Improved YOLOv11-based oxygenation and aeration state detection method and system

Abstract

The invention discloses an oxygen aeration state detection method based on improved YOLOv11, and the method comprises the steps: 1, embedding a large-kernel separable convolution attention mechanism LSKA in an SPPF, forming an SPPF-LSKA composite module, replacing a detection head with a lightweight shared convolution-separated batch normalization detector LSCSBD, and obtaining an improved YOLOv11 model; 2, bubble images generated when the micropore oxygenation system works are collected, a training set and a test set are constructed, the improved YOLOv11 model is trained through the training set, and the performance of the model is verified through the test set; and step 3, applying the trained improved YOLOv11 model to state detection of an actual aeration system, including deploying the trained model to an edge computing device, collecting a target aeration area image in real time, performing model reasoning, judging an aeration state according to the detected bubble number, and outputting early warning information if the aeration state is abnormal. The system has strong anti-interference and high recognition capability in a complex breeding environment, training data is easy to construct, and the aeration state can be accurately monitored in real time.

Description

Technical Field

[0001] This invention relates to the field of visual recognition technology in aquaculture, and in particular to a method and system for detecting aeration status based on an improved YOLOv11. Background Technology

[0002] The aeration status of a microporous aeration system is the most direct way to assess whether the aerator is operating properly and whether the aeration is functioning correctly. Traditional manual inspection methods for monitoring the working status of microporous aeration systems suffer from low efficiency, strong subjectivity, and poor real-time performance. Acoustic detection methods are susceptible to interference when acquiring acoustic signals. However, image recognition methods can directly monitor the aerator's operation and also indicate whether the microporous aeration system has failed due to problems such as pipe connections or leaks. Summary of the Invention

[0003] This invention addresses the aforementioned problems by providing a method and system for detecting aeration status based on an improved YOLOv11 algorithm. In this invention, the YOLOv11n model refers to a model built upon the YOLO series target detection network architecture, the specific structure of which can be implemented with reference to the publicly available YOLOv8n network. This invention, considering the aeration characteristics of microporous aeration systems, utilizes images collected by pondside cameras and employs the improved YOLOv11 algorithm to determine and identify the aeration status. This enables rapid identification and detection of the working status of microporous aeration systems, offering fast detection speed, good implementability, and simple system construction.

[0004] This invention provides a method for detecting aeration status based on an improved YOLOv11, characterized by the following steps: Step 1, constructing an improved target detection model, in which a large-kernel separable convolutional attention mechanism (LSKA) is embedded in the Spatial Pyramid Pooling-Fast (SPPF) module to form an SPPF-LSKA composite module, and replacing the detection head with a lightweight shared convolutional-separated batch normalization (LSCSBD) detector. Step 1: Obtain the improved YOLOv11 model by detecting bubbles generated during the operation of the microporous aeration system; Step 2: Collect bubble images generated during the operation of the microporous aeration system, construct training and testing sets, train the improved YOLOv11 model using the training set, and verify the model performance using the testing set; Step 3: Apply the trained improved YOLOv11 model to the state detection of the actual aeration system, deploy the trained model to an edge computing device, collect images of the target aeration area in real time and perform model inference, determine the aeration state based on whether bubble clusters are detected, and output early warning information if abnormal.

[0005] Preferably, step 1 specifically includes: step 1-1, embedding the LSKA mechanism into the SPPF module to form an SPPF-LSKA composite module, wherein the LSKA mechanism is constructed by decomposing a large kernel convolution into multiple cascaded one-dimensional convolution kernels; step 1-2, replacing the traditional detection head with a lightweight shared convolution-separated batch normalization detector LSCSBD, wherein the LSCSBD adopts a "shared convolutional layer - independent batch normalization layer" mechanism and configures independent BN layers for bubble features at different levels.

[0006] Preferably, in step 1-1, the LSKA mechanism is used to capture long-range dependencies in image features while reducing computational load.

[0007] Preferably, in steps 1-2, the LSCSBD further includes a scale adaptation layer for adapting to detection targets of different scales.

[0008] Preferably, in step 2, the bubble images used to construct the training and test sets include bubble images of the micropore aeration system collected under different breeding environments, different lighting conditions, different shooting angles, and different distances.

[0009] Preferably, step 3 specifically includes: step 3-1, acquiring images or real-time video streams of the target aeration area; step 3-2, inputting the images or real-time video streams into a trained improved YOLOv11 model, with the model outputting the detection results of bubbles in the images; step 3-3, determining the aeration status based on whether bubble clusters are detected, specifically including: setting a preset bubble cluster quantity threshold; if the number of bubble clusters detected by the model is greater than or equal to the threshold, the aeration system is determined to be working normally; if the number of detected bubble clusters is less than the threshold, the aeration system is determined to be working abnormally; step 3-4, displaying the aeration status in real time, and if an abnormal status is determined, pushing early warning information to the operation and maintenance terminal.

[0010] Preferably, in step 3, the edge computing device acquires the video stream of the aeration area in real time through the RTSP protocol, extracts the images frame by frame and inputs them into the trained improved YOLOv11 model for model inference, and outputs the bounding boxes, confidence scores and bubble cluster counts of the bubbles in the bubble image.

[0011] This invention provides an aeration status detection system based on an improved YOLOv11, characterized by the following features: an image acquisition module for acquiring images or video streams of the aeration area; a model processing module equipped with a trained improved YOLOv11 model for detecting bubble clusters in the images acquired by the image acquisition module; and a status judgment and output module for judging the aeration status based on the detection results of the model processing module and outputting the judgment results.

[0012] Preferably, the system is deployed on an edge computing device and is equipped with a local display interface and a remote communication interface.

[0013] The present invention provides an electronic device, including a memory and a processor. The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the method described above.

[0014] Technical effect Compared with the prior art, the oxygenation aeration status detection method and system based on the improved YOLOv11 of the present invention has the following characteristics and benefits: 1. This method is based on the YOLOv11-SPPF-LSKA-LSCSBD model, which has a strong ability to suppress background noise and can identify aeration status under complex aquaculture environments, strong light, poor shooting angles, etc.

[0015] 2. The training is relatively easy and the training dataset is easy to construct. By manually adjusting the working mode of the micro-hole aerator according to the location and shooting angle of the camera in the aquaculture pond, the image data required for the training dataset can be collected.

[0016] 3. The application of this method enables real-time detection and monitoring of the aeration status of the microporous aerator. The aeration status can reflect whether each component of the system, including the blower, pipeline, and aeration disc, is working properly. Attached Figure Description

[0017] The above and other objects, features, and advantages of this application will become more apparent from the following detailed description of the embodiments in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain the application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 This is a flowchart of the oxygenation and aeration status detection method based on the improved YOLOv11 in Embodiment 1 of the present invention; Figure 2 This is a structural diagram of the LSKA attention mechanism in Embodiment 1 of the present invention; Figure 3 This is a structural diagram of the LSCSBD module in Embodiment 1 of the present invention; Figure 4 This is a flowchart illustrating the application of the aeration status detection method based on the improved YOLO11 in Embodiment 2 of the present invention. Figure 5 This is a structural diagram of the aeration state detection model constructed based on the improved YOLO11 in Embodiment 2 of the present invention; Figure 6 This is a graph showing the accuracy changes of different target detection models during the training process in the comparative model of this invention; Figure 7 This is a comparison diagram of the water ripple interference experiment results between Embodiment 2 of the present invention and the comparative model, wherein... Figure 7 (a) Figure 7 (b) Figure 7 (c) Figure 7 (d) shows the experimental results of water ripple interference for the original image, YOLO, YOLOv11-LSCSBD, and YOLOv11-SPPF-LSKA-LSCSBD models, respectively.

[0019] Figure 8 This is a comparison diagram of the strong light interference experiment results between Embodiment 2 of the present invention and the comparative model, wherein... Figure 8 (a) Figure 8 (b) Figure 8 (c) The images show the experimental results of strong light interference for the original image, YOLOv11, and YOLOv11-SPPF-LSKA-LSCSBD models, respectively. Figure 9 This is a side-view experimental comparison diagram of Embodiment 2 of the present invention and the comparative model, wherein... Figure 9 (a) Figure 9 (b) Figure 9 (c) shows the side-view experimental results of the original image, YOLOv11, and YOLOv11-SPPF-LSKA-LSCSBD models, respectively. Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easy to understand, the following embodiments, in conjunction with the accompanying drawings, specifically illustrate a method and system for detecting aeration status based on an improved YOLOv11.

[0021] Example 1 This embodiment provides a method for detecting aeration status based on the improved YOLOv11.

[0022] Figure 1 This is a flowchart of an aeration status detection method based on the improved YOLOv11 in an embodiment of the present invention.

[0023] like Figure 1 As shown, the oxygenation and aeration status detection method based on the improved YOLOv11 in this embodiment includes the following steps: Step S1, Model Construction: Based on YOLOv11n, an improved object detection model is constructed. Specific improvements include: Step S1-1: The LSKA lightweight attention mechanism is embedded into the Spatial Pyramid Pooling - Fast (SPPF) module to form the SPPF-LSKA composite module. The Large Separable Kernel Attention (LSKA) captures long-range dependencies by decomposing the large kernel convolution into a cascaded one-dimensional kernel.

[0024] Figure 2 This is a structural diagram of the LSKA attention mechanism in Embodiment 1 of the present invention.

[0025] Figure 3 This is a structural diagram of the LSCSBD module in Embodiment 1 of the present invention; Step S1-2: Replace the traditional detection head with a lightweight shared convolutional-separated batch normalization detector (LSCSBD). The detection head adopts a "shared convolutional layer - independent batch normalization layer" mechanism, configures independent BN layers to adapt to bubble features at different levels, and solves the problem of inconsistent target scale through scale layers.

[0026] Step S2, Model Training: Train the model using the pre-built model. Specific steps include: Step S2-1, Dataset: Collect bubble images generated by the micropore aeration system under different breeding environments, lighting conditions, shooting angles and distances. After labeling with Label-Studio, construct the YOLO dataset and divide it into training and testing sets in a 7:3 ratio for training the improved target detection model.

[0027] Step S2-2: Adjust the dataset images to 640×640 pixels, set the batch size to 16 and the number of training rounds to 100, train the improved model, and verify the model performance through the test set. Step S3, Model Application: Apply the trained model to the state detection of the actual aeration system, specifically including the following: Step S3-1, Data Acquisition: Acquire images or real-time video streams of the target aeration area using a camera, ensuring that the acquisition range covers the working area of ​​the aeration equipment.

[0028] Step S3-2, Model Inference: Input the acquired video into the trained improved YOLOv11 model, and the model outputs the detection results of bubbles in the image.

[0029] Step S3-3, Aeration status determination: Set a threshold for the number of bubble clusters. If the number of bubble clusters detected by the model is greater than or equal to the threshold, the aeration system is determined to be working normally; if the number of bubble clusters is less than the threshold, the aeration system is determined to be abnormal.

[0030] Step S3-4, Result Output: Real-time display of aeration status (normal / abnormal). If it is determined to be abnormal, push warning information to the operation and maintenance terminal.

[0031] Furthermore, after completing the above steps, a working status detection model for the micropore aeration system is obtained. The model can be deployed according to actual aquaculture conditions to monitor the working status of the micropore aeration system.

[0032] Example 2 This embodiment provides an oxygenation and aeration status detection system based on the improved YOLOv11.

[0033] The working process of the aeration status detection system based on the improved YOLOv11 in this embodiment is as follows: Using steps S1-S3 in Example 1, bubble images generated by the microporous aeration system are acquired; and the bubble images are input into the aeration state detection model to obtain the detection results of bubbles in the output image of the aeration state detection model.

[0034] Figure 4 This is a flowchart illustrating the application of the aeration status detection method based on the improved YOLO11 in this embodiment of the invention.

[0035] Specifically, in this embodiment, as Figure 4 As shown, the trained model weight file and inference code are deployed to an edge computing device (Raspberry Pi), the image reading interface is configured (supporting real-time video stream reading and single image input), the inference speed is optimized, and the model processing frame rate is kept stable at more than 15fps to meet the real-time detection requirements.

[0036] Using a camera by the pond, images of bubbles are captured. The video stream of the aeration area is acquired in real time via the RTSP protocol. The edge computing device extracts the images frame by frame, performs model inference, and outputs the bounding boxes, confidence scores, and quantity statistics of the bubbles in the images.

[0037] On the local display screen of the edge computing device, a bubble detection box is displayed in real time. When an aeration abnormality is detected, an early warning information (including the time of the abnormality, the pond number, and a real-time screenshot) is pushed to the operation and maintenance terminal through the communication module to remind staff to troubleshoot the fault in time (such as checking the operation status of the blower, the sealing of the pipeline, and whether the aeration disc is blocked).

[0038] Figure 5 The diagram shows the structure of the aeration state detection model based on the improved YOLO11 in this embodiment of the invention.

[0039] like Figure 5 As shown, the overall architecture of the model is mainly divided into three core modules: Backbone, Neck (feature fusion layer), and Head (detection head).

[0040] Specifically, an attention mechanism LSKA module is introduced into the SPPF module of the YOLOv11 network model backbone to enhance the ability to capture long-range features, and the feature representation is further optimized through the C2PSA module.

[0041] Furthermore, the detection head is replaced by an LSCSBD module to detect the three scale feature maps output by Neck, and finally output the bounding box of the bubble (corresponding to the bubble annotation box in the Output on the right), so as to realize the identification and localization of bubbles in the aeration state.

[0042] The working principle is as follows: Images of bubbles or real-time video from the micropore aeration system, captured by cameras deployed in the aquaculture ponds, are transmitted over a network to an edge computing device. A trained model runs on the edge computing device, processing the input images at a rate of 15 frames per second in real time. It extracts bubble features from the images and uses built-in logic to identify key information such as the number and distribution of bubbles. Based on the identified bubble features, the model determines whether the aeration system is functioning correctly. If a specified number of bubbles are not detected for five consecutive seconds, it indicates a potential malfunction in the aeration system (such as clogged aeration discs or equipment shutdown). The system immediately triggers an alarm, prompting staff to investigate the fault and enabling real-time monitoring and early warning of abnormalities in the aeration system's operating status.

[0043] Figure 6 This is a graph showing the accuracy changes of different target detection models during the training process in the comparative model of this invention; Figure 7 This is a comparison diagram of the water ripple interference experiment results between Embodiment 2 of the present invention and the comparative model, wherein... Figure 7 (a) Figure 7 (b) Figure 7 (c) Figure 7 (d) shows the experimental results of water ripple interference for the original image, YOLO, YOLOv11-LSCSBD, and YOLOv11-SPPF-LSKA-LSCSBD models, respectively.

[0044] Figure 8 This is a comparison diagram of the strong light interference experiment results between Embodiment 2 of the present invention and the comparative model, wherein... Figure 8 (a) Figure 8 (b) Figure 8 (c) The images show the experimental results of strong light interference for the original image, YOLOv11, and YOLOv11-SPPF-LSKA-LSCSBD models, respectively. Figure 9 This is a side-view experimental comparison diagram of Embodiment 2 of the present invention and the comparative model, wherein... Figure 9 (a) Figure 9 (b) Figure 9 (c) shows the side-view experimental results of the original image, YOLOv11, and YOLOv11-SPPF-LSKA-LSCSBD models, respectively.

[0045] like Figures 6-9The figures show the performance comparison results of the improved model and the comparative model in Embodiments 1 and 2 of the present invention under various interference environments such as water ripples, strong light, and side view, which together verify the advantages of the model of the present invention in terms of accuracy and robustness.

[0046] Table 1 shows the ablation experiment results in this embodiment.

[0047] Table 1 Table 2 shows the results of the comparative experiments in this embodiment.

[0048] Table 2 As shown in Tables 1 and 2, these are the ablation experiment results of the model of the present invention under different component configurations and the comparative experiment results with a variety of mainstream models, which comprehensively verify the overall superiority of the model of the present invention in terms of detection accuracy, recall rate and computational efficiency.

[0049] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0050] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0051] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0052] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0053] The above description has been given for illustrative and descriptive purposes. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for detecting aeration status based on an improved YOLOv11, characterized in that, Includes the following steps: Step 1: Construct an improved target detection model. The model embeds a large kernel separable convolutional attention mechanism LSKA in the fast spatial pyramid pooling module SPPF to form an SPPF-LSKA composite module. The detection head is replaced with a lightweight shared convolutional-separated batch normalization detector LSCSBD to obtain an improved YOLOv11 model. Step 2: Collect bubble images generated during the operation of the microporous oxygenation system, construct training and testing sets, train the improved YOLOv11 model using the training set, and verify the model performance using the testing set. Step 3: Apply the trained improved YOLOv11 model to the state detection of the actual aeration system, including: The trained model is deployed to an edge computing device to collect images of the target aeration area in real time and perform model inference. The aeration status is determined based on whether bubble clusters are detected, and an early warning message is output if there is an abnormality.

2. The method for detecting aeration status based on improved YOLOv11 according to claim 1, characterized in that: Step 1 specifically includes: Step 1-1: Embed the LSKA mechanism into the SPPF module to form an SPPF-LSKA composite module, wherein the LSKA mechanism is constructed by decomposing a large kernel convolution into multiple cascaded one-dimensional convolution kernels; Steps 1-2: Replace the traditional detection head with a lightweight shared convolutional-separated batch normalization detector (LSCSBD). The LSCSBD adopts a "shared convolutional layer - independent batch normalization layer" mechanism and configures independent BN layers for bubble features at different levels.

3. The method for detecting aeration status based on improved YOLOv11 according to claim 2, characterized in that: In step 1-1, the LSKA mechanism is used to capture long-range dependencies in image features while reducing computational load.

4. The method for detecting aeration status based on improved YOLOv11 according to claim 2, characterized in that: In steps 1-2, the LSCSBD also includes a scale adaptation layer for adapting to detection targets of different scales.

5. The method for detecting aeration status based on improved YOLOv11 according to claim 1, characterized in that: In step 2, the bubble images used to construct the training and test sets include bubble images of the micropore aeration system collected under different breeding environments, different lighting conditions, different shooting angles, and different distances.

6. The method for detecting aeration status based on improved YOLOv11 according to claim 1, characterized in that: Step 3 specifically includes: Step 3-1: Acquire images or real-time video streams of the target aeration area; Step 3-2: Input the image or real-time video stream into the trained improved YOLOv11 model, and the model outputs the detection results of bubbles in the image; Step 3-3: Determine the aeration status based on whether bubble clusters are detected. Specifically, this includes: setting a preset bubble cluster quantity threshold; if the number of bubble clusters detected by the model is greater than or equal to the threshold, the aeration system is determined to be working normally; if the number of detected bubble clusters is less than the threshold, the aeration system is determined to be working abnormally. Steps 3-4: Display the aeration status in real time. If an abnormal status is detected, push an early warning message to the operation and maintenance terminal.

7. The method for detecting aeration status based on improved YOLOv11 according to claim 1, characterized in that: In step 3, the edge computing device acquires the video stream of the aeration area in real time through the RTSP protocol, extracts the images frame by frame and inputs them into the trained improved YOLOv11 model for model inference, and outputs the bounding boxes, confidence scores and bubble cluster counts of the bubbles in the bubble image.

8. An aeration status detection system based on an improved YOLOv11, characterized in that, For implementing the method as described in any one of claims 1 to 7, comprising: The image acquisition module is used to acquire images or video streams of the aeration area; The model processing module is equipped with a pre-trained improved YOLOv11 model, which is used to perform bubble cluster target detection on the images acquired by the image acquisition module. The status judgment and output module is used to judge the aeration status based on the detection results of the model processing module and output the judgment result.

9. The oxygenation and aeration status detection system based on the improved YOLOv11 according to claim 8, characterized in that: The system is deployed on an edge computing device and is equipped with a local display interface and a remote communication interface.

10. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method as described in any one of claims 1-7.

Images