is inserted into the flow path pipeline (331) structure.

4. The multi-source information online monitoring and early warning system for facility aquaculture according to claim 1 is characterized in that: The environmental detection module (4) is installed outside the device housing (101), and includes an air temperature and humidity sensor (401), a CO2 concentration sensor (402), a wind speed sensor (403), and an illuminance sensor (404).

5. The multi-source information online monitoring and early warning system for facility aquaculture according to claim 1 is characterized in that: On the upper part of the front of the device housing (101) of the multifunctional monitoring device (1), a human-computer interaction module (7) is installed, which is used for on-site device operation and program setting, and at the same time for displaying water quality parameter information, environmental parameter information, fish growth characteristics, fish behavior characteristics, fish disease assessment, and fish status assessment data information.

6. A method for online monitoring and early warning of multi-source information of facility aquaculture based on the online monitoring and early warning system of multi-source information of facility aquaculture according to claim 1, characterized in that: The remote monitoring and warning module (11) selects water quality pH, ammonia nitrogen, COD, turbidity, and total phosphorus as water quality pollution evaluation indicators. The standard index is obtained by the ratio of the measured value of each water quality index to its evaluation standard. Then, the weight of the water quality index is calculated by the improved entropy method, and the standard index and the weight are linearly added and calculated, etc. to obtain the comprehensive pollution index, which is used to evaluate the pollution situation of the comprehensive water quality. The central controller (2) uses the edge computing method to realize fish disease assessment, growth characteristic measurement, and behavior characteristic detection, and realizes fish status assessment based on fish growth characteristic data. Among them, the central controller (2) uses an improved object detection deep learning model, that is, the SK-YOLOv5s model, to realize underwater fish disease detection. Aiming at the problem that the underwater light intensity is low, resulting in blurred and color-distorted fish body images, the SK-YOLOv5s model fuses the visual attention mechanism SKNet and the object detection deep learning model YOLOv5s. The segmentation network SKNet is used to preprocess the image to obtain a clear fish body disease image, and then the segmentation network is fused into the Backbone end of YOLOv5s to form a feature extraction network that focuses on pixel-level information, strengthening the recognition ability of blurred fish body diseases.

7. The method for online monitoring and early warning of multi-source information of facility aquaculture according to claim 6 is characterized in that: The specific method for the remote monitoring and warning module (11) to evaluate the water quality pollution degree by the pollution index method is as follows: Step 1: Set the single pollution index as: Where, f(P) is the pollution index of water pollutants; C pH , S pH They are the measured value and limit standard of pH respectively; are the measured value and limit standard of ammonia nitrogen, C COD , S COD are the measured value and limit standard of COD respectively; C TURB , S TURB are the measured value and limit standard of turbidity respectively; C TP , S TP They are the measured value and limit standard of total phosphorus respectively; Step 2: The entropy weight calculation of water quality indicators is consistent with the time series. The water quality data contains 5 water quality indicators, represented by the matrix [x ij ] m×n To represent, first standardize the data to get the standardized matrix [y ij ] m×n , at this time the entropy value of the water quality index is: Where f(H) represents the entropy value, f(p) represents the probability of the element in the i-th row and j-th column of the normalized matrix, and f(k) = -1 / ln m, y i represents the i-th row element in the standardized matrix, and m represents the length of the water quality data sequence of the constructed matrix; Then the entropy weight calculation formula of the water quality index is: In the formula, f(ω * ) is the improved entropy weight; n is the number of water quality indicators, f(H pH ) is the entropy value of the pH indicator, f(H NH3-N ) is the entropy value of the indicator ammonia nitrogen, f(H COD ) is the entropy value of COD, f(H TURB ) is the entropy value of the index turbidity, f(H TP ) is the entropy value of the indicator total phosphorus; Step 3: The water quality pollution degree uses the improved entropy method weight to calculate the comprehensive pollution index, and the calculation formula is: In the formula, CWQI is the comprehensive pollution index; according to the water quality comprehensive pollution index, the water quality status is divided into five grades, and its pollution degree is evaluated: Grade Ⅰ: CWQI ≤ 0.7, the pollution degree is relatively clean; Grade Ⅱ: 0.7 < CWQI ≤ 1.0, the pollution degree is still clean; Grade Ⅲ: 1.0 < CWQI ≤ 2.0, the pollution degree is slightly polluted; Grade Ⅳ: 2.0 < CWQI ≤ 3.0, the pollution degree is moderately polluted; Grade Ⅴ: CWQI > 3.0, the pollution degree is severely polluted; among them, the pollution levels of Grade Ⅰ to Ⅱ are within the standard limit, and the pollution levels of Grade Ⅲ to Ⅴ exceed the warning limit. For the over-limit value, the remote monitoring and warning module (11) gives an alarm notification and feeds back the abnormal index data to relevant personnel.

8. The method for online monitoring and early warning of multi-source information of facility aquaculture according to claim 6, characterized in that: The specific method for the central controller (2) to evaluate the fish status is as follows: Step 1: Set the fish weight calculation formula as: f(WE)=f(L)*ω 11 +f(WI)*ω 12 +f(HD)*ω 13 +f(TW)*ω 14 +f(TL)*ω 15 +b1 Among them, ω 11 is the weight coefficient of body length, ω 12 is the weight coefficient of body width, ω 13 is the weight coefficient of the head diameter, ω 14 is the weight coefficient of the caudal peduncle width, ω 15 is the weight coefficient of the caudal peduncle length, b1 is a constant; f(WE), f(L), f(WI), f(HD), f(TW), and f(TL) represent body weight, body length, body width, head diameter, caudal peduncle width, and caudal peduncle length, respectively; Step 2: The fish status evaluation formula is: f(D, ID, N, H) = max(1, f(FI, HD, RO)) * W1 + f(WE) * W2 + f(SC, AC, S) * W3 + f(OF, NF, HS, MS, LS) * W4 + B where W1 is the weight coefficient of disease characteristics, W2 is the weight coefficient of body weight characteristics, W3 is the weight coefficient of color and斑纹 characteristics, W4 is the weight coefficient of behavior characteristics, and B is a constant; f(D, ID, N, H) represents the fish status index, f(FI, HD, RO) represents the disease status index, f(SC, AC, S) represents the fish color and斑纹 characteristics index, and f(OF, NF, HS, MS, LS) represents the fish behavior characteristics index; f(FI, HD, RO) = max(f(FI), f(HD), f(RO)) f(SC,AC,S)=f(SC)*ω 21 +f(AC)*ω 22 +f(S)*ω 23 +b2 Among them, f(FI), f(HD), and f(RO) represent the degree of fish body damage, hemorrhagic disease, and body rot, respectively; max(f(FI), f(HD), and f(RO)) represent the maximum degree of disease; f(SC), f(AC), and f(S) represent the back color, abdominal color, and pattern depth index of fish, respectively; ω 21 ,ω 22 ,ω 23 They represent the weight coefficients of the fish back color, belly color and stripe depth index respectively, and b2 is a constant; f(OF, NF) and f(HS, MS, LS) represent the fish feeding state index and swimming intensity index respectively; max(f(OF), f(NF)) and max(f(HS), f(MS), f(LS)) represent whether the fish is in the feeding state and whether it is in the maximum swimming intensity respectively; According to the fish status index f(D, ID, N, H), the fish status is divided into four grades or categories, and its health level is evaluated: Grade Ⅰ (H): 0 < f(D, ID, N, H) ≤ 2, the fish is healthy; Grade Ⅱ (N): 2 < f(D, ID, N, H) ≤ 5, the fish is normal; Grade Ⅲ (ID): 5 < f(D, ID, N, H) ≤ 10, the fish is about to die; Grade Ⅳ (D): f(D, ID, N, H) > 10, the fish is already dead; The status levels of Grade Ⅰ to Ⅱ are normal status, and the status levels of Grade Ⅲ to Ⅴ are abnormal status. For the abnormal status, the remote monitoring and warning module (11) issues an alarm notification and feedbacks the degree of abnormality to relevant personnel.

Images