Energy-sustaining fish school environment monitoring method and device

Abstract

The embodiments of the present specification disclose an energy-self-sustaining fish environment monitoring method and device, the method comprising: charging a battery based on a preset energy-self-sustaining module in a current breeding area, so as to start a fish behavior detection module in the current breeding area based on the discharge of the battery; collecting environmental data of the current breeding based on a multi-sensor module; calling a fish video image collected by a preset binocular camera, and preprocessing the fish video image to obtain a video image of a fish school to be identified; evaluating the growth status and behavior of the fish school video image to be identified based on a fish category recognition model, a growth status recognition model and a behavior recognition model; extracting environmental data according to the fish growth status and fish behavior to obtain environmental parameters affecting the environmental data, and calculating the environmental parameters affecting the environmental parameters based on a regression fitting method to obtain the optimal environmental parameters affecting the environmental parameters, so as to adjust the current breeding area based on the optimal environmental parameters affecting the environmental parameters.

Description

Technical Field

[0001] The present invention relates to the field of scientific aquaculture technology, and in particular to a method and device for monitoring the environment of fish populations with self-sustaining energy. Background Art

[0002] Sea ranching is an artificial breeding system built in shallow seas or bays, which aims to provide a breeding and reproduction site for marine organisms. Compared with land-based breeding, sea ranching has the advantages of less land, environmental friendliness, and abundant resources. For some countries, sea ranching can be a way to increase food supply and create economic opportunities. At the same time, due to its environmental friendliness and sustainability, sea ranching is also believed to have a positive impact on protecting local natural ecosystems.

[0003] In offshore ranches, there are still some shortcomings in the process of detecting and monitoring fish schools. For example, water currents, suspended matter, algae, etc. in the marine environment will affect light propagation and image quality. The morphology and behavior of fish schools may change in different environments and conditions, resulting in poor adaptability of detection algorithms. The monitoring system in offshore ranches needs to process a large amount of images and data, the monitoring parameters are single, and the maintenance and management costs are high. Summary of the invention

[0004] In order to solve the above technical problems, one or more embodiments of the present specification provide an energy-self-sustaining fish environment monitoring method.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of the present specification provide an energy-sustaining fish swarm environment monitoring method, the method comprising:

[0007] Based on the preset energy self-sustaining module of the current breeding area, the battery is charged to start the fish school behavior detection module of the current breeding area based on the discharge of the battery; wherein the preset energy self-sustaining module includes: a solar power generation component and a wave power generation component;

[0008] Collecting the current breeding environment data according to the multi-sensor module in the fish school behavior detection module;

[0009] Calling the video image of the school of fish collected by the preset binocular camera according to the collection time of the environmental data, and preprocessing the video image of the school of fish to obtain the video image of the school of fish to be identified;

[0010] Based on the fish category recognition model, and the growth status recognition model and behavior recognition model corresponding to each fish category, the video image of the school of fish to be recognized evaluates the growth status and fish behavior in the current breeding area;

[0011] The environmental data is extracted according to the growth status and behavior of fish in the current breeding area to obtain the influencing environmental parameters in the environmental data, and the influencing environmental parameters are calculated based on a regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters.

[0012] Optionally, in one or more embodiments of the present specification, the preset energy self-sustaining module based on the current breeding area charges the battery to start the fish school behavior detection module in the current breeding area based on the discharge of the battery, specifically including:

[0013] The voltage and current of the solar panel are measured and transmitted to the main controller, so as to track the maximum power generation power of the solar power generation assembly based on the control of the mechanical structure by the main controller, so as to adjust the solar panel in the solar power generation assembly in real time; wherein the solar panel is connected to the mechanical structure for controlling the steering;

[0014] The changing voltage is boosted to a preset first voltage value by a boost circuit connecting the solar panel and the wave generator, and the battery is charged based on a battery charging integrated circuit corresponding to the battery; wherein the first voltage value is 36V;

[0015] The output voltage of the battery is processed based on the voltage stabilizing module connected to the battery to obtain the processed output voltage, and the output voltage is obtained based on the connection between each component in the preset energy self-sustaining module and the battery to start the fish school behavior detection module;

[0016] Wherein, after charging the battery based on the battery charging integrated circuit corresponding to the battery, the method further includes:

[0017] The battery management module based on the battery charging integrated circuit detects the charging voltage of the battery, and if it is determined that the voltage of the battery is greater than a preset second voltage value, the battery charging integrated circuit is turned off; wherein the second voltage value is 25.2V.

[0018] Optionally, in one or more embodiments of the present specification, collecting the current aquaculture environment data according to the multi-sensor module in the fish school behavior detection module specifically includes:

[0019] According to the main processor in the fish school behavior detection module, a data acquisition instruction is sent to the coprocessor of the multi-sensor model;

[0020] The coprocessor analyzes the data collection instruction to obtain data collection rules to control each of the multi-sensor modules to collect environmental data based on the data collection rules; wherein the environmental data includes: water body data and meteorological data;

[0021] The environmental data is transmitted to the co-processor to forward the environmental data to the main controller based on the co-processor.

[0022] Optionally, in one or more embodiments of the present specification, calling a video image of a school of fish collected by a preset binocular camera according to the collection time of the environmental data, and preprocessing the video image of the school of fish to obtain a video image of the school of fish to be identified specifically includes:

[0023] Determine the collection time of the environmental data based on the main controller, and call the fish school video image of the collection time collected by the preset binocular camera based on the collection time of the environmental data;

[0024] Based on different channel values ​​of each pixel point in the video image of the school of fish, the mean value of each channel in the video image of the school of fish is calculated based on the different channel values ​​of each pixel point; wherein the channels include: R color channel, G color channel, and B color channel;

[0025] Determine a brightness channel in the school of fish video image, correct other channels of the school of fish video image based on the mean value of the brightness channel, and merge the corrected channels with the brightness channel to obtain a color-corrected school of fish video image;

[0026] The fish school video image is subjected to noise filtering based on a preset median filter to obtain a fish school video image to be identified.

[0027] Optionally, in one or more embodiments of the present specification, before evaluating the growth status and behavior of fish in the current breeding area based on the fish category recognition model, and the growth status recognition model and behavior recognition model corresponding to each fish category, the method further includes:

[0028] The preset convolutional neural network is trained based on image samples of various types of fish to obtain a fish category recognition model;

[0029] Performing morphological compensation on image samples of each type of fish to obtain compensated image samples, extracting feature data from the compensated image samples based on a corner detection algorithm, inputting the feature data into a preset expert experience model for training, and obtaining a growth status recognition model corresponding to each fish category;

[0030] Based on a preset feature extractor, behavioral feature data in a video data set of each fish behavior is extracted, and the behavioral feature data is input into a long short-term memory network to perform sequence modeling on the behavioral features to obtain an initial behavior recognition model, so as to train the initial behavior recognition model to obtain a behavior recognition model.

[0031] Optionally, in one or more embodiments of the present specification, the extracting feature data from the compensated image sample based on a corner detection algorithm specifically includes:

[0032] The autocorrelation function of the corner detection algorithm is defined as:

[0033] E(u,v)=∑ x,y w(x,y)[I(x+u,y+u)-I(x,y)] 2 ; Where (x, y) is the pixel coordinate position corresponding to the window, (u, v) is the offset of the window, I(x, y) represents the image grayscale, w(x, y) is the window function, and the w weight coefficient corresponding to all pixels in the window is 1;

[0034] The estimate of the autocorrelation function is estimated as: Among them, M is a 2*2 matrix which can be obtained by the inverse of the image:

[0035] The matrix M is diagonalized to obtain the grayscale change rate of each point in the sample image in the horizontal and vertical directions. If it is determined that the grayscale change rate is less than a preset threshold, the point is extracted as feature data.

[0036] Optionally, in one or more embodiments of the present specification, the environmental data is extracted according to the growth status and behavior of fish in the current breeding area to obtain the influencing environmental parameters in the environmental data, specifically including:

[0037] The environmental data corresponding to the growth status and behavior of fish in the current breeding area is used as the current sample set, so as to decentralize the sample data in the sample set and obtain standard sample data;

[0038] Calculating the covariance matrix of the standard sample data to perform eigenvalue decomposition on the covariance matrix to obtain an eigenvector corresponding to the maximum eigenvalue;

[0039] The corresponding eigenvectors are standardized to construct an eigenvector matrix based on the standardized eigenvectors, and the sample data in the current sample set are transformed according to the eigenvector matrix to obtain a sample set after dimensionality reduction, so that the environmental data corresponding to the sample set after dimensionality reduction is used as the parameters affecting the environment.

[0040] Optionally, in one or more embodiments of the present specification, the influencing environmental parameters are calculated based on a regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters, specifically including:

[0041] Constructing a discrete function of the environmental impact parameter and the fish growth condition;

[0042] Perform differential fitting according to the discrete function to obtain the optimal influencing environment parameters;

[0043] Whether to make adjustments is determined based on the standard parameter value range corresponding to the optimal environmental impact parameter and the values ​​of each optimal environmental impact parameter in the current breeding environmental data.

[0044] One or more embodiments of the present specification provide an energy-sustaining fish environment monitoring device, the device comprising: a main controller, an energy-sustaining module connected to the main controller, a multi-sensor module, and a preset binocular camera;

[0045] The main controller is based on the Raspberry Pi series MPU and is equipped with a Linux operating system to control the operation of each connected device.

[0046] The energy self-sustaining module is used to charge the battery based on the energy self-sustaining module composed of the solar power generation component and the wave power generation component, so as to start each module in the current breeding area based on the discharge of the battery;

[0047] The multi-sensor module is used to collect environmental data of the current breeding;

[0048] The preset binocular camera is used to collect fish school video images according to the collection time of the environmental data, and pre-process the fish school video images to obtain the fish school video images to be identified;

[0049] The main controller is further used to evaluate the growth status and behavior of fish in the current breeding area based on the fish category recognition model, as well as the growth status recognition model and behavior recognition model corresponding to each fish category, using the video image of the school of fish to be identified; extract the environmental data according to the growth status and behavior of fish in the current breeding area to obtain the influencing environmental parameters in the environmental data, and calculate the influencing environmental parameters based on a regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters.

[0050] Optionally, in one or more embodiments of the present specification, the device further includes: a communication module based on a 433 communication module, a wifi module and a bluetooth module, for performing data interaction in a variety of communication modes.

[0051] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0052] The solar power generation components and wave power generation components are used to charge the battery to form an energy self-sustaining module, and then the battery discharge is used to power each module, so that the overall equipment can achieve self-sufficient energy supply capabilities. The multi-sensor module realizes the perception of water body data and meteorological data, and combines multiple models to identify and analyze fish behavior, which facilitates the analysis of the relationship between environmental data and fish behavior and the extraction of environmental data that affects fish behavior, making it easier to make efficient and scientific adjustments to offshore ranch farming. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0054] Figure 1 A schematic flow chart of a method for monitoring fish swarm environment with energy self-sustaining provided in an embodiment of this specification;

[0055] Figure 2 A cross-sectional schematic diagram of a current breeding area provided in an embodiment of this specification;

[0056] Figure 3 A work flow chart of an energy-sustaining fish swarm environment monitoring method provided in an embodiment of this specification;

[0057] Figure 4 A logic diagram of an overall scheme for energy-sustaining fish swarm environmental supervision provided in the embodiments of this specification;

[0058] Figure 5 A schematic diagram of a partial flow chart of an algorithm provided in an embodiment of this specification;

[0059] Figure 6 A schematic diagram of the basic network structure of a fish category recognition model provided in an embodiment of this specification;

[0060] Figure 7 A schematic diagram of the principle of a corner detection algorithm provided in an embodiment of this specification;

[0061] Figure 8 A schematic diagram of the algorithm result of a long short-term memory network provided in an embodiment of this specification;

[0062] Fig. 9 A schematic diagram of component connections of an energy self-sustaining module provided in an embodiment of this specification;

[0063] Fig.10 A schematic diagram of the equipment connection of an energy-sustaining fish environment monitoring device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0064] The embodiments of this specification provide an energy-sustaining fish environment monitoring method and device.

[0065] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0066] like Figure 1 As shown, the embodiment of this specification provides an energy-sustaining fish school environment monitoring method and device, and the method specifically includes the following steps:

[0067] S101: charging a battery based on a preset energy self-sustaining module in a current breeding area, so as to start a fish school behavior detection module in the current breeding area based on the discharge of the battery; wherein the preset energy self-sustaining module includes: a solar power generation component and a wave power generation component.

[0068] In order to ensure the energy supply of the fish environment monitoring equipment, an energy self-sustaining module including a solar power generation component and a wave power generation component is set in the embodiment of this specification, so as to charge the battery based on the preset energy self-sustaining module in the current breeding area, avoiding the problem that the external power supply easily causes insufficient power supply to affect daily monitoring and the high power supply cost. After charging the battery based on the solar power generation component and the wave power generation component, the battery can discharge to maintain the working voltage of each module in the fish behavior detection module, thereby realizing the startup and operation of each module to achieve energy persistence and self-sustaining.

[0069] Specifically, in the embodiment of the present specification, the battery is charged based on the preset energy self-sustaining module of the current breeding area, so as to start the fish school behavior detection module of the current breeding area based on the discharge of the battery, which specifically includes: in order to ensure that the solar power generation component can achieve the maximum power generation power, the voltage and current of the solar panel are first measured and transmitted to the main controller in the embodiment of the present specification, so as to control the mechanical structure to track the maximum power generation power of the solar power generation component according to the main controller, that is, to adjust the solar panel steering in the solar power generation component in real time by adjusting the rotation of the mechanical structure. Among them, it can be understood that the solar panel is connected to the mechanical structure that controls the steering. Then, the changing voltage is boosted to a preset first voltage value by the boost circuit connecting the solar panel and the wave generator, so as to charge the battery according to the battery charging integrated circuit corresponding to the battery. Among them, it should be noted that the preset first voltage value is 36V. Then, the output voltage of the battery is processed according to the voltage stabilizing module connected to the battery to obtain the processed output voltage, so as to obtain the output voltage according to the connection between each component in the preset energy self-sustaining module and the battery, and start the fish school behavior detection module. Among them, in order to ensure the safety of the charging and discharging process, after charging the battery based on the battery charging integrated circuit corresponding to the battery, the method also includes: executing battery charging voltage equalization, electrostatic protection, reverse protection, over-discharge protection, short circuit protection, trickle charging, constant current charging, constant voltage charging and other functions based on the charging and battery management module to ensure the safety of the charging and discharging process, and the battery management module based on the battery charging integrated circuit detects the charging voltage of the battery. If it can be determined that the voltage of the battery is greater than the preset second voltage value of 25.2V, then the battery charging integrated circuit is turned off.

[0070] S102: Collecting the current aquaculture environment data according to the multi-sensor module in the fish school behavior detection module.

[0071] In order to obtain the environmental data of the current breeding area, the embodiment of this specification collects the environmental data of the current breeding according to the multi-sensor module in the fish behavior detection module. It should be noted that the multi-sensor module is composed of multiple types of sensors, including but not limited to turbidity sensors, pH sensors, TDS sensors, temperature sensors, dissolved oxygen sensors, light intensity sensors, GPS sensors, etc. that can obtain relevant environmental data during the growth of fish. When collecting environmental information data, the coprocessor will parse the main control command and trigger the sensor sequence in the multi-sensor module, so as to collect the surrounding environment information in real time based on multiple sensors and send it to the coprocessor. Specifically, in one or more embodiments of this specification, collecting the environmental data of the current breeding according to the multi-sensor module in the fish behavior detection module specifically includes the following processes:

[0072] First, according to the main processor in the fish school behavior detection module, a data acquisition instruction is sent to the coprocessor of the multi-sensor model. Then, according to the coprocessor, the data acquisition instruction is parsed to obtain data acquisition rules to control each multi-sensor module to collect environmental data based on the data acquisition rules; it should be noted that the environmental data includes but is not limited to: water data such as pH value, dissolved oxygen and various ions, and meteorological data such as light intensity. Then, each sensor transmits the collected environmental data to the coprocessor, so that the environmental data is forwarded to the main controller based on the coprocessor.

[0073] Among them, it is also necessary to explain that the various components in the water body in the environmental data are closely related to the growth of fish. For example, when the pH value is lower than 6.5, the pH value of fish blood decreases, the oxygen-carrying function of hemoglobin is impaired, and fish tissues are hypoxic. Although the dissolved oxygen content in the water is normal at this time, the fish still show symptoms of hypoxia. When the pH value is too high, the ion NH4+ is converted into molecular ammonia NH3, the toxicity increases, the water body is strongly alkaline, corrodes the gill tissue of fish, causes respiratory disorders, and suffocates fish in severe cases. Strong alkaline water bodies also affect the activity of microorganisms and thus affect the degradation of organic matter by microorganisms. The dissolved oxygen content in aquaculture water should generally be 5-8 mg / L, and should be maintained at least above 4 mg / L. When hypoxia occurs, fish become restless and breathe faster. Most of them are concentrated in the surface water. When hypoxia is severe, a large number of fish float to the surface, unable to swim, and even suffocate to death. When the dissolved oxygen is oversaturated, there is generally no harm, but it can sometimes cause bubble disease in fish, especially in the seedling cultivation stage. Ammonia nitrogen in water exists as molecular ammonia and ionic ammonia. Molecular ammonia is very toxic to fish, while ionic ammonia is not only non-toxic, but also a source of nutrition for aquatic plants. When the ammonia concentration in water is too high, it will cause fish to develop septicemia. If it is too high for a long time, it will inhibit the growth and reproduction of fish, and severely poisoned people may even die.

[0074] S103: calling the video image of the school of fish collected by the preset binocular camera according to the collection time of the environmental data, and preprocessing the video image of the school of fish to obtain the video image of the school of fish to be identified.

[0075] based on Figure 3 The flowchart shown in the figure shows that after obtaining environmental data through multi-sensor modules, in order to determine whether the current environment for fish growth is suitable for ensuring the normal growth of fish based on the current behavior and growth status of fish. In the embodiment of this specification, multiple fish school video images corresponding to the acquisition time collected by the preset binocular camera will be called according to the acquisition time of the environmental data, and then the fish school video images will be pre-processed to obtain the fish school video images to be identified for subsequent operations. Figure 2 The preset binocular camera shown is installed underwater in the current breeding area and connected to the energy self-sustaining module to obtain power from the battery.

[0076] Specifically, in one or more embodiments of the present specification, the video image of the school of fish collected by the preset binocular camera is called according to the collection time of the environmental data, and the video image of the school of fish is preprocessed to obtain the video image of the school of fish to be identified, which specifically includes the following process:

[0077] First, the collection time of environmental data is determined based on the main controller, so as to call the preset binocular camera to obtain the fish school video image detected at the corresponding collection time according to the collection time of the environmental data. Then, in order to solve the problem of image distortion in the underwater environment, the fish school video image will be processed. In the embodiment of this specification, the different channel values ​​of each pixel in the fish school video image are first obtained, and then the mean of each channel in the fish school video image is calculated according to the different channel values ​​of each pixel. Among them, it can be understood that the three color channels of the fish school video image include: R color channel, G color channel, and B color channel. Then, in order to correct the distorted color in the image, it is necessary to determine the brightness channel in the fish school video image, and then correct the other channels of the fish school video image according to the mean of the brightness channel, so as to merge the corrected channel with the brightness channel to obtain the color-corrected fish school video image. In order to eliminate isolated noise points and small-size noise in the video image of the school of fish after color correction, retain the edge and detail information of the image, effectively reduce the clutter in the marine environment, and improve the detection accuracy of the school of fish, the embodiment of this specification uses a pre-set median filter to filter the noise of the school of fish video image to obtain the video image of the school of fish to be identified.

[0078] S104: Based on the fish category recognition model, and the growth status recognition model and behavior recognition model corresponding to each fish category, the video image of the school of fish to be recognized is used to evaluate the growth status and fish behavior of the fish in the current breeding area.

[0079] like Figure 5 As shown, after obtaining environmental data and video images of fish schools through hardware, it is necessary to perform identification and analysis based on the video images of fish schools. Therefore, in the embodiments of this specification, the video images of the fish schools to be identified are based on the fish category identification model, and the growth status identification model and behavior identification model corresponding to each fish category, to evaluate the growth status and behavior of fish in the current breeding area, so as to facilitate the subsequent analysis of the relationship between fish school behavior and environmental data, thereby achieving reliable maintenance and efficient management of marine ranch breeding. Furthermore, in one or more embodiments of this specification, before the video images of the fish schools to be identified are based on the fish category identification model, and the growth status identification model and behavior identification model corresponding to each fish category, to evaluate the growth status and behavior of fish in the current breeding area, the method also includes:

[0080] Through the image samples of various types of fish Figure 6 The preset convolutional neural network shown is trained to obtain a fish category recognition model. It should be noted that the convolutional neural network (CNN) is a deep learning model widely used in the fields of image recognition and computer vision. It has a wide range of applications in image recognition, target detection, face recognition, video analysis and other fields. It can classify, locate and segment images by learning and extracting features from images. Therefore, in this application, a large amount of fish picture data and annotations are used as image samples to train the preset convolutional neural network, so that the trained convolutional neural network, that is, the fish category recognition model, can achieve the output of fish.

[0081] Then, morphological compensation is performed on the image samples of each type of fish, that is, corrosion, expansion, opening and closing operations are performed in sequence to remove the noise of the fish body in the image samples and obtain compensated image samples. Then, feature data of the compensated image samples are extracted based on the corner detection algorithm, and the feature data is input into the preset expert experience model for training to obtain the growth status recognition model corresponding to each fish category. In one or more embodiments of the present specification, the feature data of the compensated image samples is extracted based on the corner detection algorithm, which specifically includes the following process:

[0082] First, define the autocorrelation function of the corner detection algorithm as:

[0083] E(u,v)=∑ x,y w(x,y)[I(x+u,y+u)-I(x,y)] 2 ; Where (x, y) is the pixel coordinate position corresponding to the window, (u, v) is the offset of the window, I(x, y) represents the image grayscale, w(x, y) is the window function, and the w weight coefficient corresponding to all pixels in the window is 1. Then the estimated autocorrelation function is:

[0084] Among them, M is a 2*2 matrix which can be obtained by the inverse of the image:

[0085] like Figure 7 As shown, the matrix M is diagonalized to obtain the grayscale change rates λ1, λ2 of each point in the sample image in the horizontal and vertical directions. If it is determined that the grayscale change rates are all less than the preset threshold, the point is extracted as feature data.

[0086] While training and obtaining the fish category recognition model and the growth status recognition model corresponding to each fish category based on the above process, the behavior feature data in the video data set of each fish behavior is extracted based on the preset feature extractor, and the behavior feature data is input into the long short-term memory network to perform sequence modeling on the behavior feature, and an initial behavior recognition model is obtained, and the initial behavior recognition model is trained to obtain the behavior recognition model. That is, in the embodiment of this specification, the corresponding behavior recognition model is obtained by training the long short-term memory network, and the specific implementation steps are:

[0087] First, a video dataset containing different fish behaviors is collected. For each video, the behavior category of each frame is labeled. Data preprocessing is performed, including video frame extraction, resizing, and image enhancement. A pre-trained CNN model is used as a feature extractor. The video image data in the video dataset is input into the CNN model. The time series model LSTM is used to model the features of the continuous video image data sequence. The feature representation learning method or the features in the time series model are input into the fully connected layer or other classifiers for training. The trained model is evaluated with a validation set, and the model is tuned according to the evaluation results. Different network architectures, hyperparameter settings, and regularization methods can be tried to improve model performance. Accuracy, recall rate, and F1 score are used as evaluation criteria for the model to obtain a behavior recognition model that meets the requirements. The model training based on visual technology through the above steps facilitates the subsequent detection of fish species, body length, and body shape index based on target recognition algorithms and scientific analysis, and improves the efficiency of detecting and judging fish behavior activities in sea ranches.

[0088] S105: Extracting the environmental data according to the fish growth status and fish behavior in the current breeding area to obtain the influencing environmental parameters in the environmental data, and calculating the influencing environmental parameters based on a regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters.

[0089] In order to realize data-driven calculation of fish growth status and parameters and to evaluate the current breeding environment, the embodiments of this specification extract environmental data according to the growth status and behavior of fish in current breeding, obtain the influencing environmental parameters in the environmental data, and calculate the influencing environmental parameters based on regression fitting to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area through the optimal influencing environmental parameters. Specifically, in one or more embodiments of this specification, in order to improve the analysis efficiency, eliminate the collinearity between variables, and find closely related variables, extract environmental data according to the growth status and behavior of fish in the current breeding area, and obtain the influencing environmental parameters in the environmental data, which specifically includes the following processes:

[0090] First, the environmental data corresponding to the fish growth status and fish behavior in the current breeding area is used as the current sample set D = (x (1) ,x (2) ,...x (m) ), thereby decentralizing the sample data in the sample set and obtaining standard sample data Then calculate the covariance matrix XX of the standard sample data T , with the covariance matrix XX T Perform eigenvalue decomposition to obtain the eigenvectors (w1, w2, ..., w n′ ). The corresponding eigenvector is normalized to construct an eigenvector matrix W based on the normalized eigenvector, and the sample data in the current sample set is transformed according to the eigenvector matrix. (i) =W T x (i) Get the sample set D′ after dimension reduction = (z (1) ,z (2) ,...z (m) ), and the environmental data corresponding to the sample set after dimensionality reduction is used as the influencing environmental parameters.

[0091] Furthermore, in order to obtain the parameter values ​​of the influencing environmental parameters, it is determined whether the environmental data of the current breeding area needs to be adjusted. In one or more embodiments of the present specification, the influencing environmental parameters are calculated based on the regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters, specifically including: first constructing a discrete function of the environmental influencing parameters and the growth status of the fish. Then, differential fitting is performed based on the discrete function to obtain the optimal influencing environmental parameters. Then, based on the standard parameter value range corresponding to the optimal environmental influencing parameters and the values ​​of each optimal environmental influencing parameter in the environmental data of the current breeding, it is determined whether to adjust.

[0092] like Figure 2 As shown, in order to monitor the fish farming environment in the farming area and provide guidance for subsequent farming, this specification will acquire environmental data and video image data based on hardware acquisition equipment, and then obtain corresponding analysis results based on AI algorithms and transmit them to user terminals based on communication equipment, so that terminal users can obtain corresponding farming information in a timely manner. In order to meet the above requirements, Fig.10 The embodiment of this specification provides a schematic diagram of the equipment connection of an energy-sustaining fish environment monitoring device. Fig.10It can be seen that in one or more embodiments of the present specification, an energy-self-sustaining fish environment monitoring device includes: a main controller, an energy-self-sustaining module connected to the main controller, a multi-sensor module, and a pre-set binocular camera.

[0093] Among them, the main controller is based on the Raspberry Pi series MPU and is equipped with the Linux operating system to control the operation of each connected device. It can be understood that the main controller based on the Raspberry Pi series MPU and is equipped with the Linux operating system has a powerful computing capability to perform model calculations on the collected data of the coprocessor.

[0094] Energy self-sustaining module for Fig. 9 The solar power generation components and wave power generation components shown are used to charge the battery, thereby starting each module, i.e., load, in the current breeding area according to the discharge of the battery.

[0095] The multi-sensor module is used to collect the current breeding environment data. The multi-sensor module includes turbidity sensor, pH sensor, TDS sensor, temperature sensor, dissolved oxygen sensor, light intensity sensor, and GPS sensor.

[0096] A binocular camera is preset to collect video images of the school of fish according to the collection time of the environmental data, and the video images of the school of fish are preprocessed to obtain the video images of the school of fish to be identified.

[0097] The main controller is also used to evaluate the growth status and behavior of fish in the current breeding area based on the fish category recognition model, the growth status recognition model and the behavior recognition model corresponding to each fish category using the video image of the fish school to be identified; extract the environmental data according to the growth status and behavior of fish in the current breeding area to obtain the influencing environmental parameters in the environmental data, and calculate the influencing environmental parameters based on the regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters.

[0098] Furthermore, in order to be able to transmit the analysis results of the main controller to the user terminal, in one or more embodiments of the present specification, the device also includes: a communication module composed of a 433 communication module, a wifi module and a Bluetooth module, which is used to interact with data under multiple communication modes. The communication module composed of multiple modules enables the device to use multiple communication modes to interact with the terminal under different requirements such as communication distance, data volume, stability, etc., thereby improving the scope of application of the device.

[0099] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0100] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0101] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.

Claims

1. A method for monitoring fish environment with energy self-sustaining, characterized in that: The method comprises: Based on the preset energy self-sustaining module of the current breeding area, the battery is charged to start the fish school behavior detection module of the current breeding area based on the discharge of the battery; wherein the preset energy self-sustaining module includes: a solar power generation component and a wave power generation component; Collecting environmental data of the current breeding area according to the multi-sensor module in the fish school behavior detection module; Calling the video image of the school of fish collected by the preset binocular camera according to the collection time of the environmental data, and preprocessing the video image of the school of fish to obtain the video image of the school of fish to be identified; Based on the fish category recognition model, and the growth status recognition model and behavior recognition model corresponding to each fish category, the video image of the school of fish to be recognized evaluates the growth status and fish behavior in the current breeding area; Extracting the environmental data according to the growth status and behavior of fish in the current breeding area to obtain influencing environmental parameters in the environmental data, and calculating the influencing environmental parameters based on a regression fitting method to obtain optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters; The preset energy self-sustaining module in the current breeding area is used to charge the battery, so as to start the fish school behavior detection module in the current breeding area based on the discharge of the battery, specifically including: The voltage and current of the solar panel are measured and transmitted to the main controller, so as to track the maximum power generation power of the solar power generation assembly based on the control of the mechanical structure by the main controller, so as to adjust the solar panel in the solar power generation assembly in real time; wherein the solar panel is connected to the mechanical structure for controlling the steering; The changing voltage is boosted to a preset first voltage value by a boost circuit connecting the solar panel and the wave generator, and the battery is charged based on a battery charging integrated circuit corresponding to the battery; wherein the first voltage value is 36V; The output voltage of the battery is processed based on the voltage stabilizing module connected to the battery to obtain the processed output voltage, and the output voltage is obtained based on the connection between each component in the preset energy self-sustaining module and the battery to start the fish school behavior detection module; Wherein, after charging the battery based on the battery charging integrated circuit corresponding to the battery, the method further includes: The battery management module based on the battery charging integrated circuit detects the charging voltage of the battery, and if it is determined that the voltage of the battery is greater than a preset second voltage value, the battery charging integrated circuit is turned off; wherein the second voltage value is 25.2V; Before evaluating the growth status and behavior of fish in the current breeding area based on the fish category recognition model, the growth status recognition model and the behavior recognition model corresponding to each fish category, the method further includes: The preset convolutional neural network is trained based on image samples of various types of fish to obtain a fish category recognition model; Performing morphological compensation on image samples of each type of fish to obtain compensated image samples, extracting feature data from the compensated image samples based on a corner detection algorithm, inputting the feature data into a preset expert experience model for training, and obtaining a growth status recognition model corresponding to each fish category; Extracting behavior feature data from a video data set of each fish behavior based on a preset feature extractor, inputting the behavior feature data into a long short-term memory network to perform sequence modeling on the behavior features, obtaining an initial behavior recognition model, and training the initial behavior recognition model to obtain a behavior recognition model; The extracting of feature data from the compensated image samples based on a corner detection algorithm specifically includes: The autocorrelation function of the corner detection algorithm is defined as: ; Where (x, y) is the pixel coordinate position corresponding to the window, (u, v) is the offset of the window, I(x, y) represents the image grayscale, w(x, y) is the window function, and the w weight coefficient corresponding to all pixels in the window is 1; The estimate of the autocorrelation function is estimated as: ; Where M is a 2*2 matrix that can be obtained by the inverse of the image: ; The matrix M is diagonalized to obtain the grayscale change rate of each point in the sample image in the horizontal and vertical directions. If it is determined that the grayscale change rate is less than a preset threshold, the point is extracted as feature data.

2. The energy-sustaining fish environment monitoring method according to claim 1 is characterized in that: Collecting environmental data of the current breeding area according to the multi-sensor module in the fish school behavior detection module specifically includes: According to the main processor in the fish school behavior detection module, a data acquisition instruction is sent to the coprocessor of the multi-sensor module; The coprocessor analyzes the data collection instruction to obtain data collection rules to control each of the multi-sensor modules to collect environmental data based on the data collection rules; wherein the environmental data includes: water body data and meteorological data; The environmental data is transmitted to the co-processor to forward the environmental data to the main controller based on the co-processor.

3. The energy-sustaining fish environment monitoring method according to claim 1 is characterized in that: The method further comprises: calling a video image of a school of fish collected by a preset binocular camera according to the collection time of the environmental data, and preprocessing the video image of the school of fish to obtain a video image of the school of fish to be identified. Specifically, the method comprises: Determine the collection time of the environmental data based on the main controller, and call the fish school video image of the collection time collected by the preset binocular camera based on the collection time of the environmental data; Based on different channel values ​​of each pixel point in the video image of the school of fish, the mean value of each channel in the video image of the school of fish is calculated based on the different channel values ​​of each pixel point; wherein the channels include: R color channel, G color channel, and B color channel; Determine a brightness channel in the school of fish video image, correct other channels of the school of fish video image based on the mean value of the brightness channel, and merge the corrected channels with the brightness channel to obtain a color-corrected school of fish video image; The fish school video image is subjected to noise filtering based on a preset median filter to obtain a fish school video image to be identified.

4. The energy-sustaining fish environment monitoring method according to claim 1 is characterized in that: The environmental data is extracted according to the fish growth status and fish behavior in the current breeding area to obtain the influencing environmental parameters in the environmental data, specifically including: The environmental data corresponding to the growth status and behavior of fish in the current breeding area is used as the current sample set, so as to decentralize the sample data in the sample set and obtain standard sample data; Calculating the covariance matrix of the standard sample data to perform eigenvalue decomposition on the covariance matrix to obtain an eigenvector corresponding to the maximum eigenvalue; The corresponding eigenvectors are standardized to construct an eigenvector matrix based on the standardized eigenvectors, and the sample data in the current sample set are transformed according to the eigenvector matrix to obtain a sample set after dimensionality reduction, so that the environmental data corresponding to the sample set after dimensionality reduction is used as the parameters affecting the environment.

5. The energy-sustaining fish environment monitoring method according to claim 1 is characterized in that: The influencing environmental parameters are calculated based on a regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters, specifically including: Constructing a discrete function of the influencing environmental parameters and the fish growth conditions; Perform differential fitting according to the discrete function to obtain the optimal influencing environment parameters; Whether to make adjustments is determined based on the standard parameter value range corresponding to the optimal environmental influencing parameter and the values ​​of each optimal environmental influencing parameter in the environmental data of the current breeding area.

6. An energy-sustaining fish environment monitoring device, characterized in that: The device comprises: a main controller, an energy self-sustaining module connected to the main controller, a multi-sensor module and a preset binocular camera; The main controller is based on the Raspberry Pi series MPU and is equipped with a Linux operating system to control the operation of each connected device. The energy self-sustaining module is used to charge the battery based on the energy self-sustaining module composed of the solar power generation component and the wave power generation component, so as to start each module in the current breeding area based on the discharge of the battery; The multi-sensor module is used to collect environmental data of the current breeding area; The preset binocular camera is used to collect fish school video images according to the collection time of the environmental data, and pre-process the fish school video images to obtain the fish school video images to be identified; The main controller is further used to evaluate the growth status and behavior of fish in the current breeding area based on the fish category recognition model, the growth status recognition model and the behavior recognition model corresponding to each fish category, based on the video image of the school of fish to be recognized; extract the environmental data according to the growth status and behavior of fish in the current breeding area to obtain the influencing environmental parameters in the environmental data, and calculate the influencing environmental parameters based on the regression fitting method to obtain the optimal influencing environmental parameters, so as to adjust the current breeding area based on the optimal influencing environmental parameters; The preset energy self-sustaining module in the current breeding area is used to charge the battery, so as to start the fish school behavior detection module in the current breeding area based on the discharge of the battery, specifically including: The voltage and current of the solar panel are measured and transmitted to the main controller, so as to track the maximum power generation power of the solar power generation assembly based on the control of the mechanical structure by the main controller, so as to adjust the solar panel in the solar power generation assembly in real time; wherein the solar panel is connected to the mechanical structure for controlling the steering; The changing voltage is boosted to a preset first voltage value by a boost circuit connecting the solar panel and the wave generator, and the battery is charged based on a battery charging integrated circuit corresponding to the battery; wherein the first voltage value is 36V; The output voltage of the battery is processed based on the voltage stabilizing module connected to the battery to obtain the processed output voltage, and the output voltage is obtained based on the connection between each component in the preset energy self-sustaining module and the battery to start the fish school behavior detection module; After charging the battery based on the battery charging integrated circuit corresponding to the battery, the method further includes: The battery management module based on the battery charging integrated circuit detects the charging voltage of the battery, and if it is determined that the voltage of the battery is greater than a preset second voltage value, the battery charging integrated circuit is turned off; wherein the second voltage value is 25.2V; Before evaluating the growth status and behavior of fish in the current breeding area based on the video image of the school of fish to be identified based on the fish category identification model, and the growth status identification model and behavior identification model corresponding to each fish category, the method further includes: The preset convolutional neural network is trained based on image samples of various types of fish to obtain a fish category recognition model; Performing morphological compensation on image samples of each type of fish to obtain compensated image samples, extracting feature data from the compensated image samples based on a corner detection algorithm, inputting the feature data into a preset expert experience model for training, and obtaining a growth status recognition model corresponding to each fish category; Extracting behavior feature data from a video data set of each fish behavior based on a preset feature extractor, inputting the behavior feature data into a long short-term memory network to perform sequence modeling on the behavior features, obtaining an initial behavior recognition model, and training the initial behavior recognition model to obtain a behavior recognition model; The extracting of feature data from the compensated image samples based on a corner detection algorithm specifically includes: The autocorrelation function of the corner detection algorithm is defined as: ; Where (x, y) is the pixel coordinate position corresponding to the window, (u, v) is the offset of the window, I(x, y) represents the image grayscale, w(x, y) is the window function, and the w weight coefficient corresponding to all pixels in the window is 1; The estimate of the autocorrelation function is estimated as: ; Where M is a 2*2 matrix that can be obtained by the inverse of the image: ; The matrix M is diagonalized to obtain the grayscale change rate of each point in the sample image in the horizontal and vertical directions. If it is determined that the grayscale change rate is less than a preset threshold, the point is extracted as feature data.

7. The energy-sustaining fish environment monitoring device according to claim 6 is characterized in that: The device also includes: a communication module composed of a 433 communication module, a wifi module and a bluetooth module, which is used for data interaction in multiple communication modes.

Images