Net cage fish school culture feed supplement evaluation method and related equipment
By combining underwater image data and water quality parameters, a multi-source information assessment method has been developed to solve the problem of insufficient or excessive feeding of fish in existing technologies, enabling precise feed delivery and improving aquaculture efficiency and profitability.
Patent Information
- Application Number
- CN202411080430.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing timed and quantitative feeding methods based on fish theoretical growth models or experience are difficult to meet the actual feeding needs of fish, which can easily lead to insufficient or excessive feeding, reducing feed utilization and aquaculture efficiency.
A fish school behavior recognition model based on underwater image data and a lightweight deep learning model are used to identify fish feeding behavior and detect uneaten food. By combining changes in water quality parameters, the feeding energy model is used to predict the feeding energy demand of fish schools, and multi-source information is integrated to assess the feeding status of fish schools.
It provides accurate feeding status assessment results, optimizes feeding strategies, and improves breeding efficiency and benefits.
Smart Images

Figure CN119007289A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to the field of computers, and more specifically, the present invention relates to a method for evaluating feed supplementation in cage fish farming and related equipment. Background Art
[0002] In deep-sea aquaculture scenarios, precise feeding of feed means feeding the right amount of feed within a specific time period based on the actual feeding needs of fish, so as to meet the nutritional needs of fish growth, metabolism, etc., and maximize the utilization rate of feed nutrition. Feed, as the main cost, accounts for about 70% of the total cost and directly affects the aquaculture benefits. At present, the feeding method is mainly based on the theoretical growth model of fish or the preset feeding amount based on experience, which greatly reduces labor costs and labor intensity. However, this feeding method is often quite different from the actual feeding needs of fish, which can easily lead to insufficient or excessive feeding. Summary of the invention
[0003] A series of simplified concepts are introduced in the Summary of the Invention, which will be further described in detail in the Detailed Description of the Invention. The Summary of the Invention does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.
[0004] In order to solve the problem that the timed and quantitative feeding method based on the theoretical growth model of fish or the feeding amount preset by experience is difficult to meet the actual feeding needs of fish, which is easy to cause insufficient or excessive feeding, in the first aspect, the present invention proposes a method for evaluating feed supplementation of cage fish aquaculture, the method comprising:
[0005] Based on the underwater image data, fish feeding behavior is identified by a fish behavior identification model, wherein the fish behavior identification model is a deep learning model that integrates fish appearance and movement information;
[0006] Residual bait detection in aquaculture boxes based on underwater image data through hierarchical feature fusion and enhanced lightweight deep learning model;
[0007] According to the changes in water quality parameters, the feeding energy demand information of fish is predicted through the feeding energy model, and the feeding status of cage-cultured fish is evaluated by combining the identified fish feeding behavior information and the detected residual bait information.
[0008] Specifically, it also includes:
[0009] Based on the physical model of underwater optical imaging, the underwater transmittance image of underwater image data is obtained according to the dark channel prior, RGB three-channel color difference and spatial texture constraints;
[0010] A hierarchical feature fusion strategy is introduced to perform cross-modal fusion of RGB features and transmittance features;
[0011] Combined with the U-shaped network architecture, a decoding module is designed to restore the image resolution step by step and restore the image details from coarse to fine.
[0012] Specifically, it also includes:
[0013] Use labelimg and labelme to roughly classify and label the objects in the underwater image data set, and divide it into training set, validation set and test set;
[0014] Perform projection transformation on multiple frames of images and perform image alignment;
[0015] Use multi-frame difference method to segment the outline of the moving target of the school of fish;
[0016] Combined with the contour information of the fish school target, the Shi-Tomas algorithm is used to extract the corner points of the moving target, and the classifier is designed to eliminate the interference of non-fish school targets;
[0017] A fish school appearance feature YOLOv9 network detector based on a single-frame image and a fish school motion target YOLOv9 network detector based on a time series are designed respectively to obtain the fish school behavior recognition model.
[0018] Specifically, it also includes:
[0019] Label the underwater bait residue image dataset in the underwater image data, and use labelimg to label the bait residue targets;
[0020] Modify the feature encoding subnetwork connection mode to generate a fine-grained feature output layer;
[0021] Enhance feature transfer and reuse, and build feature fusion and feature enhancement modules;
[0022] The number of network layers is reduced, and some residual connection modules are replaced with dense connection modules to obtain a lightweight deep learning model with hierarchical feature fusion and enhancement.
[0023] Specifically, it also includes:
[0024] According to the physiological and behavioral characteristics of fish during feeding, combined with the bait residue in the cage, a feeding intensity label is developed;
[0025] The time series features are extracted using the empirical mode decomposition method and input into the multi-scale CNN and LSTM models for training and judgment to obtain the feeding energy model.
[0026] Specifically, the method of evaluating the feeding state of cage-cultured fish by combining the identified fish feeding behavior information and the detected residual bait information includes:
[0027] Quantify the information of fish feeding behavior, residual bait and feeding energy requirement into feeding characteristic indicators;
[0028] The quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of fish schools.
[0029] Specifically, the quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of the fish school, including:
[0030] The support vector regression model is used to determine the weight of each feature, and the XGBoost ensemble learning algorithm is used to learn and predict the feeding status of fish schools in order to comprehensively evaluate the feeding status of fish schools.
[0031] In a second aspect, the present invention further provides a feed supplementation assessment device for cage fish farming, comprising:
[0032] An identification unit, used for identifying the feeding behavior of a school of fish by using a school of fish behavior identification model based on the underwater image data, wherein the school of fish behavior identification model is a deep learning model integrating the appearance and movement information of the school of fish;
[0033] A detection unit, used to detect residual bait in aquaculture boxes based on underwater image data through hierarchical feature fusion and enhanced lightweight deep learning model;
[0034] The evaluation unit is used to predict the fish feeding energy demand information through the feeding energy model according to the changes in water quality parameters, so as to evaluate the feeding status of the cage-cultured fish in combination with the identified fish feeding behavior information and the detected residual bait information.
[0035] In a third aspect, an electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to implement the steps of the cage fish aquaculture feed supplementation evaluation method as described in any one of the first aspects above when executing the computer program stored in the memory.
[0036] In a fourth aspect, the present invention further proposes a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for evaluating feed supplementation for cage fish farming according to any one of the above items in the first aspect is implemented.
[0037] In summary, the feed supplementation assessment method for caged fish farming proposed in this application is to identify the feeding behavior of fish schools through a fish school behavior recognition model based on underwater image data, wherein the fish school behavior recognition model is a deep learning model that integrates the appearance and motion information of fish schools; to detect residual bait in the farming box through a lightweight deep learning model that integrates hierarchical feature fusion and enhancement based on underwater image data; to predict the feeding energy demand information of fish schools through a feeding energy model according to changes in water quality parameters, so as to evaluate the feeding status of caged fish schools by combining the identified fish school feeding behavior information and the detected residual bait information. In this way, multi-source information is integrated to provide accurate feeding status assessment results. Feeding strategies are optimized based on the assessment results to improve farming efficiency and benefits.
[0038] The cage fish farming feed supplementation evaluation method of the present invention, other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will be understood by technicians in this field through research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present specification. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:
[0040] Figure 1 A schematic diagram of a method for evaluating feed supplementation in cage fish farming provided in an embodiment of the present application;
[0041] Figure 2 A schematic diagram of the structure of a feed supplementation assessment device for cage fish farming provided in an embodiment of the present application;
[0042] Figure 3 A schematic diagram of the structure of an electronic device for evaluating feed supplementation in cage fish farming provided in an embodiment of the present application. DETAILED DESCRIPTION
[0043] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.
[0044] In order to solve the problem that feeding fish with a fixed amount of feed according to theoretical growth model or experience is difficult to meet the actual feeding needs of fish, which may lead to insufficient or excessive feeding, please refer to Figure 1 , is a flow chart of a method for evaluating feed supplementation in cage fish farming provided in an embodiment of the present application, which may specifically include: steps S110 to S130.
[0045] S110, identifying the feeding behavior of a school of fish through a school of fish behavior recognition model based on the underwater image data, wherein the school of fish behavior recognition model is a deep learning model that integrates the appearance and movement information of the school of fish.
[0046] For example, the above-mentioned multiple underwater image data with time sequence can be understood as video data.
[0047] S120 detects residual bait in aquaculture boxes based on underwater image data through hierarchical feature fusion and enhanced lightweight deep learning model.
[0048] S130, predicting the fish feeding energy demand information through a feeding energy model according to the change of water quality parameters, so as to evaluate the feeding status of the cage cultured fish in combination with the identified fish feeding behavior information and the detected residual bait information.
[0049] In summary, the feed supplementation assessment method for caged fish farming provided in the embodiment of the present application is to identify the feeding behavior of fish schools through a fish school behavior recognition model based on underwater image data, wherein the fish school behavior recognition model is a deep learning model that integrates the appearance and motion information of fish schools; to detect residual bait in the farming box based on underwater image data through a lightweight deep learning model that integrates hierarchical feature fusion and enhancement; to predict the feeding energy demand information of fish schools through a feeding energy model according to changes in water quality parameters, so as to evaluate the feeding status of caged fish schools by combining the identified fish school feeding behavior information and the detected residual bait information. In this way, multi-source information is integrated to provide accurate feeding status assessment results. Feeding strategies are optimized based on the assessment results to improve farming efficiency and benefits.
[0050] According to some embodiments, further comprising:
[0051] Based on the physical model of underwater optical imaging, the underwater transmittance image of underwater image data is obtained according to the dark channel prior, RGB three-channel color difference and spatial texture constraints;
[0052] A hierarchical feature fusion strategy is introduced to perform cross-modal fusion of RGB features and transmittance features;
[0053] Combined with the U-shaped network architecture, a decoding module is designed to restore the image resolution step by step and restore the image details from coarse to fine.
[0054] For example, in order to integrate the advantages of traditional underwater image enhancement methods and data-driven deep learning image enhancement models, for the observed underwater RGB images / videos, firstly, based on the physical model of underwater optical imaging, the underwater transmittance image is obtained according to the dark channel prior, RGB three-channel color difference and spatial texture constraints. Secondly, in the encoder design stage, the original RGB image and the obtained transmittance image are respectively input into two independent encoders for feature extraction, and the hierarchical feature fusion strategy is introduced to cross-modally fuse the RGB features with the transmittance features to adaptively screen the iconic features; in the decoder design stage, combined with the advantages of the U-shaped symmetry learning network architecture, the decoding module is designed to restore the image resolution from coarse to fine step by step. In the loss function design process, the stability of the image content is maintained by introducing the mean square error loss, the stability of the overall structure of the image is maintained by introducing the structural similarity loss, and the stability of the image semantic information is maintained by introducing the perceptual loss function. Through the above strategies, the constructed model ensures that the enhanced underwater image is more in line with the human visual evaluation standards. The enhanced underwater image is significantly improved in contrast, color restoration and details.
[0055] In some examples, this also includes:
[0056] Use labelimg and labelme to roughly classify and label the objects in the underwater image data set, and divide it into training set, validation set and test set;
[0057] Perform projection transformation on multiple frames of images and perform image alignment;
[0058] Use multi-frame difference method to segment the outline of the moving target of the school of fish;
[0059] Combined with the contour information of the fish school target, the Shi-Tomas algorithm is used to extract the corner points of the moving target, and the classifier is designed to eliminate the interference of non-fish school targets;
[0060] A fish school appearance feature YOLOv9 network detector based on a single-frame image and a fish school motion target YOLOv9 network detector based on a time series are designed respectively to obtain the fish school behavior recognition model.
[0061] For example, for the fish school video data collected in the breeding box, based on the prior information of the fish school target color, shape structure, brightness, spatial distribution, etc., the labeling tools labelimg and labelme in target detection and semantic segmentation are used to roughly classify and annotate the targets in the data set, and then divided into training set, verification set and test set according to the random sampling method. For the enhanced fish school video sequence, the 2D projection transformation is first used to align multiple frames of images, and the multi-frame difference method is used to segment the contours of the fish school moving targets, so as to construct the fish school movement information. In order to eliminate the interference of non-target motion information such as water body fluctuations and algae, for the segmented moving targets, the Shi-Tomas algorithm is used to extract the moving target corners in combination with the unique contour information of the fish school target, and the classifier is designed based on the statistical laws of the direction and amplitude of the motion vector to eliminate the interference of non-fish school targets.
[0062] It is understandable that in order to make the most of the fish swarm feature information and ensure that the detection results are not affected by background noise, this project intends to use a dual YOLOv9 network as the basic architecture for fish swarm target detection, and design a detector based on the appearance features of the fish swarm based on a single frame image and a fish swarm motion target detector based on a time series. Specifically, for the underwater video sequence collected, the current frame image is first read, and the fish swarm appearance detector is used to determine the potential fish swarm target position in the image. For the next image frame, the local search area is obtained by image cropping technology around the central area of the fish swarm target detected by the appearance detector; at the same time, the above-mentioned frame difference method is used to obtain the fish swarm contour, and the local image and the segmented motion target contour are fused as the input of the motion target detector, and the fish swarm motion target detector is activated. The farmed fish swarm detector of the dual YOLOv9 network that integrates the fish swarm appearance and motion information adopts the "global-local" target detection idea to improve the algorithm reasoning speed and detection success rate.
[0063] In some examples, this also includes:
[0064] Label the underwater bait residue image dataset in the underwater image data, and use labelimg to label the bait residue targets;
[0065] Modify the feature encoding subnetwork connection mode to generate a fine-grained feature output layer;
[0066] Enhance feature transfer and reuse, and build feature fusion and feature enhancement modules;
[0067] The number of network layers is reduced, and some residual connection modules are replaced with dense connection modules to obtain a lightweight deep learning model with hierarchical feature fusion and enhancement.
[0068] For example, based on the prior information of small particles and spatial distribution of underwater bait residues, the open source labeling tool labelimg is used to annotate the targets in the bait image dataset. In combination with the characteristics of weak contrast, small targets and high density of underwater bait residues, this project intends to make the following improvements to the lightweight target detection model YOLOv8 to adapt to the underwater bait residue detection task. First, in order to solve the problem of too many small targets, the project intends to modify the connection mode of the feature encoding subnetwork so that the network can retain more shallow features, generate feature output layers that are more conducive to small target detection, obtain richer fine-grained information, and discard the feature output layers responsible for large target detection that are useless to the dataset. Secondly, in order to speed up the training speed, the residual network is densely connected, and it is planned to strengthen feature transfer and reuse, build feature fusion and feature enhancement modules, and focus on learning features that are more related to the detection task while solving the gradient vanishing problem in the training process of the dataset; in addition, in order to improve the calculation speed, the project intends to reduce the number of network layers and replace some residual connection modules with dense connection modules to reduce feature redundancy.
[0069] In some examples, this also includes:
[0070] According to the physiological and behavioral characteristics of fish during feeding, combined with the bait residue in the cage, a feeding intensity label is developed;
[0071] The time series features are extracted using the empirical mode decomposition method and input into the multi-scale CNN and LSTM models for training and judgment to obtain the feeding energy model.
[0072] For example, based on the physiological and behavioral characteristics of fish during feeding, combined with the bait residue in the cage, four feeding intensities of 'none', 'weak', 'medium' and 'strong' are formulated as labels for network training. According to the different feeding intensities, the three time series of water temperature, dissolved oxygen concentration and ammonia nitrogen compounds collected by the water quality sensor are divided into different segments, and the segments corresponding to the same label are composed of time series vectors, which are normalized and preprocessed; for the preprocessed time series, the empirical mode decomposition method is used to extract the time series features, and the features are input into the deep learning model composed of multi-scale CNN and LSTM to realize the extraction of deep-level features of the time series, model training and feeding intensity judgment. In the network prediction results, the higher the feeding intensity, the lower the energy demand of the fish.
[0073] In some examples, the step of evaluating the feeding status of cage-cultured fish by combining the identified fish feeding behavior information and the detected residual bait information includes:
[0074] Quantify the information of fish feeding behavior, residual bait and feeding energy requirement into feeding characteristic indicators;
[0075] The quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of fish schools.
[0076] In some examples, the quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of the fish school, including:
[0077] The support vector regression model is used to determine the weight of each feature, and the XGBoost ensemble learning algorithm is used to learn and predict the feeding status of fish schools in order to comprehensively evaluate the feeding status of fish schools.
[0078] For example, the target information of fish and bait residue detected by the above model is used to obtain the number and aggregation index of the target using high-order moments and Delaunay triangulation profile analysis, and the average swimming speed and direction of the fish are calculated using the frame difference method and optical flow method; the feeding energy prediction label of the water quality information feedback is fuzzy processed using the membership function. After integrating the quantified indicators, the support vector regression model based on grid search is used to determine the weight of each feature, and the XGBoost ensemble learning algorithm is used to realize the learning and prediction of the feeding status of the fish.
[0079] See also Figure 2 , an embodiment of the feed supplement assessment device for cage fish farming in the embodiment of the present application may include:
[0080] An identification unit 21 is used to identify the feeding behavior of a school of fish based on the underwater image data through a school of fish behavior identification model, wherein the school of fish behavior identification model is a deep learning model that integrates the appearance and movement information of the school of fish;
[0081] A detection unit 22, used for detecting residual bait in aquaculture boxes based on underwater image data through a hierarchical feature fusion and enhanced lightweight deep learning model;
[0082] The evaluation unit 23 is used to predict the fish feeding energy demand information through the feeding energy model according to the changes in water quality parameters, so as to evaluate the feeding status of the cage cultured fish in combination with the identified fish feeding behavior information and the detected residual bait information.
[0083] In summary, the feed supplementation assessment device for caged fish farming provided in the embodiment of the present application recognizes the feeding behavior of fish schools through a fish school behavior recognition model based on underwater image data, wherein the fish school behavior recognition model is a deep learning model that integrates fish school appearance and motion information; detects residual bait in the farming box through a lightweight deep learning model that integrates hierarchical feature fusion and enhancement based on underwater image data; predicts the energy demand information of fish school feeding through a feeding energy model according to changes in water quality parameters, so as to evaluate the feeding status of caged fish schools by combining the identified fish school feeding behavior information and the detected residual bait information. In this way, multi-source information is integrated to provide accurate feeding status assessment results. Feeding strategies are optimized based on the assessment results to improve farming efficiency and benefits.
[0084] like Figure 3 As shown, the embodiment of the present application further provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 320 and executable on the processor. When the processor 320 executes the computer program 311, the steps of any of the above-mentioned methods for evaluating feed supplementation for fish farming in cages are implemented:
[0085] Based on underwater image data, fish school feeding behavior is identified through fish school behavior recognition, wherein the fish school behavior recognition model is a deep learning model that integrates fish school appearance and movement information;
[0086] Residual bait detection in aquaculture boxes based on underwater image data through hierarchical feature fusion and enhanced lightweight deep learning model;
[0087] According to the changes in water quality parameters, the feeding energy demand information of fish is predicted through the feeding energy model, and the feeding status of cage-cultured fish is evaluated by combining the identified fish feeding behavior information and the detected residual bait information.
[0088] Specifically, it also includes:
[0089] Based on the physical model of underwater optical imaging, the underwater transmittance image of underwater image data is obtained according to the dark channel prior, RGB three-channel color difference and spatial texture constraints;
[0090] A hierarchical feature fusion strategy is introduced to perform cross-modal fusion of RGB features and transmittance features;
[0091] Combined with the U-shaped network architecture, a decoding module is designed to restore the image resolution step by step and restore the image details from coarse to fine.
[0092] Specifically, it also includes:
[0093] Use labelimg and labelme to roughly classify and label the objects in the underwater image data set, and divide it into training set, validation set and test set;
[0094] Perform projection transformation on multiple frames of images and perform image alignment;
[0095] Use multi-frame difference method to segment the outline of the moving target of the school of fish;
[0096] Combined with the contour information of the fish school target, the Shi-Tomas algorithm is used to extract the corner points of the moving target, and the classifier is designed to eliminate the interference of non-fish school targets;
[0097] A fish school appearance feature YOLOv9 network detector based on a single-frame image and a fish school motion target YOLOv9 network detector based on a time series are designed respectively to obtain the fish school behavior recognition model.
[0098] Specifically, it also includes:
[0099] Label the underwater bait residue image dataset in the underwater image data, and use labelimg to label the bait residue targets;
[0100] Modify the feature encoding subnetwork connection mode to generate a fine-grained feature output layer;
[0101] Enhance feature transfer and reuse, and build feature fusion and feature enhancement modules;
[0102] The number of network layers is reduced, and some residual connection modules are replaced with dense connection modules to obtain a lightweight deep learning model with hierarchical feature fusion and enhancement.
[0103] Specifically, it also includes:
[0104] According to the physiological and behavioral characteristics of fish during feeding, combined with the bait residue in the cage, a feeding intensity label is developed;
[0105] The time series features are extracted using the empirical mode decomposition method and input into the multi-scale CNN and LSTM models for training and judgment to obtain the feeding energy model.
[0106] Specifically, the method of evaluating the feeding state of cage-cultured fish by combining the identified fish feeding behavior information and the detected residual bait information includes:
[0107] Quantify the information of fish feeding behavior, residual bait and feeding energy requirement into feeding characteristic indicators;
[0108] The quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of fish schools.
[0109] Specifically, the quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of the fish school, including:
[0110] The support vector regression model is used to determine the weight of each feature, and the XGBoost ensemble learning algorithm is used to learn and predict the feeding status of fish schools in order to comprehensively evaluate the feeding status of fish schools.
[0111] Since the electronic device introduced in this embodiment is a device used to implement a cage fish farming feed supplementation evaluation device in the embodiment of the present application, based on the method introduced in the embodiment of the present application, the technical personnel in this field can understand the specific implementation mode of the electronic device of this embodiment and its various variations. Therefore, how the electronic device implements the method in the embodiment of the present application is not introduced in detail here. As long as the equipment used by the technical personnel in this field to implement the method in the embodiment of the present application is within the scope of protection of this application.
[0112] In the specific implementation process, when the computer program 311 is executed by the processor, it can achieve Figure 1 Any implementation manner in the corresponding embodiment:
[0113] Based on the underwater image data, fish feeding behavior is identified by a fish behavior identification model, wherein the fish behavior identification model is a deep learning model that integrates fish appearance and movement information;
[0114] Residual bait detection in aquaculture boxes based on underwater image data through hierarchical feature fusion and enhanced lightweight deep learning model;
[0115] According to the changes in water quality parameters, the feeding energy demand information of fish is predicted through the feeding energy model, and the feeding status of cage-cultured fish is evaluated by combining the identified fish feeding behavior information and the detected residual bait information.
[0116] Specifically, it also includes:
[0117] Based on the physical model of underwater optical imaging, the underwater transmittance image of underwater image data is obtained according to the dark channel prior, RGB three-channel color difference and spatial texture constraints;
[0118] A hierarchical feature fusion strategy is introduced to perform cross-modal fusion of RGB features and transmittance features;
[0119] Combined with the U-shaped network architecture, a decoding module is designed to restore the image resolution step by step and restore the image details from coarse to fine.
[0120] Specifically, it also includes:
[0121] Use labelimg and labelme to roughly classify and label the objects in the underwater image data set, and divide it into training set, validation set and test set;
[0122] Perform projection transformation on multiple frames of images and perform image alignment;
[0123] Use multi-frame difference method to segment the outline of the moving target of the school of fish;
[0124] Combined with the contour information of the fish school target, the Shi-Tomas algorithm is used to extract the corner points of the moving target, and the classifier is designed to eliminate the interference of non-fish school targets;
[0125] A fish school appearance feature YOLOv9 network detector based on a single-frame image and a fish school motion target YOLOv9 network detector based on a time series are designed respectively to obtain the fish school behavior recognition model.
[0126] Specifically, it also includes:
[0127] Label the underwater bait residue image dataset in the underwater image data, and use labelimg to label the bait residue targets;
[0128] Modify the feature encoding subnetwork connection mode to generate a fine-grained feature output layer;
[0129] Enhance feature transfer and reuse, and build feature fusion and feature enhancement modules;
[0130] The number of network layers is reduced, and some residual connection modules are replaced with dense connection modules to obtain a lightweight deep learning model with hierarchical feature fusion and enhancement.
[0131] Specifically, it also includes:
[0132] According to the physiological and behavioral characteristics of fish during feeding, combined with the bait residue in the cage, a feeding intensity label is developed;
[0133] The time series features are extracted using the empirical mode decomposition method and input into the multi-scale CNN and LSTM models for training and judgment to obtain the feeding energy model.
[0134] Specifically, the method of evaluating the feeding state of cage-cultured fish by combining the identified fish feeding behavior information and the detected residual bait information includes:
[0135] Quantify the information of fish feeding behavior, residual bait and feeding energy requirement into feeding characteristic indicators;
[0136] The quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of fish schools.
[0137] Specifically, the quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of the fish school, including:
[0138] The support vector regression model is used to determine the weight of each feature, and the XGBoost ensemble learning algorithm is used to learn and predict the feeding status of fish schools in order to comprehensively evaluate the feeding status of fish schools.
[0139] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0140] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0141] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0142] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0143] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0144] The present application also provides a computer program product, which includes computer software instructions. When the computer software instructions are executed on a processing device, the processing device is caused to execute the following Figure 1 The process of evaluating feed supplementation for cage fish farming in the corresponding embodiment.
[0145] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a server or a data center that includes one or more available media integration. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state drive (SSD)), etc.
[0146] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0147] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.
[0148] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0149] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0150] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.
[0151] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for evaluating feed supplementation for caged fish farming, characterized in that: include: Based on the underwater image data, the fish school feeding behavior is identified by a fish school behavior identification model, wherein the fish school behavior identification model is a deep learning model that integrates the fish school appearance and movement information; Residual bait detection in aquaculture boxes based on underwater image data through hierarchical feature fusion and enhanced lightweight deep learning model; According to the changes in water quality parameters, the feeding energy demand information of fish is predicted through the feeding energy model, and the feeding status of cage-cultured fish is evaluated by combining the identified fish feeding behavior information and the detected residual bait information.
2. The method according to claim 1, characterized in that Also includes: Based on the physical model of underwater optical imaging, the underwater transmittance image of underwater image data is obtained according to the dark channel prior, RGB three-channel color difference and spatial texture constraints; A hierarchical feature fusion strategy is introduced to perform cross-modal fusion of RGB features and transmittance features; Combined with the U-shaped network architecture, a decoding module is designed to restore the image resolution step by step and restore the image details from coarse to fine.
3. The method according to claim 1, characterized in that Also includes: Use labelimg and labelme to roughly classify and label the objects in the underwater image data set, and divide it into training set, validation set and test set; Perform projection transformation on multiple frames of images and perform image alignment; Use multi-frame difference method to segment the outline of the moving target of the school of fish; Combined with the contour information of the fish school target, the Shi-Tomas algorithm is used to extract the corner points of the moving target, and the classifier is designed to eliminate the interference of non-fish school targets; A fish school appearance feature YOLOv9 network detector based on a single-frame image and a fish school motion target YOLOv9 network detector based on a time series are designed respectively to obtain the fish school behavior recognition model.
4. The method according to claim 1, characterized in that Also includes: Label the underwater bait residue image dataset in the underwater image data, and use labelimg to label the bait residue targets; Modify the feature encoding subnetwork connection mode to generate a fine-grained feature output layer; Enhance feature transfer and reuse, and build feature fusion and feature enhancement modules; The number of network layers is reduced, and some residual connection modules are replaced with dense connection modules to obtain a lightweight deep learning model with hierarchical feature fusion and enhancement.
5. The method according to claim 1, characterized in that Also includes: According to the physiological and behavioral characteristics of fish during feeding, combined with the bait residue in the cage, a feeding intensity label is developed; The time series features are extracted using the empirical mode decomposition method and input into the multi-scale CNN and LSTM models for training and judgment to obtain the feeding energy model.
6. The method according to claim 1, characterized in that The method of evaluating the feeding status of fish in cage culture by combining the identified fish feeding behavior information and the detected residual bait information includes: Quantify the information of fish feeding behavior, residual bait and feeding energy requirement into feeding characteristic indicators; The quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of fish schools.
7. The method according to claim 6, characterized in that The quantified feeding characteristic indicators are integrated to comprehensively evaluate the feeding status of the fish school, including: The support vector regression model is used to determine the weight of each feature, and the XGBoost ensemble learning algorithm is used to learn and predict the feeding status of fish schools in order to comprehensively evaluate the feeding status of fish schools.
8. A feed supplementation assessment device for cage fish farming, characterized in that: include: An identification unit, used for identifying the feeding behavior of a school of fish by using a school of fish behavior identification model based on the underwater image data, wherein the school of fish behavior identification model is a deep learning model integrating the appearance and movement information of the school of fish; A detection unit, used to detect residual bait in aquaculture boxes based on underwater image data through hierarchical feature fusion and enhanced lightweight deep learning model; The evaluation unit is used to predict the fish feeding energy demand information through the feeding energy model according to the changes in water quality parameters, so as to evaluate the feeding status of the cage-cultured fish in combination with the identified fish feeding behavior information and the detected residual bait information.
9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the method for evaluating feed supplementation for cage fish farming as described in any one of claims 1 to 7 when executing the computer program stored in the memory.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for evaluating feed supplementation for cage fish farming according to any one of claims 1 to 7 is implemented.
Citation Information
Patent Citations
Method and device for evaluating ingestion intensity of fish school
CN114419432A
Intelligent bait casting boat
CN114467824A
Intelligent fish feeding method and system based on multi-factor comprehensive control
CN114847209A
Aquatic animal cultivation assisting system, lift device, feeding device, aquatic animal cultivation method, and aquatic animal cultivation assisting program
WO2020250330A1
Cited By
Multi-dimensional feature collaborative quantitative recognition method and model for fish ingestion intensity
CN120014507A
Multi-dimensional feature cooperative fish feeding intensity quantification identification method and model
CN120014507B
Fish feeding method and device based on multi-task active learning framework
CN120108046A
Typhoon disaster prevention and after-disaster rapid recovery method for cage culture in deep and far sea
CN120323377A
