Soft-shell crab molting state monitoring and automatic catching device

Abstract

The invention discloses a soft-shell crab molting state monitoring and automatic capturing device, and particularly relates to the technical field of soft-shell crab molting state monitoring and automatic capturing, and the soft-shell crab molting state monitoring and automatic capturing device comprises a monitoring module, an environment self-adaption module, a capturing module, a control module and a remote monitoring module. According to the soft-shell crab molting state monitoring and automatic capturing device, through cooperative application of multi-sensor data fusion and a deep learning algorithm, full-process intelligent monitoring and automatic capturing of the soft-shell crab molting state are achieved; by combining technical means such as environment self-adaptive adjustment, wireless low-power-consumption communication, large-scale multi-pond linkage and night weak light optimization, efficient, stable and safe operation of the system in various breeding scenes is guaranteed, and meanwhile, modular software and hardware design and a distributed cloud edge collaborative architecture have a good application prospect. The device has excellent expansibility, maintenance convenience and intelligent operation and maintenance capability, and the automation degree and production benefits of soft-shell crab culture are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of soft-shell crab molting status monitoring and automatic capture, and in particular to a soft-shell crab molting status monitoring and automatic capture device. Background Art

[0002] With the growing market demand for soft-shell crabs, traditional artificial aquaculture methods rely primarily on manual catching and visually determining molting timing. This is not only inefficient, but can also easily damage the crabs due to delayed judgment or improper operation, affecting yield and commodity value. To improve production efficiency, some studies have attempted to install single video surveillance or environmental sensors such as temperature and dissolved oxygen in aquaculture ponds. However, these applications are often isolated, lacking comprehensive analysis of the multi-stage characteristics of the molting process, and are difficult to achieve full-time, all-area coverage.

[0003] In recent years, deep learning technology has begun to be applied in the field of intelligent livestock and aquaculture, realizing animal behavior recognition through target detection and tracking in video images;

[0004] At the same time, multi-parameter sensor networks have also achieved certain results in environmental monitoring. However, the existing systems still have the following major shortcomings: First, the monitoring method is single and the multi-source data fusion capability is weak, making it difficult to accurately distinguish the early, middle and complete stages of molting;

[0005] Second, it has poor environmental adaptability and cannot operate stably in low-light, nighttime, or power-limited scenarios;

[0006] Third, there is a lack of supporting solutions for automated capture and transportation, and a lot of manual work is still required;

[0007] Fourth, it is difficult to support multi-pool linkage and large-scale expansion, and the operation and maintenance costs are high;

[0008] Fifth, there is a lack of a unified cloud-edge collaborative architecture and online model update mechanism, making it difficult to continuously optimize recognition accuracy and system performance.

[0009] The above technical defects restrict the intelligent, automated and large-scale development of soft-shell crab farming. Summary of the Invention

[0010] The main purpose of the present invention is to provide a soft-shell crab molting status monitoring and automatic capture device, which can effectively solve the problems in the background technology.

[0011] To achieve the above object, the technical solution adopted by the present invention is:

[0012] A soft-shell crab molting status monitoring and automatic capture device comprises: a monitoring module, an environment adaptation module, a capture module, a control module and a remote monitoring module;

[0013] The monitoring module includes a high-resolution low-light camera, an infrared sensor, a water pressure sensor, and a water quality sensor arranged at a predetermined position in the aquaculture pond;

[0014] The monitoring module includes a deep learning processing unit for identifying the three stages of early molting, mid-molting and complete molting based on the image data collected by the camera and the data collected by the sensor;

[0015] The environment adaptation module is used to adjust the camera exposure parameters and infrared sensor sensitivity according to the light intensity, water temperature and water quality sensor detection results collected by the camera;

[0016] The capture module includes a robotic arm with multiple degrees of freedom, with a flexible gripper or a silicone suction cup installed at the end of the robotic arm;

[0017] The capture module includes a conveying device for transferring the captured soft-shell crabs to a temporary holding pond or a packaging area;

[0018] The control module includes a programmable logic controller for controlling the motion trajectory and grasping action of the robotic arm according to the position information and posture information output by the deep learning processing unit;

[0019] The remote monitoring module includes a cloud server for receiving data uploaded by the monitoring module and the control module to realize monitoring status display and abnormality warning.

[0020] Preferably, the deep learning processing unit adopts a convolutional neural network model and is trained based on labeled soft-shell crab molting images and video data to automatically identify the three stages of early molting, mid-molting and completed molting.

[0021] Preferably, the infrared sensor is arranged in an area where crabs frequently move, and identifies the crab's position and assists in judging its molting behavior by collecting information on heat source changes.

[0022] Preferably, the environmental adaptation module automatically adjusts the camera exposure parameters, infrared sensor sensitivity and threshold parameters of the deep learning processing unit according to the water temperature, light intensity and water quality sensor detection results to adapt to changes in the breeding environment.

[0023] Preferably, the communication interface uses wired Ethernet and wireless Wi-Fi for real-time data transmission between the monitoring module, the environment adaptation module, the control module and the remote monitoring module.

[0024] Preferably, the cloud server generates push notifications for abnormal molting status information, and stores and analyzes received historical monitoring data and captured data for use in breeding performance evaluation and optimization decisions.

[0025] Preferably, the robotic arm has six degrees of freedom and can quickly capture soft-shell crabs through a planned motion trajectory;

[0026] The clamping claws are designed with flexible materials to prevent mechanical damage to the soft-shell crab.

[0027] Preferably, the conveying device adopts a speed-adjustable conveyor belt, and its transmission speed is set according to the body size and capture frequency of the soft-shell crab to ensure the safety and stability of the soft-shell crab during the transfer process.

[0028] Preferably, the water quality sensor is used to detect dissolved oxygen, pH value and ammonia nitrogen concentration in real time, and transmit the detection results to the environmental adaptation module for adjusting relevant parameters during monitoring and capturing operations.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. The present invention achieves accurate identification of the molting stage of soft-shell crabs by integrating deep learning and multi-sensor data. It uses a variety of equipment such as high-resolution cameras, infrared sensors, and water pressure sensors to comprehensively monitor the molting process of soft-shell crabs. The deep learning processing unit performs data fusion through a temporal convolutional network and an improved ResNet-50 model to identify the different stages of molting, thereby ensuring efficient and accurate capture of the action, greatly improving capture efficiency and accuracy.

[0031] 2. The environmental adaptation module of the present invention automatically adjusts the camera gain, infrared sensor sensitivity, and the threshold of the data processing model by real-time monitoring of water temperature, light, environmental parameters, and water quality changes. This adaptive adjustment mechanism ensures the stable operation of the equipment under various environmental conditions and improves the adaptability of the device. In particular, it can still efficiently monitor the molting process and automatically capture it in cases of insufficient light or frequent changes in water quality.

[0032] 3. In wireless communication scenarios, the present invention adopts ZigBee Mesh network and Wi-Fi 6 technology to solve the limitations of traditional wired wiring and realize wireless remote monitoring and data transmission of the system. In particular, the low-power design supports solar energy or small energy storage power supply systems, enabling the device to operate stably in environments with insufficient power supply, which is particularly suitable for island or temporary breeding scenarios.

[0033] 4. The present invention supports multi-pool linkage management and can uniformly monitor and schedule tasks for multiple breeding ponds through a central control system. The redundant design of the RS-485 bus and Ethernet ensures the stability and real-time performance of data transmission. The system can dynamically adjust the work tasks of the robotic arm and realize shared services for multiple pools, greatly improving the management efficiency of the breeding park and reducing operating costs.

[0034] 5. In weak light or nighttime environments, the present invention ensures accurate soft-shell crab status monitoring under low-light conditions by adding LED fill lights and highly sensitive infrared sensors. Combined with a night-specific model and an emergency warning system, it can monitor and push important abnormal events in real time, ensuring that the on-duty personnel can respond in the first time, thereby improving the safety and reliability of the system and ensuring the safety and efficiency of the breeding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the overall process of the present invention. DETAILED DESCRIPTION

[0036] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0037] like Figure 1 As shown, a soft-shell crab molting status monitoring and automatic capture device includes: a monitoring module, an environment adaptation module, a capture module, a control module and a remote monitoring module;

[0038] The monitoring module includes a high-resolution low-light camera, an infrared sensor, a water pressure sensor, and a water quality sensor arranged at a predetermined position in the aquaculture pond;

[0039] The monitoring module includes a deep learning processing unit for identifying the three stages of early molting, mid-molting and complete molting based on the image data collected by the camera and the data collected by the sensor;

[0040] The environment adaptation module is used to adjust the camera exposure parameters and infrared sensor sensitivity according to the light intensity, water temperature and water quality sensor detection results collected by the camera;

[0041] The capture module includes a robotic arm with multiple degrees of freedom, with a flexible gripper or a silicone suction cup installed at the end of the robotic arm;

[0042] The capture module includes a conveying device for transferring the captured soft-shell crabs to a temporary holding pond or a packaging area;

[0043] The control module includes a programmable logic controller for controlling the motion trajectory and grasping action of the robotic arm according to the position information and posture information output by the deep learning processing unit;

[0044] The remote monitoring module includes a cloud server for receiving data uploaded by the monitoring module and the control module to realize monitoring status display and abnormality warning.

[0045] Furthermore, the deep learning processing unit adopts a convolutional neural network model and is trained based on labeled soft-shell crab molting images and video data to automatically identify the three stages of molting: early molting, mid-molting and molting completion.

[0046] Furthermore, the infrared sensor is arranged in the area where crabs are active frequently, and the position of the crabs is identified by collecting the information of heat source changes, and the molting behavior is assisted in judging.

[0047] Furthermore, the environmental adaptation module automatically adjusts the camera exposure parameters, infrared sensor sensitivity and threshold parameters of the deep learning processing unit according to the water temperature, light intensity and water quality sensor detection results to adapt to changes in the breeding environment.

[0048] Furthermore, the communication interface uses wired Ethernet and wireless Wi-Fi to transmit data in real time between the monitoring module, the environment adaptation module, the control module and the remote monitoring module.

[0049] Furthermore, the cloud server generates push notifications for abnormal molting status information, and stores and analyzes received historical monitoring data and captured data for use in breeding performance evaluation and optimization decision-making.

[0050] Furthermore, the robotic arm has six degrees of freedom and can quickly capture soft-shell crabs through a planned motion trajectory;

[0051] The clamping claws are designed with flexible materials to prevent mechanical damage to the soft-shell crab.

[0052] Furthermore, the conveying device adopts a speed-adjustable conveyor belt, and its transmission speed is set according to the body size and capture frequency of the soft-shell crab to ensure the safety and stability of the soft-shell crab during the transfer process.

[0053] Furthermore, the water quality sensor is used to detect dissolved oxygen, pH value and ammonia nitrogen concentration in real time, and transmit the detection results to the environmental adaptation module for adjusting relevant parameters during monitoring and capturing operations.

[0054] Example 1: This example is to achieve the purpose of deploying a capture device in a soft-shell crab breeding pond under a basic scenario.

[0055] Configuration and function overview:

[0056] 1. Monitoring module

[0057] Camera system: High-resolution low-light cameras are deployed with night vision sensitivity of 0.01 lux, focal length of 8mm, frame rate of 30fps. The four cameras are installed at a height of approximately 1.2m, with an inclination angle of approximately 30°. Adjacent fields of view overlap by 20%, and a panoramic view of the aquaculture pond is formed through real-time stitching.

[0058] Infrared and water pressure sensors: Four rows of infrared pyroelectric sensors are arranged equidistantly above the water surface, with a detection distance of about 3m, a viewing angle of 120°, and an interval of 2m. Water pressure sensors are arranged along the mainstream direction at the bottom of the pool, with an accuracy of ±0.5kPa and a sampling frequency of 10Hz.

[0059] Water quality monitoring: Multi-parameter online sensors, including dissolved oxygen 0-20mg / L, pH electrode 0-14pH, ammonia nitrogen online analyzer 0-10mg / L, with a data update cycle of approximately 5 minutes.

[0060] 2. Environmental Adaptation Module

[0061] Real-time acquisition of water temperature (20–28°C), ambient light (0–10,000 lux), and water quality parameters. Through built-in PID or similar adaptive control algorithms, the camera exposure parameters, infrared sensor sensitivity, and deep learning model thresholds are adjusted to ensure robustness and accuracy of monitoring during environmental fluctuations.

[0062] 3. Capture module

[0063] Robotic arm: It adopts a six-degree-of-freedom robotic arm with a maximum load of about 5kg, a maximum movement speed of 1m / s, and a repeatability accuracy of ±0.1mm; its end is equipped with a flexible gripper made of silicone material, with an opening force of about 2N and a clamping force of about 8N, taking into account both softness and stability during grasping.

[0064] Conveying system: Adjustable speed conveyor belt, with a width of about 200mm and an adjustable speed of 0.1-1.0m / s, equipped with a pneumatic pushing device to ensure that the acceleration during the transfer process is less than 0.5g, reducing the impact on soft-shell crabs.

[0065] 4. Control module

[0066] Based on a programmable logic controller (PLC) or industrial PC, it receives the position and posture information output by the deep learning processing unit, plans the motion trajectory of the robotic arm in real time, issues drive instructions, and links the conveying system to complete the automatic capture and transfer of soft-shell crabs.

[0067] 5. Remote monitoring module

[0068] Upload monitoring data and operating status to the cloud server via wired or wireless networks to achieve a visual monitoring interface, historical record query and abnormal warning push.

[0069] Specific implementation process:

[0070] 1. Equipment installation and network access

[0071] Build brackets around the aquaculture pond to secure the camera and infrared sensor. Lay out water pressure and water quality sensor lines on the pond bottom and water surface, and lead them to the control cabinet.

[0072] Place the deep learning processing unit and GPU module into a protective box, connect them to a power source, and complete the local network configuration, such as Ethernet or Wi-Fi. The remote monitoring module establishes a secure channel with the cloud server.

[0073] 2. Model preparation and verification

[0074] Import labeled soft-shell crab molting images and video data, train a TCN and ResNet-50 fusion model in the development environment, verify recognition accuracy and latency on a test set, and adjust hyperparameters to ensure that the accuracy of the three levels of discrimination ("early molting," "mid-molting," and "completed molting") meets requirements.

[0075] Deploy the trained model to the edge computing unit and start the offline or online calibration process.

[0076] 3. Sensor and camera calibration

[0077] Calibrate camera exposure and focus one by one to obtain clear images, and adjust the sensitivity of infrared sensors on site to ensure reliable detection under different lighting conditions;

[0078] Verify the reading accuracy of water pressure and water quality sensors and compare and calibrate them with standard instruments.

[0079] 4. Linkage debugging

[0080] Set the recognition threshold and grasping strategy parameters on the control interface to simulate the shell-molting scenario. Use models or real samples to verify the coordination between the deep learning unit's detection result output, PLC instruction issuance, and robotic arm action execution.

[0081] Adjust the gripping path and speed of the robotic arm to ensure smooth linkage between the gripper movement and the conveying system.

[0082] 5. System acceptance and trial operation

[0083] Monitor the first molting events in the actual aquaculture environment, record the capture success rate, false positive rate and safety of soft-shell crabs, and fine-tune the thresholds, control parameters and process strategies based on the feedback;

[0084] After completing the system integration, formulate daily maintenance plans and personnel operation manuals, and train farm operators to be familiar with the operation interface and exception handling procedures.

[0085] The second embodiment is to realize the use of the capture device in a wireless network scenario.

[0086] Configuration and function overview:

[0087] 1. Communication Network

[0088] Wireless sensor network: uses a ZigBee Mesh network with a node data rate of approximately 250kbps, a coverage radius of approximately 100m, and can be expanded to approximately 50 nodes. It is used for data transmission from infrared, water pressure, and water quality sensors. Camera data is carried via Wi-Fi 6, ensuring real-time video streaming and a network packet loss rate of less than 1%.

[0089] Edge and cloud collaboration: Edge processors, such as the ARM Cortex-A72×4 core, deploy a lightweight YOLOv5n model to complete preliminary image detection. More refined recognition is performed in the cloud or on more powerful computing nodes, and model updates are periodically issued.

[0090] 2. Online learning and model updating

[0091] Adopting the Federated Learning (FedAvg) framework: each edge node uses the collected data to fine-tune the model locally, uploading only the gradient or updating the parameters. The central server aggregates and generates a new global model, which is distributed to each node, improving the adaptability of the model in multiple pools and different environments.

[0092] 3. Low power management

[0093] Sensors and edge nodes support sleep-wake mode, with idle power consumption less than 50mW. They are compatible with solar energy or small energy storage power supplies, and dynamically schedule communication and computing resources to further reduce overall energy consumption.

[0094] 4. Capture Improvements

[0095] A pneumatic blowing device is added to the original flexible gripper. The blowing pressure is about 0.2-0.5MPa and is adjustable. When the robotic arm approaches the soft-shell crab, air blowing is performed first, which reduces the clamping time by about 30% and reduces the risk of concentrated force on the back of the crab.

[0096] Specific implementation process:

[0097] 1. Network deployment and node configuration

[0098] In the target aquaculture area, ZigBee mesh network node locations are divided according to the terrain. Infrared, water pressure, and water quality sensors are installed and node addresses and routes are configured. Wi-Fi 6 access points are set up and antenna positions are adjusted to ensure that there are no blind spots in camera video coverage.

[0099] Configure security authentication and access permissions between the edge processor and cloud server, and deploy a lightweight YOLOv5n model.

[0100] 2. Federated Learning Process Construction

[0101] Deploy federated learning scripts on each edge node and central server, initialize global model weights, set upload frequency and aggregation strategy, and ensure regular updates even when network bandwidth is limited;

[0102] Perform initial training and testing in a multi-pool environment to check local fine-tuning effects and global model performance.

[0103] 3. Hardware and clamping debugging

[0104] Calibrate the synchronization timing between the pressure of the pneumatic blowing device and the action of the gripper: test the gripping effect before and after blowing on the actual sample, and adjust the optimal airflow pressure range and the timing parameters of the gripper opening / closing;

[0105] Verify the linkage between blowing and grabbing to ensure that the capture action is efficient and does not harm the soft-shell crab.

[0106] 4. Low power consumption and power supply test

[0107] Connect sensors and edge nodes to solar or mobile power systems, test sleep-wake cycles and network recovery speeds, and ensure continuous system operation at night or under cloudy conditions. Monitor power consumption and power status, and adjust wake-up strategies if necessary.

[0108] 5. System acceptance and operation monitoring

[0109] Run for more than 48 hours to test network stability, recognition accuracy, and federated learning update effects; optimize network topology and model hyperparameters based on monitoring results;

[0110] Eventually, an operating guide is formed to train on-site technicians to perform maintenance and troubleshooting.

[0111] Embodiment 3: This embodiment is to achieve the purpose of using the capture device in a multi-pool linkage scenario.

[0112] Configuration and function overview:

[0113] 1. Network and bus architecture

[0114] It adopts industrial-grade RS-485 bus with a baud rate of 115200bps, multi-master and multi-slave mode and Ethernet PLC dual-redundant link design to ensure that the switching time is less than 100ms when any path fails, achieving high availability and real-time data transmission.

[0115] 2. Centralized scheduling and load balancing

[0116] The central control center is connected to multiple robotic arm clusters, such as 8 arms or more, and adopts a dynamic task scheduling algorithm, combined with the shortest job priority strategy and real-time load monitoring, to evenly distribute capture tasks among the arms, thereby improving the overall throughput by about 45%.

[0117] 3. Parallel transmission and multi-channel output

[0118] It adopts a grid-type chain belt structure with a bandwidth of about 300mm, supports simultaneous operation of 4 parallel channels, and the throughput of each channel can reach about 120 pieces per hour, meeting the needs of large-scale capture.

[0119] 4. Monitoring client and early warning

[0120] The monitoring client supports filtering by aquaculture pond number, displaying multiple video channels in real time, viewing historical track playback and performing fault diagnosis, and sending warnings via dual channels of voice calls and SMS, with end-to-end latency ≤2s.

[0121] Specific implementation process:

[0122] 1. Network cabling and redundancy construction

[0123] Lay out RS-485 buses and Ethernet redundant links between each aquaculture pond, install PLCs and switches, configure active and standby nodes, and verify the link switching mechanism;

[0124] Deploy monitoring display screens and operation terminals in the central control room, and build a server environment for dynamic task scheduling.

[0125] 2. Verification of robotic arm and conveying system

[0126] Verify the range of motion, repeatability, and grasping timing of each robotic arm one by one, and calibrate the speed, synchronization logic, and safety collision detection of each channel on the grid-type chain belt;

[0127] Each channel is linked to the robotic arm, and the reliability and throughput efficiency of multi-channel parallel operations are tested by simulating shell molting events.

[0128] 3. Monitoring client access and view configuration

[0129] Import the mapping between pool numbers and devices into the monitoring client, design the partition view layout, and test the real-time video streaming, warning notification, and historical playback functions to ensure rapid location and response when an anomaly occurs.

[0130] Set up multi-level permission management to limit the access scope and operating permissions of operators, maintenance personnel and visitors.

[0131] 4. Multi-pool task switching test

[0132] Use simulated or real molting samples to conduct cross-pool task switching tests, verify the central dispatcher's control delay and responsiveness of the robotic arm, and test the reliability of voice / SMS warnings under different network conditions.

[0133] Adjust scheduling algorithm parameters, network priorities, and warning thresholds based on test results.

[0134] 5. Training and acceptance

[0135] Prepare system operation manuals and maintenance procedures, organize technical team training; conduct trial runs in actual farms to observe system stability and troubleshooting efficiency, further optimize and complete acceptance.

[0136] Embodiment 4: This embodiment is to achieve the purpose of using the capture device in low-light and nighttime scenes.

[0137] Configuration and function overview:

[0138] 1. Active fill light and camera optimization

[0139] Install an adjustable color temperature LED fill light with a color temperature range of approximately 3000–6500K and adjustable brightness from 0-100%. It is synchronized with the camera trigger to ensure clear and usable images at night or in low-light environments.

[0140] 2. Dual-mode recognition algorithm

[0141] The deep learning processing unit adds an IR-STC (Infrared Stacked Convolutional Network) branch for infrared and visible light fusion, which fuses different spectral data in a 1:1 weighted ratio, achieving a nighttime recognition accuracy of approximately 96%;

[0142] The number of infrared sensor rows has been increased to four, and the detection distance has been increased to approximately 5m. The sensitivity can be adjusted in multiple levels, and the maximum noise equivalent power (NEP) is less than 50μW.

[0143] 3. Accurate positioning and capture

[0144] The robotic arm's gripper is embedded with a circular micro LED light strip to form a point light source mark, which cooperates with the camera for 3D ranging and positioning. The positioning error is less than 1mm, ensuring high-precision grasping in low-light conditions.

[0145] 4. Emergency warning system

[0146] The system integrates a three-level early warning mechanism consisting of App push, SMS and voice calls. If any abnormal activity or equipment failure is detected at night, the on-duty personnel will be notified immediately to ensure timely response.

[0147] Specific implementation process:

[0148] 1. Linkage layout of fill light and camera

[0149] An LED light group with adjustable color temperature is installed around the breeding pond to complete the connection with the camera hardware trigger signal, and the fill light brightness and color temperature are adjusted to ensure clear images without overexposure in different night scenes.

[0150] 2. Algorithm deployment and model switching

[0151] A night-specific IR-STC model branch is loaded into the deep learning processing unit, which can be set to automatically switch or fuse processing according to lighting conditions. A large amount of data collected at night is offline labeled and the model is fine-tuned to optimize the recognition effect.

[0152] 3. Infrared sensing and positioning calibration

[0153] Adjust the sensitivity of the infrared sensor, test the heat source detection distance, compare it with the camera image, optimize the fusion strategy, calibrate the point light source position on the robot arm gripper light strip, and cooperate with the camera system to perform 3D ranging calibration.

[0154] 4. Early warning platform integration and testing

[0155] Build a push platform, integrate apps, SMS, and voice interfaces, simulate abnormal nighttime scenarios such as sudden events or equipment failures, and verify the timeliness and reliability of the three-level push notification system; record latency and optimize network and interface parameters.

[0156] 5. Night trial operation and optimization

[0157] Operate continuously for at least 7 days in actual nighttime or low-light environments, summarize recognition accuracy, warning hit rate, and system reliability data, and continuously fine-tune the algorithm threshold and fill-light strategy based on the operating data; after optimization is completed, form a standard operation and maintenance process.

[0158] Embodiment 5: This embodiment is to achieve the purpose of park integration and intelligent management.

[0159] Configuration and function overview:

[0160] 1. Distributed cloud-edge collaborative cluster

[0161] This embodiment builds a private cloud Kubernetes cluster in the campus computer room, deploys TensorFlowServing and ONNXRuntime services, elastically expands inference nodes on demand, supports model inference requests of each single device, and efficiently collaborates with edge nodes and cloud services through secure channels.

[0162] 2. BI and data reports

[0163] The platform has a built-in BI module that automatically collects and analyzes indicators such as the molting cycle (statistics by day / week / month), captures success rate curves, and environmental parameter correlations, and generates visual dashboards to assist in operational decision-making and optimization.

[0164] 3. Modular hardware design

[0165] The robotic arm and conveying interface adopt a unified JAE standard modular interface to achieve hot-swap replacement, with on-site replacement time less than 5 minutes, improving maintenance efficiency and equipment reliability.

[0166] 4. Multi-terminal and security management

[0167] The remote monitoring client supports multi-resolution adaptive display on PCs, tablets, mobile phones, etc.; permission management is divided into three levels: "administrator", "operator" and "visitor". All operation logs are stored in encrypted form to meet security regulations such as GDPR and ISO27001.

[0168] Specific implementation process:

[0169] 1. Cluster deployment and service configuration

[0170] Build a Kubernetes private cloud cluster in the campus computer room, configure the TensorFlowServing and ONNXRuntime deployment environment, establish service discovery and heartbeat detection mechanisms, and ensure high-availability connections between each inference node and edge devices;

[0171] Add the edge computing nodes of individual on-site devices to the cluster network to achieve unified management and scheduling.

[0172] 2. BI module and data warehouse connection

[0173] Deploy a BI platform and connect it to the data warehouse to analyze data such as molting status, capture records, and environmental parameters collected by each device. Configure periodic automatic reports and custom dashboards to support multi-dimensional data visualization.

[0174] Define monitoring indicators and alarm thresholds based on operational needs to provide an intuitive basis for decision-making.

[0175] 3. Hardware module replacement and hot-swap verification

[0176] The interface modules of the robotic arm and conveying device were gradually replaced on site to verify the compatibility of the standardized interface and the speed of the hot-swap process (less than 5 minutes). The system recovery time during the replacement process was recorded to ensure that the interference of equipment maintenance on production was minimized.

[0177] 4. Multi-terminal client deployment and testing

[0178] Install or connect to the adaptive monitoring client on different devices such as PCs, tablets, and mobile phones to debug the multi-resolution interface; test remote control, real-time video, historical data access, and warning reception functions to ensure consistency and stability of the operating experience.

[0179] 5. Permissions and security policy implementation

[0180] Establish a hierarchical authority management system and configure functional permissions corresponding to roles; deploy a log audit system to encrypt and record all operations and conduct regular reviews; conduct penetration testing and security assessments to verify that encryption storage and access control mechanisms comply with GDPR and ISO27001 requirements;

[0181] After completing the security policy test, the overall platform acceptance is carried out to ensure that the cluster, network and each module meet the requirements of high availability, high security and maintainability.

[0182] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A device for monitoring and automatically capturing the molting status of soft-shell crabs, characterized in that: include: Monitoring module, environmental adaptation module, capture module, control module and remote monitoring module; The monitoring module includes a high-resolution low-light camera, an infrared sensor, a water pressure sensor, and a water quality sensor arranged at a predetermined position in the aquaculture pond; The monitoring module includes a deep learning processing unit for identifying the three stages of early molting, mid-molting and complete molting based on the image data collected by the camera and the data collected by the sensor; The environment adaptation module is used to adjust the camera exposure parameters and infrared sensor sensitivity according to the light intensity, water temperature and water quality sensor detection results collected by the camera; The capture module includes a robotic arm with multiple degrees of freedom, with a flexible gripper or a silicone suction cup installed at the end of the robotic arm; The capture module includes a conveying device for transferring the captured soft-shell crabs to a temporary holding pond or a packaging area; The control module includes a programmable logic controller for controlling the motion trajectory and grasping action of the robotic arm according to the position information and posture information output by the deep learning processing unit; The remote monitoring module includes a cloud server for receiving data uploaded by the monitoring module and the control module to realize monitoring status display and abnormality warning.

2. The device for monitoring and automatically capturing the molting status of soft-shell crabs according to claim 1, characterized in that: The deep learning processing unit adopts a convolutional neural network model and is trained based on labeled soft-shell crab molting images and video data to automatically identify the three stages of molting: early molting, mid-molting, and molting completion.

3. The device for monitoring and automatically capturing the molting status of soft-shell crabs according to claim 1, wherein: The infrared sensor is arranged in an area where crabs frequently move, and identifies the crab's position and assists in judging its molting behavior by collecting information on heat source changes.

4. The device for monitoring and automatically capturing the molting status of soft-shell crabs according to claim 1, wherein: The environmental adaptation module automatically adjusts the camera exposure parameters, infrared sensor sensitivity and threshold parameters of the deep learning processing unit according to the water temperature, light intensity and water quality sensor detection results to adapt to changes in the breeding environment.

5. The device for monitoring and automatically capturing the molting status of soft-shell crabs according to claim 1, characterized in that: The communication interface uses wired Ethernet and wireless Wi-Fi for real-time data transmission between the monitoring module, the environment adaptation module, the control module and the remote monitoring module.

6. The soft-shell crab molting status monitoring and automatic capture device according to claim 1, characterized in that: The cloud server generates push notifications for abnormal molting status information, and stores and analyzes received historical monitoring data and captured data for use in aquaculture performance evaluation and optimization decision-making.

7. The soft-shell crab molting status monitoring and automatic capture device according to claim 1, characterized in that: The robotic arm has six degrees of freedom and can quickly capture soft-shell crabs through a planned motion trajectory; The clamping claws are designed with flexible materials to prevent mechanical damage to the soft-shell crab.

8. The soft-shell crab molting status monitoring and automatic capture device according to claim 1, characterized in that: The conveying device adopts a speed-adjustable conveyor belt, and its transmission speed is set according to the body size and capture frequency of the soft-shell crab to ensure the safety and stability of the soft-shell crab during the transfer process.

9. The soft-shell crab molting status monitoring and automatic capture device according to claim 1, characterized in that: The water quality sensor is used to detect dissolved oxygen, pH value and ammonia nitrogen concentration in real time, and transmit the detection results to the environmental adaptive module for adjusting relevant parameters during monitoring and capture operations.

Images