A non-contact method and system for monitoring the quality and behavior characteristics of live shrimp

Abstract

This invention discloses a non-contact method for monitoring the quality and behavioral characteristics of live shrimp, comprising the following steps: acquiring video frames and calculating the conversion ratio between pixels and physical dimensions; using a pose estimation model to infer for each frame of image, and simultaneously outputting the shrimp bounding box and coordinates of at least three key points; assigning and maintaining a unique tracking ID to the detected shrimp; calculating phenotypic parameters including body length, curl, and color features based on the bounding box and key points of the tracking ID; and performing quality rating based on the phenotypic parameters. An embodiment of this invention also discloses a non-contact system for monitoring the quality and behavioral characteristics of live shrimp. Using this invention, continuous and automated monitoring of shrimp populations in aquaculture ponds can be achieved without contact or interference with the normal lives of shrimp, utilizing a single-view monitoring camera.

Description

Technical Field

[0001] This invention relates to the technical field of computer vision for aquaculture monitoring, and in particular to a non-contact intelligent monitoring method and system for the quality and behavioral characteristics of live shrimp. Background Technology

[0002] In the farming of Litopenaeus vannamei (whiteleg shrimp), with the promotion of intensive farming models and the continuous increase in shrimp stocking density, disease risks and stress problems have become increasingly prominent. The health status of shrimp directly affects their growth rate, survival rate, feed conversion rate, and final commercial value. Therefore, timely and accurate monitoring and assessment of the growth status and health level of the farmed population is crucial for achieving refined management and risk early warning.

[0003] In actual production, the health status of shrimp is manifested through a series of quantifiable morphological characteristics and behavioral patterns. Growth rate (body length change), posture (such as abnormal curling), and body color (such as reddish or whitish discoloration) are intuitive indicators for judging their physiological state. In terms of behavior, movement speed, activity level, and response to external stimuli are important indicators for measuring their vitality and stress level. Traditional monitoring methods rely on manual observation and sampling measurements by aquaculture workers, which suffers from problems such as strong subjectivity, low efficiency, discontinuous data, and limited sample size, making it difficult to support the needs of large-scale, high-density modern aquaculture management.

[0004] Currently, existing solutions related to the technical concept of this invention mainly include the following paths: (1) Machine vision phenotyping measurement path based on static images Implementation scheme: As described in Chinese Patent Publication No. CN118077627A, "A Machine Vision-Based Device for Measuring the Body Length of South American Shrimp and a Feeding Device", static or quasi-static images of shrimp are collected in a specific device, the shrimp outline is extracted using image processing algorithms, and pixel-to-physical size conversion is performed based on a calibration object, thereby automatically measuring parameters such as body length.

[0005] (2) Fine morphological quantization path based on complex multi-keypoint model Implementation plan: Drawing on the technical approach of facial key point detection (such as 68 key points), we attempt to annotate a large number of key points on shrimp or other marine organisms in order to obtain richer morphological parameters through detailed contour and pose modeling.

[0006] Source of existing technology: Wu Xiaoping et al., Multi-pose face recognition based on facial key points and incremental clustering [J]. Laser & Optoelectronics Progress, 2019, 56(14): 141002. (3) Biobehavioral monitoring pathway based on contact sensors Implementation: In fish behavior research, electronic tags or sensors are often implanted or attached to record the individual's movement and physiological signals.

[0007] Source of existing technology: Horiguchi Y, et al. Development of a remote monitoring system for stress response in fish[J]. Fish Physiology and Biochemistry, 2025. and Wang L, et al. Recent advances in wearable flexible sensors[J]. Advanced Materials Technologies, 2024. Main disadvantages: Not applicable to shrimp: Shrimp are crustaceans that need to molt periodically. Any sensors that are attached or implanted will fall off with the molt, making long-term monitoring impossible.

[0008] Causes severe stress and damage: Shrimp have fragile bodies, and physical attachment can seriously affect their normal behavior and even lead to death.

[0009] High cost and difficult deployment: The high cost of the sensors themselves and their deployment and maintenance makes them unsuitable for large-scale farming operations.

[0010] (4) Deep learning-based hierarchical method for general object detection Implementation scheme: As shown in the patent document CN111666986A, target detection models such as YOLOv4 are used. By constructing a dataset and training a neural network, the size grading or variety identification of crayfish can be achieved.

[0011] The main drawbacks of existing technologies are: 1. Static measurement method, detached from the original breeding site Measurements typically require removing shrimp from the aquaculture water and placing them in specific measuring containers, causing stress to the animals and making in-situ, continuous monitoring impossible. Furthermore, only static phenotypic information from single-frame images can be obtained, completely failing to quantify key dynamic health indicators such as shrimp movement speed, activity level, and behavioral patterns.

[0012] 2. Inability to track individuals: In a group environment, it is impossible to distinguish between different individuals. Each measurement is a "one-time" snapshot, and it is impossible to obtain the process of individual behavioral changes.

[0013] 3. The annotation cost is too high, and it lacks practicality and ease of deployment. Existing technologies mostly require labeling dozens of key points for each sample, which consumes enormous human and time resources. This results in a large number of model parameters, slow inference speed, and difficulty in achieving real-time, high-throughput video processing in aquaculture sites. In addition, for Litopenaeus vannamei, which is small in size, has varied postures, and suffers from severe mutual occlusion in water, a large number of key points are prone to false positives or false negatives, resulting in low robustness and practicality of the model in real and complex environments.

[0014] 4. Failed to achieve individual-level behavioral monitoring; limited functionality. Existing technologies mainly focus on the classification or simple counting of static images, failing to achieve continuous tracking of each individual in a group, and failing to integrate phenotyping, behavioral recognition and health assessment into the same system. Summary of the Invention

[0015] The technical problem to be solved by the embodiments of the present invention is to provide a non-contact method and system for monitoring the quality and behavioral characteristics of live shrimp, which uses a single-view monitoring camera to achieve continuous and automated monitoring of shrimp groups in the aquaculture pond without contact or interference with the normal life of shrimp.

[0016] To address the aforementioned technical problems, embodiments of the present invention provide a non-contact method for monitoring the quality and behavioral characteristics of live shrimp, comprising the following steps: S1: Acquire video frames and calculate the conversion ratio between pixels and physical size. S2: Use the pose estimation model to infer for each frame of image, and output the shrimp bounding box and the coordinates of at least three key points; S3: Assign and maintain a unique tracking ID for the detected shrimp; S4: Calculate phenotypic parameters including body length, curvature, and color features based on the trajectory record of the tracking ID; S5: Rating the quality based on the phenotypic parameters.

[0017] Furthermore, S2 also includes the following steps: S21: The step of acquiring shrimp images and annotating each shrimp in each image, wherein the annotation operation includes the coordinates of key points such as the rectangular bounding box, the corresponding rostrum point, the midpoint of the end of the carapace, and the end point of the 7th telson. S22: Use the YOLO pose estimation model to perform inference on each frame of the image.

[0018] Furthermore, step S3 also includes: S31: Multi-target tracking using the ByteTrack algorithm; S32: When a new shrimp individual appears in a video frame, a brand new and unique tracking ID is assigned to it through the ByteTrack algorithm. In subsequent consecutive frames, the newly detected bounding box is matched with the existing tracking trajectory based on the motion model and appearance similarity, ensuring that the ID of the same shrimp remains unchanged until it leaves the screen or is occluded within a set time. S33: Maintain a fixed-length queue for each display ID to store the sequence of bounding box center point coordinates for that individual in the most recent N frames.

[0019] Furthermore, S4 also includes the step of: Body length is calculated using the following method: The point of the forehead is recorded as The midpoint of the cephalothorax is The end point of the 7th telson is ; Calculate pixel distance: , ; Pixel body length of shrimp ; Calculate the actual body length using the scaling factor k: ; The degree of curl is calculated using the following method: Construct two vectors: with the midpoint C at the end of the cephalothorax as the common starting point, pointing to the frontal horn point T and the end point E of the caudal segment, respectively;

[0020]

[0021] Calculate the angle θ between the two vectors:

[0022] Calculate the curl C: ; Color statistics are performed using the following methods: Within the bounding box of each shrimp, a rectangular region with a width and height equal to half that of the original bounding box is defined, centered on the center point of the bounding box, as the region of interest for color sampling; Extract the values ​​of all pixels within the region of interest in the RGB color space; The average value of the blue channel pixels in the background of the aquaculture pond was used as the baseline. ; For shrimp detected subsequently, the relative deviation between the current shrimp body's blue channel mean and the background baseline value is calculated. .

[0023] Furthermore, S5 also includes the steps of: calculating motion speed and classifying behavior; The steps for calculating the motion speed include: Leveraging the ID correlation provided by multi-object tracking, we can obtain the same ID in the current frame t and the previous frame. coordinates of the center point of the bounding box and ; Calculate pixel displacement ; Combining the scaling factor k and the video frame interval time Calculate the instantaneous velocity v: ; The steps for classifying behavior include: Based on preset speed thresholds, the instantaneous motion state of shrimp is divided into four categories: Rest: ; crawl: ; Swimming: ; catapult: .

[0024] Furthermore, S5 also includes a behavioral statistics step: performing statistical analysis on the cumulative duration and proportion of each behavioral state to quantify the overall vitality level of the group.

[0025] Furthermore, the quality rating is determined using the following thresholds: excellent: and ; good: and ; generally: or ; Poor: and ; in, The threshold for curl degree grading. The threshold for color deviation grading.

[0026] Accordingly, embodiments of the present invention also provide a non-contact shrimp live quality and behavioral characteristic monitoring system, comprising: An image acquisition device is fixedly installed above the aquaculture pond to acquire video frames by taking a single-view overhead photograph of the monitoring area. A computer processing unit is configured to execute steps S2 to S5; The data storage and visualization module is used to save raw images, processing results, and analysis logs, and provides a graphical user interface for real-time display of monitoring screens.

[0027] Implementing the embodiments of the present invention has the following beneficial effects: 1. This invention provides a non-invasive, high-throughput in-situ monitoring system: without contacting or interfering with the normal life of shrimp, it uses a single-view monitoring camera to achieve continuous and automated monitoring of shrimp populations in aquaculture ponds.

[0028] 2. This invention achieves a lightweight unification of target detection and core phenotypic extraction: It adopts a single YOLO pose estimation model and efficiently and stably completes individual positioning, body length measurement and pose (curl) assessment of shrimp using only three core anatomical key points. While ensuring practicality, it greatly reduces model training costs and computational complexity.

[0029] 3. This invention enables continuous tracking and behavior analysis at the individual level: By combining a multi-target tracking algorithm, a unique ID is assigned to each shrimp, enabling accurate calculation of its movement trajectory and instantaneous speed, and automatically classifying behavior patterns to provide refined data for vitality assessment.

[0030] 4. This invention constructs a multi-dimensional integrated health assessment system: integrating morphology (curl), appearance (color), and dynamic behavior (speed) indicators to establish a set of quantitative quality rating standards, providing an intuitive and scientific basis for decision-making in aquaculture management. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the method of the present invention; Figure 2 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings.

[0033] The first aspect of this invention provides a non-contact method for monitoring the quality and behavioral characteristics of live shrimp, such as... Figure 1 As shown, the following steps are performed.

[0034] S1: Acquire video frames and calculate the conversion ratio between pixels and physical size.

[0035] An image acquisition device is fixedly installed above the aquaculture pond to capture a single-view overhead image of the monitoring area. A calibrated object of known physical size (such as a ruler) must be placed within the field of view.

[0036] In the initial or periodic calibration process, two marker points of known length on the calibration object are manually calibrated, and their pixel distance is calculated. Let the actual physical distance of the calibration object be... (Unit: cm), then the conversion factor k between pixels and centimeters can be calculated:

[0037] In all subsequent calculations involving physical dimensions, this scaling factor k will be used for conversion. If not calibrated, the system will output relative values ​​in pixels, which will affect the accuracy of physical measurements such as body length.

[0038] S2: Use the pose estimation model to infer for each frame of image, and output the shrimp bounding box and the coordinates of at least three key points.

[0039] Model training is performed using the following steps: Dataset preparation: Collect a large number of shrimp images. Use annotation tools to annotate each shrimp in each image. The annotations include: a bounding box; and the coordinates of three keypoints, corresponding to: the temple, the midpoint of the end of the carapace, and the end of the 7th telson segment.

[0040] Model Selection and Training: YOLOv8-pose was chosen as the base model. This model combines the YOLOv8 object detection model with a keypoint detection head. The model was fine-tuned using the labeled dataset described above, resulting in a customized model weight file specifically designed for Litopenaeus vannamei detection and keypoint recognition.

[0041] Use the trained model to perform the following steps: When processing video, the trained YOLO pose estimation model is invoked for inference on each frame of the video. For each input frame, the model outputs the following information for all detected shrimp individuals at once: bbox: bounding box coordinates (class, x_center, y_center, width, height); keypoints: Normalized coordinates (x, y, visibility) of three keypoints.

[0042] S3: Assign and maintain a unique tracking ID for the detected shrimp.

[0043] To achieve continuous analysis at the individual level, this invention employs the ByteTrack algorithm for multi-target tracking. The ByteTrack algorithm primarily correlates inter-frame data based on the position and size of the detection boxes in consecutive frames, resulting in high computational efficiency and good performance.

[0044] ID Assignment: When a new shrimp appears in the video, the ByteTrack algorithm assigns it a brand new, unique tracking ID. In subsequent frames, the algorithm matches the newly detected bounding box with the existing tracking trajectory based on motion model and appearance similarity, thus ensuring that the ID of the same shrimp remains unchanged until it leaves the frame or is severely occluded.

[0045] Track recording: The system maintains a fixed-length deque for each display ID to store the sequence of bounding box center point coordinates of that individual in the most recent N frames (e.g., N=30) for subsequent drawing of motion trajectories.

[0046] S4: Calculate phenotypic parameters, including body length, curvature, and color features, based on the trajectory record of the tracking ID.

[0047] Based on the keypoint coordinates output by the model, the system automatically calculates the following phenotypic parameters for each shrimp (identified by ID): Body length (L) calculation: Body length is the most basic indicator for measuring the growth status of shrimp. This invention employs a robust body length estimation method: the main axis of the shrimp's body is decomposed into two segments: the cephalothorax and the abdomen.

[0048] The point of the forehead is recorded as The midpoint of the cephalothorax is The end point of the 7th telson is .

[0049] Calculate pixel distance: , .

[0050] Pixel body length of shrimp .

[0051] Calculate the actual body length using the scaling factor k: (Unit: centimeters)

[0052] Curl (C) calculation: Curl is a key behavioral indicator for assessing stress or health status in shrimp. When shrimp are frightened, stressed, or in poor health, they often exhibit a curled-up posture.

[0053] Construct two vectors: with the midpoint C at the end of the cephalothorax as the common starting point, pointing to the frontal horn point T and the end point E of the caudal segment, respectively.

[0054]

[0055]

[0056] Calculate the angle θ (in degrees) between two vectors:

[0057] Calculate the curl C:

[0058] When the shrimp's body is fully extended, points T, C, and E are approximately collinear, and the degree of curl C approaches 0°; when the body is severely curled, θ decreases, and the degree of curl C increases significantly.

[0059] Color feature extraction: Changes in shrimp body color (such as turning red or white) are important visual indicators of abnormal physiological states (such as stress, hypoxia, and disease).

[0060] Identification Region: Within the bounding box of each shrimp, a rectangular region with a width and height equal to half the original bounding box is defined, centered on the center point of the box. This region is used as the Region of Interest (ROI) for color sampling. This region is chosen to avoid background interference and focus on the main color of the shrimp's abdomen.

[0061] Color statistics: Extract the values ​​of all pixels within the ROI in the RGB color space. Considering that shrimp in a healthy state are usually bluish-gray or translucent, while in an abnormal state they often appear reddish or whitish, this invention pays special attention to the mean value of the blue (B) channel. .

[0062] Benchmark Calibration: In the initial stage of system deployment, the average value of the blue channel pixels in the background of the aquaculture pond is calibrated as a benchmark value. .

[0063] Calculate color deviation: For subsequently detected shrimp, calculate the relative deviation between the current shrimp body's blue channel mean and the background baseline value. .

[0064] S5: Rating the quality based on the phenotypic parameters.

[0065] First, we calculate the movement speed and classify the behavior.

[0066] Motion speed calculation and behavior classification: Velocity calculation: Utilizing the ID correlation provided by multi-target tracking, obtain the same ID in the current frame t and the previous frame. coordinates of the center point of the bounding box and Calculate pixel displacement .

[0067] Combining the scaling factor k and the video frame interval time (For example, for 30fps video, (seconds), calculate the instantaneous velocity v:

[0068] Behavioral Classification: Based on preset speed thresholds, the instantaneous movement states of shrimp are classified into four categories. Thresholds and These are empirical parameters that can be dynamically adjusted within the system. The reference values ​​provided by this technology are: v1=10, v2=50, in cm / s.

[0069] Rest: (Almost stationary or moving extremely slowly) crawl: (Using walking legs to move slowly on the bottom of the pool) Swimming: (Swimming at a normal speed using the abdominal foot) catapult: (A sudden backward movement or jump caused by rapid abdominal contraction when startled) Behavioral statistics: The system can provide speed data and perform statistical analysis on demand on the cumulative duration and proportion of each behavioral state to quantify the overall vitality level of the group.

[0070] Next, a quality rating will be conducted.

[0071] This invention integrates curl degree C and color deviation Two core indicators were used to establish a four-level quality rating system. The rating logic and reference thresholds are as follows:

[0072] in, The threshold for curl degree grading. These are the color deviation grading thresholds. These thresholds can be calibrated through preliminary experiments. Based on preliminary experiments, this technique can provide reference values: C1=9.8°, C2=26.9°, C3=44.0°.

[0073] This embodiment also includes data storage and visualization steps: all analysis results are stored in a structured manner. In each frame, a data record is generated for each detected shrimp, including but not limited to the following fields: Frame number, timestamp, display ID, bounding box coordinates, keypoint coordinates, volume length (L), curl (C), blue channel quantity (B), instantaneous velocity (v), behavior category, quality level.

[0074] This data is saved in CSV file format for easy querying, statistics and analysis later.

[0075] A second aspect of this invention provides a non-contact system for monitoring the quality and behavioral characteristics of live shrimp, such as... Figure 2 As shown, it includes: The image acquisition device 101 is fixedly installed above the aquaculture pond and is used to acquire video frames by taking a single-view overhead shot of the monitoring area.

[0076] The image acquisition device can be a regular RGB camera, an industrial camera, etc., fixedly installed above the aquaculture pond, providing a single-view overhead view of the monitoring area. A calibration object of known physical size (such as a ruler) must be placed within the field of view.

[0077] The computer processing unit 102, which may be a CPU, GPU or NPU, is responsible for running the core algorithm program and is configured to execute steps S2 to S5 in the method of this embodiment. The data storage and visualization module 103 is used to save the original images, processing results, and analysis logs, and provides a graphical user interface for real-time display of monitoring screens.

[0078] The system stores all analysis results in a structured format. In each frame, a data record is generated for each detected shrimp, containing, but not limited to, the following fields: Frame number, timestamp, display ID, bounding box center coordinates, volume length (L), curl (C), blue channel quantity (B), instantaneous velocity (v), behavior category, quality level.

[0079] This data is saved in CSV file format for easy querying, statistics and analysis later.

[0080] The key points of the embodiments of the present invention are as follows: 1. Integrated "Detection-Location-Tracking" Architecture: This architecture deeply couples the YOLO-based target detection and keypoint joint model (especially YOLOv8-pose) with the ByteTrack multi-target tracking algorithm. This forms the foundation for achieving fine-grained management at the individual level.

[0081] 2. Phenotypic parameter calculation method based on three core key points: For the Litopenaeus vannamei, this method focuses on three key points: the rostrum (T), the midpoint of the cephalothorax (C), and the telson (E). Based on these, a segmented body length calculation formula (d_TC + d_CE) and a curling degree calculation formula based on the vector angle (supplementary angle) are derived. This constitutes the innovative point of core morphological quantification.

[0082] 3. Color quantization method based on a specific ROI: A central region (e.g., half the area) is defined within the detection bounding box to extract the mean value of the blue channel, which is then compared with the background baseline value. This method effectively eliminates background interference and utilizes the sensitivity of the blue channel to abnormal color changes in shrimp bodies, achieving low-cost and stable body color quantization.

[0083] 4. Multi-dimensional data fusion quality rating system: By fusing static phenotypic (curl, body color) and dynamic behavioral (speed) data, a multi-dimensional and quantifiable shrimp health assessment and early warning scheme is constructed.

[0084] 5. Lightweight, non-contact system solution: The entire solution only requires a monocular camera and conventional computing equipment, without the need for complex 3D sensors or contact devices. It has low deployment costs, simple maintenance, and is suitable for aquaculture site environments.

[0085] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A non-contact method for monitoring the quality and behavioral characteristics of live shrimp, characterized in that, Includes the following steps: S1: Acquire video frames and calculate the conversion ratio between pixels and physical size; S2: Use the pose estimation model to infer for each frame of image, and output the shrimp bounding box and the coordinates of at least three key points; S3: Assign and maintain a unique tracking ID for the detected shrimp; S4: Calculate phenotypic parameters, including body length, curvature, and color features, based on the bounding box and key points of the tracking ID; S5: Rating the quality based on the phenotypic parameters.

2. The non-contact method for monitoring the quality and behavioral characteristics of live shrimp according to claim 1, characterized in that, S2 further includes the following steps: S21: Acquire shrimp images and label each shrimp in each image. The labeling operation includes the coordinates of key points such as the rectangular bounding box, the corresponding rostrum, the midpoint of the end of the cephalothorax, and the end of the 7th telson. S22: Use the YOLO pose estimation model to perform inference on each frame of the image.

3. The non-contact method for monitoring the quality and behavioral characteristics of live shrimp according to claim 1, characterized in that, S3 further includes the following steps: S31: Multi-target tracking using the ByteTrack algorithm; S32: When a new shrimp individual appears in a video frame, a brand new and unique tracking ID is assigned to it through the ByteTrack algorithm. In subsequent consecutive frames, the newly detected bounding box is matched with the existing tracking trajectory based on the motion model and appearance similarity, ensuring that the ID of the same shrimp remains unchanged until it leaves the screen or is occluded within a set time. S33: Maintain a fixed-length queue for each display ID to store the sequence of bounding box center point coordinates for that individual in the most recent N frames.

4. The non-contact method for monitoring the quality and behavioral characteristics of live shrimp according to claim 3, characterized in that, S4 further includes the following steps: Body length is calculated using the following method: The point of the forehead is recorded as The midpoint of the cephalothorax is The end point of the 7th telson is ; Calculate pixel distance: ， ； Pixel body length of shrimp ; Calculate the actual body length using the scaling factor k: ; The degree of curl is calculated using the following method: Construct two vectors: with the midpoint C at the end of the cephalothorax as the common starting point, pointing to the frontal horn point T and the end point E of the caudal segment, respectively; Calculate the angle θ between the two vectors: Calculate the curl C: ； Color statistics are performed using the following methods: Within the bounding box of each shrimp, a rectangular region with a width and height equal to half that of the original bounding box is defined, centered on the center point of the bounding box, as the region of interest for color sampling; Extract the values ​​of all pixels within the region of interest in the RGB color space; The average value of the blue channel pixels in the background of the aquaculture pond was used as the baseline. ; For shrimp detected subsequently, the relative deviation between the current shrimp body's blue channel mean and the background baseline value is calculated. .

5. The non-contact method for monitoring the quality and behavioral characteristics of live shrimp according to claim 4, characterized in that, S5 further includes the steps of: calculating motion speed and classifying behavior; The steps for calculating the motion speed include: Leveraging the ID correlation provided by multi-object tracking, we can obtain the same ID in the current frame t and the previous frame. coordinates of the center point of the bounding box and ; Calculate pixel displacement ; Combining the scaling factor k and the video frame interval time Calculate the instantaneous velocity v: ； The steps for classifying behavior include: Based on preset speed thresholds, the instantaneous motion state of shrimp is divided into four categories: Rest: ; crawl: ; Swimming: ; catapult: .

6. The non-contact method for monitoring the quality and behavioral characteristics of live shrimp according to claim 5, characterized in that, The S5 also includes a behavioral statistics step: performing statistical analysis on the cumulative duration and proportion of each behavioral state to quantify the overall vitality level of the group.

7. The non-contact method for monitoring the quality and behavioral characteristics of live shrimp according to claim 5 or 6, characterized in that, The quality rating is determined based on the following thresholds: excellent: and ; good: and ; generally: or ; Poor: and ; in, The threshold for curl degree grading. The threshold for color deviation grading.

8. A non-contact system for monitoring the quality and behavioral characteristics of live shrimp, characterized in that, include: An image acquisition device is fixedly installed above the aquaculture pond to acquire video frames by taking a single-view overhead photograph of the monitoring area. A computer processing unit is configured to perform the steps S2 to S5 as described in claim 6; The data storage and visualization module is used to save raw images, processing results, and analysis logs, and provides a graphical user interface for real-time display of monitoring footage.

Images