Intelligent acquisition system and method based on automatic measurement of fish growth data

Abstract

The embodiment of the invention discloses an intelligent collection system and method based on automatic measurement of fish growth data, and the system comprises a data collection module which is used for collecting fish image data, fish sound data and breeding environment data; the data transmission module is used for stably transmitting the data acquired by the sensing layer module to a data processing center through a wireless transmission technology; the data processing module is used for processing and analyzing the transmitted data and realizing fish target positioning segmentation, feature point detection, growth parameter calculation, weight estimation and growth cycle prediction; and the data management module is used for performing associative storage and visual display on the processing result and automatically generating an analysis report. According to the invention, the automation of fish growth data acquisition and analysis is realized, and the precision is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aquaculture monitoring, and relates to but is not limited to an intelligent collection system and method based on automatic measurement of fish growth data. BACKGROUND

[0002] In the global fishery industry, fish farming occupies an important position. With the continuous rise of market demand for aquatic products, large-scale farming has become an inevitable trend of fishery development. Real-time and accurate grasp of the growth status of farmed fish is of great significance to improve farming efficiency and ensure sustainable development of the fishery industry. The size of fish, such as body length, body height, and body weight, is a key indicator of its growth trend.

[0003] In the prior art, fish size monitoring mainly relies on sampling and manual measurement, but this method is extremely inefficient and cannot meet the high-throughput needs of large-scale farms. Moreover, this measurement method is highly subjective and the measurement results are easily affected by the experience and state of the measurer. Furthermore, this method can cause stress damage to fish, affecting their normal growth and development and even causing death. In addition, although machine vision-based measurement technology can reduce human intervention to some extent, it still has the following defects: strict environmental requirements, which can easily lead to a decrease in measurement accuracy due to uneven lighting, turbid water, and other interference factors; single data dimension, which cannot fully reflect the growth status of fish; and low automation, which requires manual image screening and parameter calibration.

[0004] Therefore, there is an urgent need for a more comprehensive intelligent fish growth data collection system to address the related problems of insufficient measurement accuracy, single data dimension, and low automation in the prior art, to greatly improve the automation and precision of fish growth data collection and analysis, and to improve farming efficiency and ensure sustainable development of the fishery industry. SUMMARY

[0005] The present application provides an intelligent collection system and method based on automatic measurement of fish growth data.

[0006] The technical solution of the present application is as follows: The embodiment of the application provides an intelligent acquisition system based on automatic fish growth data acquisition, which comprises a data acquisition module, a data transmission module, a data processing module and a data management module, wherein: the data acquisition module is used for acquiring and preprocessing original multi-source heterogeneous data in the fish growth process to obtain multi-source heterogeneous data, the multi-source heterogeneous data comprising fish image data, fish acoustic data and breeding environment data; the data transmission module is used for transmitting the multi-source heterogeneous data to a data processing center by combining 5G, Wi-Fi 6 and LoRa wireless transmission technology; the data processing module is used for target positioning segmentation of the fish image data through a Mask R-CNN model to generate a fish segmentation mask, feature point detection of the fish acoustic data to generate an acoustic feature matrix, fusion of the fish segmentation mask and the acoustic feature matrix based on a Transformer model, and determination of fish behavior categories and confidence through a cross attention mechanism; determination of fish growth parameters through an HRNet model based on the fish segmentation mask, determination of fish weight prediction values through an improved LSTM model integrated with an attention mechanism based on the fish growth parameters and the acoustic feature matrix; prediction of fish growth cycle prediction values required to reach a specific growth stage through a multi-layer LSTM model based on the fish growth parameters, acoustic features corresponding to the fish behavior categories and breeding environment data; and the data management module is used for associated storage of processing whole-process data, real-time display of fish growth curves, fish weight change trends, fish behavior characteristic change trends and environmental parameter change information on a visual interface, and generation of automatic analysis reports based on preset periods and preset trigger conditions.

[0007] The technical solution provided in this application, in the data acquisition module, collects and preprocesses raw multi-source heterogeneous data during the fish growth process to obtain multi-source heterogeneous data, including fish image data, fish acoustic data, and aquaculture environment data. This achieves multi-dimensional acquisition of fish growth status, avoids stress damage caused by traditional manual measurement, and provides a high-quality data foundation for subsequent analysis, ensuring the comprehensiveness and reliability of the system's data source. In the data transmission module, the multi-source heterogeneous data is transmitted to the data processing center by combining 5G, Wi-Fi 6, and LoRa wireless transmission technologies. This ensures the stability and continuity of the data link in complex aquaculture environments, effectively prevents data loss, and meets the transmission requirements of high speed, low latency, and long coverage, ensuring the system's real-time performance. In the data processing module, Mask... The R-CNN model performs target localization and segmentation on the fish image data, generating a fish body segmentation mask. It then performs feature point detection on the fish acoustic data, generating an acoustic feature matrix. Based on the Transformer model, it fuses the fish body segmentation mask and the acoustic feature matrix, and uses a cross-attention mechanism to determine the fish behavior category and confidence level, achieving accurate fish body segmentation and identification of complex behaviors. This significantly improves the robustness of perception in real underwater environments. Furthermore, by complementing and associating visual and acoustic information, it enables automated and high-precision judgment of key behaviors such as fish feeding and stress. Based on the fish body segmentation mask, the HRNet model determines fish growth parameters, achieving automated and high-precision measurement of fish size. This overcomes the shortcomings of traditional methods, which rely on single parameters and manual intervention, providing a quantitative basis directly applicable to aquaculture management decisions. Based on the fish growth parameters and the acoustic feature matrix, an improved LSTM model with an integrated attention mechanism determines the predicted fish weight, achieving a seamless transition without interfering with other fish growth. High-precision weight estimation of fish can replace traditional, stress-intensive weighing methods. By integrating morphological and behavioral characteristics for prediction, it achieves more scientific weight estimation and significantly improves prediction accuracy. The model uses a multi-layer LSTM to predict the growth cycle required for fish to reach specific growth stages, based on fish growth parameters, acoustic characteristics corresponding to fish behavior categories, and aquaculture environment data. This enables scientific and forward-looking prediction of fish future growth trajectories. Furthermore, by quantifying environmental impacts and incorporating them into the model, it provides precise decision support for optimizing aquaculture strategies and mitigating risks. The data management module links and stores data throughout the entire process, ensuring the reliability and maintainability of the analysis results from raw data. A visual interface displays real-time fish growth curves, fish weight change trends, fish behavioral characteristic change trends, and environmental parameter changes. Based on preset cycles and trigger conditions, it generates automated analysis reports, transforming from passive querying to proactive intelligent decision support and greatly improving management efficiency.

[0008] Optionally, the data transmission module includes a high-speed transmission unit, a medium-short-range transmission unit, a backup transmission unit, and an intelligent routing decision unit, wherein: the high-speed transmission unit is used to transmit the fish image data to the cloud server in real time using 5G wireless transmission technology; the medium-short-range transmission unit is used to transmit the fish acoustic data and the aquaculture environment data at high speed over short distances using Wi-Fi 6 wireless transmission technology; the backup transmission unit is used to aggregate multi-source heterogeneous data for transmission using LoRa wireless transmission technology, providing redundancy backup for the high-speed transmission unit and the medium-short-range transmission unit; and the intelligent routing decision unit is used to monitor the network status in real time and automatically switch between 5G, Wi-Fi 6, and LoRa wireless transmission technologies based on the monitoring results.

[0009] Optionally, the data processing module includes a visual feature extraction unit, an acoustic feature extraction unit, a growth parameter calculation unit, and a weight prediction unit, wherein: the visual feature extraction unit is used to perform adaptive bilateral filtering and illumination correction on the fish image data, locate the fish target using the YOLOv8 model to obtain bounding boxes, and use the bounding boxes as candidate regions, and then apply a mask... The R-CNN model performs instance segmentation and outputs a pixel-level fish body segmentation mask. The acoustic feature extraction unit is used to perform frame segmentation and feature point detection and extraction on the fish acoustic data, generate an acoustic feature matrix, and fuse the fish body segmentation mask and the acoustic feature matrix based on the Transformer model. The visual modality and acoustic modality are associated through a cross-attention mechanism to obtain the fish behavior category and confidence level. The growth parameter calculation unit is used to detect predefined key fish feature points on the fish body segmentation mask using the HRNet model, and calculate the fish body length, fish body width, and fish total length based on the pixel coordinates of the key fish feature points and a preset pixel resolution. The fish body length is the straight-line distance from the snout to the end of the caudal fin, and the fish body width is the vertical distance from the highest point of the body height to the starting point of the pelvic fin. The weight prediction unit is used to standardize the fish growth parameters and the acoustic feature matrix, and then input them into an improved LSTM model with an integrated attention mechanism to obtain the predicted fish weight value.

[0010] Optionally, the data processing module further includes a growth cycle prediction unit, which is specifically used to: receive the time series data of fish growth parameters, acoustic features corresponding to the fish behavior categories, and aquaculture environment data, perform standardization processing, and fuse them to generate a time series dataset; input the time series dataset into a multi-layer LSTM model for prediction, and output the predicted growth cycle value required for the fish to reach a specific growth stage.

[0011] Optionally, the data management module includes a data association and storage unit, a visualization unit, and an automated report generation unit, wherein: the data association and storage unit is used to establish an original data table to store index information of the original multi-source heterogeneous data, and to establish a processing result table to store the data processing results. The processing result table is associated with the original data table through a unique batch identifier and an associated timestamp; the visualization unit is used to display fish growth curves, fish weight change trends, fish behavioral characteristic change trends, and environmental parameter change information in real time in a graphical form on a visualization interface; the automated report generation unit is used to receive abnormal signals in the data processing process in real time and automatically extract abnormal data to generate early warning reports; at a fixed time each day, it automatically calls the database to extract fish growth parameters, behavioral characteristics, and aquaculture environment data for the corresponding time period to generate a daily fish growth analysis report including growth rate analysis and health status assessment; at a fixed time each week, it automatically analyzes the growth parameter growth, behavioral characteristic frequency, and environmental data average for the week to generate a weekly growth cycle assessment report.

[0012] Secondly, embodiments of this application provide an intelligent data acquisition method based on automatically measuring fish growth data. The method is applied to an intelligent data acquisition system based on automatically measuring fish growth data. The system includes a data acquisition module, a data transmission module, a data processing module, and a data management module. The method includes: acquiring and preprocessing raw multi-source heterogeneous data during the fish growth process to obtain multi-source heterogeneous data, including fish image data, fish acoustic data, and aquaculture environment data; transmitting the multi-source heterogeneous data to a data processing center using a combination of 5G, Wi-Fi 6, and LoRa wireless transmission technologies; and transmitting the data via a mask. The R-CNN model performs target localization and segmentation on the fish image data, generating a fish body segmentation mask. It then performs feature point detection on the fish acoustic data, generating an acoustic feature matrix. Based on a Transformer model, it fuses the fish body segmentation mask and the acoustic feature matrix, and uses a cross-attention mechanism to determine the fish behavior category and confidence level. Based on the fish body segmentation mask, an HRNet model determines the fish growth parameters. Based on the fish growth parameters and the acoustic feature matrix, an improved LSTM model with an integrated attention mechanism determines the predicted fish weight. The model then uses a multi-layer LSTM model to predict the growth cycle required for the fish to reach a specific growth stage. All data from the entire processing flow is stored in association, and the fish growth curve, fish weight change trend, fish behavior characteristic change trend, and environmental parameter change information are displayed in real-time on a visualization interface. An automated analysis report is generated based on preset cycles and preset trigger conditions.

[0013] Thirdly, embodiments of this application provide an electronic device, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements the steps in the above-mentioned intelligent data acquisition method based on automatic measurement of fish growth data.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in the above-described intelligent data acquisition method based on automatically measuring fish growth data.

[0015] The beneficial effects of the technical solutions provided in this application include at least the following: This application provides an intelligent data acquisition system and method based on automatic measurement of fish growth data. In the data acquisition module, raw multi-source heterogeneous data from the fish growth process is collected and preprocessed to obtain multi-source heterogeneous data, including fish image data, fish acoustic data, and aquaculture environment data. This achieves multi-dimensional acquisition of fish growth status, avoiding the stress damage caused by traditional manual measurement, and providing a high-quality data foundation for subsequent analysis, ensuring the comprehensiveness and reliability of the system's data source. In the data transmission module, the multi-source heterogeneous data is transmitted to the data processing center by combining 5G, Wi-Fi 6, and LoRa wireless transmission technologies. This ensures the stability and continuity of the data link in complex aquaculture environments, effectively preventing data loss, and meeting the transmission requirements of high speed, low latency, and long coverage, ensuring system real-time performance. In the data processing module, Mask... The R-CNN model performs target localization and segmentation on the fish image data, generating a fish body segmentation mask. It then performs feature point detection on the fish acoustic data, generating an acoustic feature matrix. Based on the Transformer model, it fuses the fish body segmentation mask and the acoustic feature matrix, and uses a cross-attention mechanism to determine the fish behavior category and confidence level, achieving accurate fish body segmentation and identification of complex behaviors. This significantly improves the robustness of perception in real underwater environments. Furthermore, by complementing and associating visual and acoustic information, it enables automated and high-precision judgment of key behaviors such as fish feeding and stress. Based on the fish body segmentation mask, the HRNet model determines fish growth parameters, achieving automated and high-precision measurement of fish size. This overcomes the shortcomings of traditional methods, which rely on single parameters and manual intervention, providing a quantitative basis directly applicable to aquaculture management decisions. Based on the fish growth parameters and the acoustic feature matrix, an improved LSTM model with an integrated attention mechanism determines the predicted fish weight, achieving a seamless transition without interfering with other fish growth. High-precision weight estimation of fish can replace traditional, stress-intensive weighing methods. By integrating morphological and behavioral characteristics for prediction, it achieves more scientific weight estimation and significantly improves prediction accuracy. The model uses a multi-layer LSTM to predict the growth cycle required for fish to reach specific growth stages, based on fish growth parameters, acoustic characteristics corresponding to fish behavior categories, and aquaculture environment data. This enables scientific and forward-looking prediction of fish future growth trajectories. Furthermore, by quantifying environmental impacts and incorporating them into the model, it provides precise decision support for optimizing aquaculture strategies and mitigating risks. The data management module links and stores data throughout the entire process, ensuring the reliability and maintainability of the analysis results from raw data. A visual interface displays real-time fish growth curves, fish weight change trends, fish behavioral characteristic change trends, and environmental parameter changes. Based on preset cycles and trigger conditions, it generates automated analysis reports, transforming from passive querying to proactive intelligent decision support and greatly improving management efficiency. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 A schematic diagram of an intelligent data acquisition system based on automatic measurement of fish growth data provided in this application embodiment; Figure 2 A flowchart illustrating an intelligent data acquisition method based on automatic measurement of fish growth data, provided in this application embodiment; Figure 3 This is a schematic diagram of the hardware entity of an electronic device provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0019] It should be noted that the terms "first, second, and third" used in the embodiments of this application are merely to distinguish similar objects and do not represent a specific order of objects. It is understood that "first, second, and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0020] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments of this application pertain. It should also be understood that terms such as those defined in general dictionaries should be understood to have a meaning consistent with their meaning in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0021] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0022] In view of the current problems in the field of aquaculture monitoring technology for fish growth data collection, this application provides an intelligent data collection system and method based on automatic measurement of fish growth data.

[0023] The technical solution of this application is described below, starting with the system implementation of this application.

[0024] Please refer to Figure 1 The illustration shows a schematic diagram of an intelligent data acquisition system based on automatic measurement of fish growth data provided in an embodiment of this application. Figure 1 As shown, the system includes a data acquisition module 01, a data transmission module 02, a data processing module 03, and a data management module 04. The data acquisition module 01 is used to collect and preprocess raw multi-source heterogeneous data during the fish growth process to obtain multi-source heterogeneous data, which includes fish image data, fish acoustic data, and aquaculture environment data. The data transmission module 02 is used to transmit the multi-source heterogeneous data to the data processing center using a combination of 5G, Wi-Fi 6, and LoRa wireless transmission technologies. The data processing module 03 is used to transmit the multi-source heterogeneous data via a mask... The R-CNN model performs target localization and segmentation on the fish image data, generating a fish body segmentation mask. It then performs feature point detection on the fish acoustic data, generating an acoustic feature matrix. Based on the Transformer model, it fuses the fish body segmentation mask and the acoustic feature matrix, and uses a cross-attention mechanism to determine the fish behavior category and confidence level. Based on the fish body segmentation mask, the HRNet model determines the fish growth parameters. Based on the fish growth parameters and the acoustic feature matrix, an improved LSTM model with an integrated attention mechanism determines the predicted fish weight. The fish growth parameters, the acoustic features corresponding to the fish behavior categories, and the aquaculture environment data are then used to predict the growth cycle required for the fish to reach a specific growth stage using a multi-layer LSTM model. The data management module 04 is used to correlate and store the data throughout the entire processing flow, displaying the fish growth curve, fish weight change trend, fish behavior characteristic change trend, and environmental parameter change information in real time on a visualization interface, and generating automated analysis reports based on preset cycles and preset trigger conditions.

[0025] In this embodiment, the data acquisition module 01 forms a fusion monitoring network by deploying multiple sensors to achieve comprehensive acquisition of fish growth data. Specifically, it includes multi-view high-definition cameras, multi-band acoustic sensors, and various environmental sensors. The multi-view high-definition cameras are deployed at the top, sides, and bottom of the fish farming pond to ensure the capture of clear images and videos of fish in different swimming postures. The cameras have a frame rate of over 30fps and feature autofocus, wide dynamic range, and high frame rate shooting capabilities to adapt to complex underwater lighting conditions and rapid fish movement, thereby providing high-quality raw data for subsequent analysis of fish image data. The multi-band acoustic sensor incorporates a multi-band signal acquisition device covering the 20Hz-20kHz frequency band to capture acoustic signals such as water flow disturbances during swimming, pecking sounds during feeding, and gill activity sounds during breathing. Furthermore, the multi-band acoustic sensor integrates a miniature signal amplification circuit and a digital filtering circuit, which can amplify weak acoustic signals while effectively suppressing environmental noise interference such as water flow and equipment vibration. The multi-band acoustic sensor is evenly installed on the inner wall, center of the bottom, and four corners of the aquaculture pond. Some sensors are suspended in the middle layer of the pond by buoyancy devices to ensure comprehensive coverage of different water layers where fish are active. The sensor collects fish acoustic signals in real time at a sampling frequency of 500Hz and converts the analog acoustic signals into digital signals through a built-in signal conversion device, simultaneously extracting characteristic parameters such as signal frequency, amplitude, and duration. The environmental sensors include temperature sensors, dissolved oxygen sensors, pH sensors, ammonia nitrogen sensors, and turbidity sensors. These sensors are evenly deployed in the aquaculture water to collect aquatic environmental parameters in real time. At the same time, meteorological sensors are deployed around the aquaculture ponds to monitor meteorological information such as ambient temperature, humidity, and light intensity, thereby providing data support for subsequent analysis of the impact of the environment on fish growth.

[0026] In this embodiment, the original multi-source heterogeneous data obtained from monitoring is preprocessed to obtain multi-source heterogeneous data including fish image data, fish acoustic data, and aquaculture environment data. Specifically, for the original image data, the Laplacian operator is used to calculate the edge gray-level variance and blurry images with variances below a preset threshold are removed. The number and overlap of fish in the original images are detected based on a lightweight YOLOv8nano model, and invalid images with no fish or excessive overlap are removed. Overexposed or underexposed images with abnormal lighting are removed by combining gray-level histograms. Finally, valid fish images containing 12 fish, with no obvious overlap and meeting the clarity standard are retained. For the original acoustic data, steady-state noise such as water flow and power frequency is removed by spectral subtraction. The validity of the data is judged based on dual thresholds of frequency band and amplitude. Signal frames containing the 100-5000Hz frequency band and meeting the short-term energy standard are retained. Five consecutive frames that meet the condition are determined to be valid segments, thereby removing acoustic data without features or with distortion. For environmental data, obvious faulty data is quickly removed through physical thresholds and constant value tests. Sudden abrupt changes are identified using the sliding window mean and standard deviation. Missing data is filled in using historical means to ensure data continuity and rationality. The preprocessed multi-source heterogeneous data carries a unique batch identifier and a unified collection timestamp, providing high-quality input for subsequent data transmission and processing.

[0027] In this embodiment, the data transmission module 02 constructs a multi-mode transmission network by integrating 5G, Wi-Fi 6, and LoRa wireless transmission technologies to transmit multi-source heterogeneous data to the data processing center, achieving reliable transmission of multi-source heterogeneous data. The data transmission module 02 includes a high-speed transmission unit, a medium-short-range transmission unit, a backup transmission unit, and an intelligent routing decision unit. The high-speed transmission unit uses 5G wireless transmission technology to transmit large amounts of high-definition fish image data and video data, leveraging the high bandwidth and low latency characteristics of 5G to ensure efficient transmission of visual data with high real-time requirements to the cloud server. The medium-short-range transmission unit uses Wi-Fi 6 wireless transmission technology to transmit fish acoustic data and aquaculture environment data at high speeds over short distances, meeting the concurrent data transmission requirements within medium-short distances. The backup transmission unit uses LoRa wireless transmission technology, utilizing its long-distance transmission capability and strong anti-interference ability as a transmission backup mechanism. In areas with weak network signals, it aggregates multi-source heterogeneous data for reliable transmission, providing redundancy backup for the 5G transmission of the high-speed transmission unit and the Wi-Fi 6 transmission of the medium-short-range transmission unit. The intelligent routing decision unit is used to monitor the network status parameters of each transmission link in real time, such as signal strength, transmission rate and packet loss rate, and automatically select the optimal transmission path among 5G, Wi-Fi 6 and LoRa wireless transmission technologies based on a preset decision algorithm. When any transmission technology experiences signal attenuation or network congestion, it can seamlessly switch to other transmission technologies. Through this adaptive, multi-backup reliable data transmission system, the continuity and integrity of data transmission can be ensured.

[0028] In this embodiment, the data processing module 03 includes a visual feature extraction unit, an acoustic feature extraction unit, a growth parameter calculation unit, and a weight prediction unit. Specifically, the visual feature extraction unit preprocesses and segments the acquired fish image data. First, adaptive bilateral filtering is applied to the fish image data, dynamically adjusting the filtering weights in the spatial and grayscale domains to effectively suppress particle noise caused by water scattering while preserving fish edge details. Second, based on Retinex theory, the fish image data is decomposed into illumination and reflection components. Logarithmic transformation is applied to the illumination component to suppress overexposed areas, and gamma correction is applied to the reflection component to enhance low-light details, thereby improving the contrast between the fish and the background by more than 25%. Third, the processed fish image data is input into the YOLOv8 model for fish target localization, outputting the bounding box coordinates and class probability of each fish target. Then, the detected bounding boxes are used as candidate regions, and a mask is applied... The R-CNN model is used for instance segmentation. This model uses ResNet-101-FPN as the backbone network for feature extraction. It generates a 7×7 feature map through the RoIAlign layer. The mask branch contains 4 convolutional layers and outputs a 28×28 binarized mask, finally obtaining a pixel-level fish body segmentation mask with a mask mIoU of over 0.85.

[0029] In this embodiment, the acoustic feature extraction unit performs frame-by-frame processing and feature point detection and extraction on the fish acoustic data. The frame length is 20ms, the overlap rate is 50%, and a Hanning window is added to each frame to suppress spectral leakage. Mel spectrogram features and temporal features of each frame are extracted. The Mel spectrogram is obtained through an 80-dimensional Mel filter bank transformation. The temporal features include short-time energy, zero-crossing rate, and spectral centroid, ultimately generating an acoustic feature matrix. Further, the fish segmentation mask and acoustic feature matrix are fused based on a Transformer model. This model includes a visual encoder and an acoustic encoder. The visual encoder flattens the fish segmentation mask and reduces its dimensionality to a 512-dimensional feature vector through three convolutional layers. The acoustic encoder transforms the acoustic feature matrix into a 512-dimensional feature sequence through two LSTM layers and a linear layer, thereby achieving the fusion of the fish segmentation mask and acoustic feature matrix. Finally, an 8-head multi-head attention mechanism is used in the cross-attention layer to achieve deep interaction between the visual and acoustic modalities, learn the correspondence between fish body movements and sound features, and finally output the probability distribution and confidence of five core fish behavior categories: feeding, normal swimming, stress swimming, stillness and breathing through a fully connected layer.

[0030] In this embodiment, in the growth parameter calculation unit, 12 predefined key fish feature points are detected on the fish body segmentation mask using the HRNet model. These feature points include the snout tip, the posterior edge of the gill cover, the start and end points of the dorsal fin, the start and end points of the caudal fin, the start and end points of the pelvic fin, the start and end points of the anal fin, the midpoint of the body length, and the highest point of the body height. The model input is uniformly the fish body segmentation mask. Features are extracted through parallel high-resolution convolutional layers, and 12 heatmaps are output. Each heatmap corresponds to the probability distribution of a feature point. The model training process uses the mean squared error loss function and performs 150 rounds of training using the Adam optimizer. The first 50 rounds freeze the first two layers of the backbone network, and the last 100 rounds unfreeze the entire network and use a cosine annealing strategy to adjust the learning rate. Finally, the average pixel error of the feature points is less than or equal to 3 pixels. The fish's body length, body width, and total length are calculated based on the pixel coordinates of the detected key feature points and the pre-calibrated pixel resolution. The fish's body length is the Euclidean distance from the snout to the end of the caudal fin, the fish's body width is the vertical distance from the highest point of the body height to the origin of the pelvic fin, and the total length of the fish takes into account the complete measurement of the caudal fin filament structure.

[0031] In this embodiment, in the weight prediction unit, growth parameters such as fish body length, fish body width, and total fish length, along with acoustic behavioral features, are standardized and then input into an improved LSTM model. This improved LSTM model includes a feature attention mechanism and a two-layer LSTM network. The feature attention mechanism dynamically assigns weights to different features; for example, the weight for body length in adult fish is 0.3, and the weight for feeding sound amplitude in juvenile fish is 0.25. The first layer of the LSTM network contains 64 hidden units, and the second layer contains 32 hidden units. After processing, the model finally obtains the predicted fish weight. The model training uses a sliding window method, predicting the weight on day 6 based on data from the past 5 days. It is trained for 300 epochs on a dataset containing 50,000 manually weighed labels, and the validation set's mean absolute percentage error is less than or equal to 5%.

[0032] In this embodiment, the growth cycle prediction unit receives and standardizes the time-series data of fish growth parameters, acoustic features corresponding to the fish behavior categories, and aquaculture environment data, and fuses them to generate a time-series dataset. This time-series dataset is then input into a multi-layer LSTM model for prediction. The model employs a three-layer stacked structure: the first layer has 64 hidden nodes to capture short-term dependencies, the second layer has 32 hidden nodes to extract medium-term patterns, and the third layer has 16 hidden nodes to mine long-term trends. Each layer is followed by a Dropout layer to prevent overfitting. The model training uses a sliding window method, with continuous data from the past 14 days as input features and the number of days required to reach marketable size as the target output. The Adam optimizer and mean squared error loss function are selected. Training stops when the validation set loss no longer decreases after 10 consecutive rounds. Finally, the model outputs the predicted growth cycle value required for fish to reach a specific growth stage, providing a scientific basis for aquaculture management decisions.

[0033] In this embodiment, the data management module includes a data association and storage unit, a visualization unit, and an automated report generation unit. The data association and storage unit establishes a raw data table to store index information of raw, multi-source heterogeneous data, such as image storage paths, acoustic waveform file paths, and raw environmental parameter values. The raw data table is associated with a unique batch identifier, acquisition timestamp, and device ID. A processing result table is established to store data processing results, such as fish growth parameters, predicted fish weight, fish behavior categories, and confidence levels. This processing result table is associated with the raw data table through a unique batch identifier and associated timestamp. This unique batch identifier is used throughout the entire process from data acquisition to data processing result generation to ensure the integrity and traceability of the data chain. The visualization unit displays fish growth curves, fish weight change trends, fish behavior characteristic change trends, and environmental parameter change information in real-time in a graphical format on a visualization interface. The automated report generation unit is used to receive abnormal signals in the data processing process in real time, and automatically extract abnormal data associated with the abnormal period after receiving the abnormal signal, generate an abnormal report with an early warning level, and push the report to the user terminal. At a fixed time every day, it automatically calls the database to extract fish growth parameters, behavioral characteristics and aquaculture environment data for the corresponding time period, and generates a daily fish growth analysis report including growth rate analysis and health status assessment. At a fixed time every week, it automatically analyzes the growth parameter growth, behavioral characteristic frequency and environmental data average of the week, and generates a weekly growth cycle assessment report.

[0034] In summary, the intelligent data acquisition system based on automatic fish growth data provided in this application embodiment collects and preprocesses raw multi-source heterogeneous data during the fish growth process in the data acquisition module, obtaining multi-source heterogeneous data, including fish image data, fish acoustic data, and aquaculture environment data. This achieves multi-dimensional acquisition of fish growth status, avoids the stress damage caused by traditional manual measurement, and provides a high-quality data foundation for subsequent analysis, ensuring the comprehensiveness and reliability of the system's data source. In the data transmission module, the multi-source heterogeneous data is transmitted to the data processing center by combining 5G, Wi-Fi 6, and LoRa wireless transmission technologies, ensuring the stability and continuity of the data link in complex aquaculture environments, effectively preventing data loss, and balancing the transmission requirements of high speed, low latency, and long coverage, ensuring the system's real-time performance. In the data processing module, Mask... The R-CNN model performs target localization and segmentation on the fish image data, generating a fish body segmentation mask. It then performs feature point detection on the fish acoustic data, generating an acoustic feature matrix. Based on the Transformer model, it fuses the fish body segmentation mask and the acoustic feature matrix, and uses a cross-attention mechanism to determine the fish behavior category and confidence level, achieving accurate fish body segmentation and identification of complex behaviors. This significantly improves the robustness of perception in real underwater environments. Furthermore, by complementing and associating visual and acoustic information, it enables automated and high-precision judgment of key behaviors such as fish feeding and stress. Based on the fish body segmentation mask, the HRNet model determines fish growth parameters, achieving automated and high-precision measurement of fish size. This overcomes the shortcomings of traditional methods, which rely on single parameters and manual intervention, providing a quantitative basis directly applicable to aquaculture management decisions. Based on the fish growth parameters and the acoustic feature matrix, an improved LSTM model with an integrated attention mechanism determines the predicted fish weight, achieving a seamless transition without interfering with other fish growth. High-precision weight estimation of fish can replace traditional, stress-intensive weighing methods. By integrating morphological and behavioral characteristics for prediction, it achieves more scientific weight estimation and significantly improves prediction accuracy. The model uses a multi-layer LSTM to predict the growth cycle required for fish to reach specific growth stages, based on fish growth parameters, acoustic characteristics corresponding to fish behavior categories, and aquaculture environment data. This enables scientific and forward-looking prediction of fish future growth trajectories. Furthermore, by quantifying environmental impacts and incorporating them into the model, it provides precise decision support for optimizing aquaculture strategies and mitigating risks. The data management module links and stores data throughout the entire process, ensuring the reliability and maintainability of the analysis results from raw data. A visual interface displays real-time fish growth curves, fish weight change trends, fish behavioral characteristic change trends, and environmental parameter changes. Based on preset cycles and trigger conditions, it generates automated analysis reports, transforming from passive querying to proactive intelligent decision support and greatly improving management efficiency.

[0035] The above is a description of the system embodiments of this application. Based on the foregoing embodiments, the method embodiments of this application are described below.

[0036] Please refer to Figure 2 The diagram illustrates a flowchart of an intelligent data acquisition method based on automatically measuring fish growth data, provided in an embodiment of this application. This method is applied to, for example... Figure 1 The diagram illustrates an intelligent data acquisition system based on automatically measuring fish growth data. For details not disclosed in the method embodiments, please refer to the system embodiments. The system includes a data acquisition module, a data transmission module, a data processing module, and a data management module. Figure 2 As shown, the method includes the following steps S210 to S240.

[0037] Step S210: Collect and preprocess the original multi-source heterogeneous data during the fish growth process to obtain multi-source heterogeneous data, which includes fish image data, fish acoustic data, and aquaculture environment data.

[0038] Step S220: The multi-source heterogeneous data is transmitted to the data processing center by combining 5G, Wi-Fi 6 and LoRa wireless transmission technologies.

[0039] In this embodiment, 5G wireless transmission technology is used to transmit the fish image data to the cloud server in real time; Wi-Fi 6 wireless transmission technology is used to transmit the fish acoustic data and the aquaculture environment data at high speed over short distances; LoRa wireless transmission technology is used to aggregate multi-source heterogeneous data for transmission, providing redundant backup for the high-speed transmission unit and the medium-short distance transmission unit; the network status is monitored in real time and the 5G, Wi-Fi 6 and LoRa wireless transmission technologies are automatically switched based on the monitoring results.

[0040] Step S230: The fish image data is segmented and localized using the Mask R-CNN model to generate a fish body segmentation mask. Feature points are detected in the fish acoustic data to generate an acoustic feature matrix. The fish body segmentation mask and the acoustic feature matrix are fused using a Transformer model, and the fish behavior category and confidence level are determined using a cross-attention mechanism. Based on the fish body segmentation mask, the fish growth parameters are determined using an HRNet model. Based on the fish growth parameters and the acoustic feature matrix, the fish weight prediction value is determined using an improved LSTM model with an integrated attention mechanism. The fish growth parameters, the acoustic features corresponding to the fish behavior categories, and the aquaculture environment data are predicted using a multi-layer LSTM model to determine the predicted growth cycle value required for the fish to reach a specific growth stage.

[0041] In this embodiment, the fish image data undergoes adaptive bilateral filtering and illumination correction. Fish target localization is performed using a YOLOv8 model to obtain bounding boxes. These bounding boxes are used as candidate regions, and instance segmentation is performed using a Mask R-CNN model to output pixel-level fish body segmentation masks. The fish acoustic data is framed and feature point detection is performed to generate an acoustic feature matrix. The fish body segmentation mask and the acoustic feature matrix are fused using a Transformer model. A cross-attention mechanism is used to associate the visual and acoustic modalities to obtain the fish behavior category and confidence level. A predefined key fish feature point is detected on the fish body segmentation mask using an HRNet model. The fish body length, fish body width, and total length are calculated based on the pixel coordinates of the key fish feature points and a preset pixel resolution. The fish body length is the straight-line distance from the snout to the end of the caudal fin, and the fish body width is the vertical distance from the highest point of the body height to the starting point of the pelvic fin. The fish growth parameters and the acoustic feature matrix are standardized and then input into an improved LSTM model with an integrated attention mechanism to obtain the predicted fish weight.

[0042] In this embodiment of the application, the fish growth parameters, acoustic features corresponding to the fish behavior categories, and aquaculture environment data of the time series are received, standardized, and fused to generate a time series dataset; the time series dataset is input into a multi-layer LSTM model for prediction, and the predicted growth cycle value required for the fish to reach a specific growth stage is output.

[0043] Step S240: The data of the entire process is associated and stored, and the fish growth curve, fish weight change trend, fish behavior characteristic change trend and environmental parameter change information are displayed in real time on the visualization interface. An automated analysis report is generated based on the preset cycle and preset trigger conditions.

[0044] In this embodiment, an original data table is established to store the index information of the original multi-source heterogeneous data, and a processing result table is established to store the data processing results. The processing result table is associated with the original data table through a unique batch identifier and an associated timestamp. The fish growth curve, fish weight change trend, fish behavior characteristic change trend, and environmental parameter change information are displayed in real time in a visualization interface in a graphical form. Abnormal signals during the data processing process are received in real time and abnormal data is automatically extracted to generate an early warning report. At a fixed time each day, the database is automatically called to extract fish growth parameters, behavioral characteristics, and aquaculture environment data for the corresponding time period to generate a daily fish growth analysis report including growth rate analysis and health status assessment. At a fixed time each week, the growth parameter growth, behavioral characteristic frequency, and environmental data mean of the week are automatically analyzed to generate a weekly growth cycle assessment report.

[0045] It should be noted that, in the embodiments of this application, if the above-mentioned intelligent data acquisition method based on automatic measurement of fish growth data is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0046] Correspondingly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the steps of the intelligent data acquisition method based on automatically measuring fish growth data described in any of the above embodiments. Correspondingly, embodiments of this application also provide a computer program product. When executed by a processor of an electronic device, the computer program product is used to implement the steps of the intelligent data acquisition method based on automatically measuring fish growth data described in any of the above embodiments.

[0047] Based on the same technical concept, this application provides an electronic device for implementing an intelligent data acquisition method based on automatically measuring fish growth data as described in the above method embodiments. Figure 3 This is a hardware entity diagram of an electronic device provided in an embodiment of this application, such as... Figure 3 As shown, the electronic device 300 includes a memory 310 and a processor 320. The memory 310 stores a computer program that can run on the processor 320. When the processor 320 executes the program, it implements the steps in any of the intelligent data acquisition methods based on automatic measurement of fish growth data described in the embodiments of this application.

[0048] The memory 310 is configured to store instructions and applications executable by the processor 320, and can also cache data to be processed or already processed by the processor 320 and various modules in the electronic device (e.g., image data, audio data, voice communication data and video communication data), which can be implemented by flash memory or random access memory (RAM).

[0049] When the processor 320 executes the program, it implements the steps of an intelligent data acquisition method based on automatically measuring fish growth data, as described above. The processor 320 typically controls the overall operation of the electronic device 300.

[0050] The aforementioned processor can be at least one of the following: Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), Controller, Microcontroller, and Microprocessor. It is understood that other electronic devices can also implement the functions of the aforementioned processor, and this application does not specifically limit the specific implementation.

[0051] The aforementioned computer storage media / memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM), etc.; or it can be various electronic devices that include one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0052] It should be noted that the descriptions of the storage medium and device embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0053] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0054] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0055] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0056] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0057] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0058] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause the device automatic test line to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0059] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0060] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0061] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent data acquisition system based on automatic measurement of fish growth data, characterized in that, The system includes: The data acquisition module is used to collect and preprocess raw multi-source heterogeneous data during the fish growth process to obtain multi-source heterogeneous data, which includes fish image data, fish acoustic data and aquaculture environment data. The data transmission module is used to transmit the multi-source heterogeneous data to the data processing center by combining 5G, Wi-Fi 6 and LoRa wireless transmission technologies; The data processing module is used to perform target localization and segmentation on the fish image data using a Mask R-CNN model to generate a fish body segmentation mask; to perform feature point detection on the fish acoustic data to generate an acoustic feature matrix; to fuse the fish body segmentation mask and the acoustic feature matrix based on a Transformer model; and to determine the fish behavior category and confidence level through a cross-attention mechanism. Based on the fish body segmentation mask, the module uses an HRNet model to determine fish growth parameters; and based on the fish growth parameters and the acoustic feature matrix, the module uses an improved LSTM model with an integrated attention mechanism to determine the predicted fish weight. Finally, the module uses a multi-layer LSTM model to predict the growth cycle required for the fish to reach a specific growth stage. The data management module is used to associate and store data throughout the entire processing process. It displays fish growth curves, fish weight change trends, fish behavioral characteristic change trends, and environmental parameter change information in real time on a visual interface, and generates automated analysis reports based on preset cycles and preset trigger conditions.

2. The system according to claim 1, characterized in that, The data transmission module includes a high-speed transmission unit, a medium-to-short-distance transmission unit, a guaranteed transmission unit, and an intelligent routing decision unit, wherein: The high-speed transmission unit is used to transmit the fish image data to the cloud server in real time using 5G wireless transmission technology. The short-to-medium range transmission unit is used to transmit the fish acoustic data and the aquaculture environment data at high speed over short distances using Wi-Fi 6 wireless transmission technology; The backup transmission unit is used to aggregate heterogeneous data from multiple sources using LoRa wireless transmission technology and then transmit it, providing redundant backup for the high-speed transmission unit and the medium-short distance transmission unit. The intelligent routing decision unit is used to monitor network status in real time and automatically switch between 5G, Wi-Fi 6 and LoRa wireless transmission technologies based on the monitoring results.

3. The system according to claim 1, characterized in that, The data processing module includes a visual feature extraction unit, an acoustic feature extraction unit, a growth parameter calculation unit, and a weight prediction unit, wherein: The visual feature extraction unit is used to perform adaptive bilateral filtering and illumination correction on the fish image data, locate the fish target using the YOLOv8 model to obtain bounding boxes, use the bounding boxes as candidate regions, perform instance segmentation using the Mask R-CNN model, and output pixel-level fish body segmentation masks. The acoustic feature extraction unit is used to perform frame segmentation and feature point detection and extraction on the fish acoustic data, generate an acoustic feature matrix, and fuse the fish body segmentation mask and the acoustic feature matrix based on the Transformer model. The visual modality and acoustic modality are associated through a cross-attention mechanism to obtain the fish behavior category and confidence level. The growth parameter calculation unit is used to detect predefined key fish feature points on the fish body segmentation mask using the HRNet model, and calculate the fish body length, fish body width and fish total length based on the pixel coordinates of the key fish feature points and the preset pixel resolution. The fish body length is the straight-line distance from the snout to the end of the caudal fin, and the fish body width is the vertical distance from the highest point of the body height to the starting point of the pelvic fin. The weight prediction unit is used to standardize the fish growth parameters and the acoustic feature matrix, and then input them into the improved LSTM model with integrated attention mechanism to obtain the predicted fish weight value.

4. The system according to claim 1, characterized in that, The data processing module further includes a growth cycle prediction unit, which is specifically used for: The system receives the fish growth parameters, acoustic features corresponding to the fish behavior categories, and aquaculture environment data from the time series, performs standardization processing, and merges them to generate a time series dataset. The time series dataset is input into a multi-layer LSTM model for prediction, and the predicted growth cycle value required for fish to reach a specific growth stage is output.

5. The system according to claim 1, characterized in that, The data management module includes a data association and storage unit, a visualization display unit, and an automated report generation unit, wherein: The data association storage unit is used to establish an index information for storing original multi-source heterogeneous data in the original data table, and to establish a processing result table to store data processing results. The processing result table is associated with the original data table through a unique batch identifier and an associated timestamp. The visualization unit is used to display fish growth curves, fish weight change trends, fish behavioral characteristic change trends, and environmental parameter change information in real time in a graphical form on the visualization interface. The automated report generation unit is used to receive abnormal signals in the data processing process in real time and automatically extract abnormal data to generate early warning reports; at fixed times every day, it automatically calls the database to extract fish growth parameters, behavioral characteristics and aquaculture environment data for the corresponding time period to generate a daily fish growth analysis report including growth rate analysis and health status assessment; at fixed times every week, it automatically analyzes the growth parameter growth, behavioral characteristic frequency and environmental data average of the week to generate a weekly growth cycle assessment report.

6. A smart data acquisition method based on automatic measurement of fish growth data, characterized in that, An intelligent data acquisition system based on automatic measurement of fish growth data is applied, the system comprising a data acquisition module, a data transmission module, a data processing module, and a data management module, the method comprising: Raw multi-source heterogeneous data during the fish growth process are collected and preprocessed to obtain multi-source heterogeneous data, which includes fish image data, fish acoustic data, and aquaculture environment data. The multi-source heterogeneous data is transmitted to the data processing center by combining 5G, Wi-Fi 6, and LoRa wireless transmission technologies. The fish image data is segmented using a Mask R-CNN model to generate a fish segmentation mask. Feature point detection is performed on the fish acoustic data to generate an acoustic feature matrix. The fish segmentation mask and the acoustic feature matrix are fused using a Transformer model, and the fish behavior category and confidence level are determined through a cross-attention mechanism. Based on the fish segmentation mask, the fish growth parameters are determined using an HRNet model. Based on the fish growth parameters and the acoustic feature matrix, the predicted fish weight is determined using an improved LSTM model with an integrated attention mechanism. The fish growth parameters, the acoustic features corresponding to the fish behavior categories, and the aquaculture environment data are then used to predict the growth cycle required for the fish to reach a specific growth stage. The system associates and stores data throughout the entire process, and displays fish growth curves, fish weight change trends, fish behavioral characteristic change trends, and environmental parameter change information in real time on a visual interface. It also generates automated analysis reports based on preset cycles and preset trigger conditions.

7. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method of claim 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method of claim 6.

Images