Fish disease detection method and system based on deep learning image processing technology

Abstract

The invention relates to a fish disease detection method based on a deep learning image processing technology. The method comprises the following steps: collecting a fish disease image and carrying out image preprocessing; segmenting the preprocessed image by adopting a layered Mama model so as to extract a fish disease feature region; and performing feature extraction on the segmented disease feature region by using a MobileNet model, and performing disease classification through a full connection layer to obtain probability distribution of disease categories. According to the method, the hierarchical Mama model is adopted for image segmentation, the MobileNet model is adopted for disease classification, the fish disease detection efficiency and accuracy can be effectively improved, meanwhile, the challenges of poor image quality, limited training data and fish disease diversity can be coped with, and high-precision segmentation results and recognition are achieved under the condition of limited training data. According to the invention, accurate disease diagnosis can be provided, breeding operators can be helped to take measures in time, and loss caused by diseases is reduced.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a fish disease detection method and system based on deep learning image processing technology. Background Art

[0002] With the rapid development of aquaculture around the world, the early detection and diagnosis of fish diseases has become a key link in ensuring aquaculture benefits, preventing the spread of diseases and maintaining ecological balance. In this context, the demand for efficient and accurate fish disease detection methods is becoming increasingly urgent. Traditional fish disease detection methods mainly rely on the empirical observation and manual diagnosis of farmers. This method is not only time-consuming and labor-intensive, but also greatly affected by subjective factors, prone to misjudgment or missed diagnosis, and difficult to meet the needs of modern aquaculture for rapid response to diseases.

[0003] At present, although there have been some attempts to detect fish diseases based on image analysis, most of these technologies remain at the level of simple image processing and pattern recognition, such as using basic image filtering, edge detection and other methods to identify certain abnormal characteristics of fish. These methods often have poor results when processing complex and changeable fish disease images, and it is difficult to accurately distinguish different types of disease characteristics, which limits their application in actual production.

[0004] Traditional fish disease detection methods have the following defects: 1. Low efficiency: Traditional manual observation methods are time-consuming and cannot meet the rapid detection needs of large-scale farms. 2. Insufficient accuracy: Detection methods that rely on manual experience and basic image processing technology are easily affected by factors such as environment and lighting, resulting in a high misdiagnosis rate. 3. Low degree of automation: Lack of efficient automated detection methods makes it difficult to achieve real-time monitoring and early warning. Summary of the invention

[0005] In view of this, it is necessary to provide a fish disease detection method and system based on deep learning image processing technology to solve the above-mentioned defects of the prior art.

[0006] In order to solve the above problems, in a first aspect, an embodiment of the present invention provides a fish disease detection method based on deep learning image processing technology, comprising:

[0007] Step S1, collecting fish disease images and performing image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations;

[0008] Step S2, using a hierarchical Mamba model to segment the preprocessed image to extract fish disease characteristic areas;

[0009] Step S3: Use the MobileNet model to extract features from the segmented disease feature areas, and classify the diseases through the fully connected layer to obtain the probability distribution of the disease categories.

[0010] Preferably, in step S2, segmenting the preprocessed image using a hierarchical Mamba model comprises:

[0011] Input the preprocessed image into the hierarchical Mamba model;

[0012] The hierarchical Mamba model extracts image features layer by layer and aggregates contextual information;

[0013] Output the segmentation mask map of the fish disease characteristic area to identify the disease characteristic area in the image.

[0014] Preferably, the hierarchical Mamba model adopts a multi-layer architecture to process fish disease images of different resolutions through four stages, gradually extracting and aggregating image features;

[0015] The fish disease images input to the hierarchical Mamba model are divided into 4x4 non-overlapping patches and embedded into a feature vector, and the fish disease images are converted into feature maps composed of feature vectors;

[0016] The feature map is processed by multiple Mamba blocks, each of which includes a Mamba operation layer, a channel similarity modulation operator, and a feed-forward network.

[0017] Preferably, in step S3, the use of the MobileNet model to extract features from the segmented disease feature areas and classify the diseases through a fully connected layer specifically includes:

[0018] Use the MobileNet model to extract features from the segmented disease feature areas;

[0019] The extracted features are input into the fully connected layer for disease classification;

[0020] The output of the fully connected layer is converted into the probability distribution of disease categories through the Softmax activation function.

[0021] Preferably, before performing feature extraction on the segmented disease feature area using the MobileNet model in step S3, the method further comprises:

[0022] The segmented fish disease feature areas are scaled, cropped, and normalized to fit the input size and range of the MobileNet model.

[0023] Preferably, after obtaining the probability distribution of the disease categories in step S3, the method further comprises:

[0024] Step S4, using multiple evaluation indicators to evaluate the classification performance of the fish disease detection model, and using a confusion matrix to analyze the performance of the model on different disease categories;

[0025] The fish disease detection model is composed of the hierarchical Mamba model and the MobileNet model; the evaluation indicators include accuracy, recall, F1 score and AUC-ROC; the F1 score is the harmonic mean of precision and recall, and AUC-ROC is the area under the receiver operating characteristic curve.

[0026] Preferably, after the performance evaluation of the fish disease detection model is completed in step S4, the method further comprises:

[0027] Step S5, converting the fish disease detection model that has passed the performance evaluation into a format suitable for deployment, and deploying it on a server or a mobile device to provide fish disease detection services.

[0028] In a second aspect, an embodiment of the present invention provides a fish disease detection system based on deep learning image processing technology, comprising:

[0029] An image preprocessing module, used for collecting fish disease images and performing image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations;

[0030] An image segmentation module is used to segment the preprocessed image using a hierarchical Mamba model to extract fish disease characteristic areas;

[0031] The image classification module is used to extract features from the segmented disease feature areas using the MobileNet model, and classify diseases through the fully connected layer to obtain the probability distribution of disease categories.

[0032] In a third aspect, the present invention further provides an electronic device, including a memory and a processor, wherein:

[0033] The memory is used to store programs;

[0034] The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the fish disease detection method based on deep learning image processing technology as described in the embodiment of the first aspect of the present invention.

[0035] In a fourth aspect, the present invention further provides a computer-readable storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the fish disease detection method based on deep learning image processing technology as described in the embodiment of the first aspect of the present invention.

[0036] The fish disease detection method and system based on deep learning image processing technology provided by the present invention adopts a hierarchical Mamba model for image segmentation and a MobileNet model for disease classification, which can effectively improve the efficiency and accuracy of fish disease detection, and can cope with the challenges of poor image quality, limited training data and diversity of fish diseases, and achieve high-precision segmentation results and recognition under limited training data conditions. The present invention realizes an automated and efficient fish disease detection system, which can monitor and warn the occurrence of fish diseases in real time, improve the automation level of fish disease detection, reduce the workload of breeding personnel, and improve production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flow chart of the fish disease detection method based on deep learning image processing technology provided by the present invention;

[0038] Figure 2 A schematic diagram of the structure of the fish disease detection model provided by the present invention;

[0039] Figure 3 A structural block diagram of a fish disease detection system based on deep learning image processing technology provided by the present invention;

[0040] Figure 4 This is a structural block diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0041] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0042] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0043] At present, the traditional method of fish disease detection mainly relies on the empirical observation and manual diagnosis of farmers. This method is not only time-consuming and labor-intensive, but also greatly influenced by subjective factors, prone to misjudgment or missed diagnosis, and difficult to meet the needs of modern aquaculture for rapid response to diseases.

[0044] In view of this, the present invention provides a fish disease detection method based on deep learning image processing technology, which uses a hierarchical Mamba model for image segmentation and a MobileNet model for disease classification, which can effectively improve the efficiency and accuracy of fish disease detection and realize automatic and efficient fish disease detection. The following will be explained and introduced through multiple embodiments.

[0045] Figure 1 This is a flow chart of the fish disease detection method based on deep learning image processing technology provided by the present invention. Figure 1 As shown, the fish disease detection method based on deep learning image processing technology includes at least steps S1 to S3, wherein:

[0046] Step S1, collecting fish disease images and performing image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations.

[0047] Specifically, in actual operation, the diseased fish can be placed in a separate breeding container in advance, and professional underwater photography equipment or smart phones and other shooting tools can be used to take images of the diseased fish in the breeding container to obtain fish disease images. During the image acquisition process, it is necessary to ensure that the image is clear and can clearly show the body surface characteristics of the fish, including color, texture, shape, etc.

[0048] After the fish disease image acquisition is completed, the collected fish disease images are subjected to denoising, image enhancement and size standardization operations. Among them, the denoising operation can use filters (such as Gaussian filtering, median filtering) to denoise the image to remove noise and impurities in the image and improve the image quality. The image enhancement operation can enhance the image features and make the disease features more obvious by adjusting parameters such as image contrast, brightness and saturation. The size standardization operation can adjust all images to a uniform size to meet the requirements of subsequent model input. Image resizing can be achieved using image scaling algorithms (such as bilinear interpolation, nearest neighbor interpolation, etc.).

[0049] In the embodiment of the present invention, denoising, image enhancement and size standardization operations are performed on the collected fish disease images, which can improve the quality of the fish disease images, highlight the disease characteristics, and provide high-quality input data for subsequent image segmentation and disease classification.

[0050] Step S2, using the hierarchical Mamba model to segment the preprocessed image to extract fish disease characteristic areas.

[0051] Figure 2 This is a schematic diagram of the fish disease detection model provided by the present invention. Figure 2 The fish disease detection model provided by the embodiment of the present invention is composed of a hierarchical Mamba model and a MobileNet model. Figure 2 In the figure, Norm refers to the normalization layer, Linear refers to the fully connected layer, FFN refers to the feedforward network, and Mamba Block refers to the Mamba block.

[0052] In step S2, the image segmentation process may include:

[0053] S21, input the preprocessed image into the hierarchical Mamba model.

[0054] S22, the hierarchical Mamba model extracts image features layer by layer and aggregates contextual information.

[0055] Specifically, the Hierarchical Mamba model is a deep learning model that uses a hierarchical architecture to process images of different resolutions. This design allows the model to capture image features at multiple scales, thereby improving the ability to understand the image content. At each stage of the model, image features are extracted layer by layer through structures such as convolutional layers and downsampling layers. These features include the shape, texture, color, etc. of the fish, as well as abnormal changes that may be caused by diseases.

[0056] While extracting image features, the Hierarchical Mamba model also aggregates contextual information. This means that the model considers the relationship between different regions in the image and how they together form a complete disease feature. Contextual information aggregation is achieved through the control hierarchical Mamba blocks in the model, which include Mamba operation layers, channel similarity modulation operators, and feedforward networks. Together, they improve the model's parameter efficiency and stability when processing visual state space models, and enhance cross-channel communication and improve feature aggregation.

[0057] S23, outputting a segmentation mask image of the fish disease characteristic region to identify the disease characteristic region in the image.

[0058] Specifically, after layer-by-layer feature extraction and context information aggregation, the hierarchical Mamba model will output a segmentation mask map. This mask map is a two-dimensional array of the same size as the input image, in which each element indicates whether the corresponding position in the input image belongs to the disease characteristic area. The segmentation mask map provides an accurate target area for subsequent disease classification by identifying the disease characteristic area in the image. This helps to reduce the interference of background noise and improve the accuracy of classification.

[0059] like Figure 2 As shown in Figure 1, the hierarchical Mamba model uses a multi-layer architecture to process fish disease images of different resolutions through four stages, gradually extracting and aggregating image features. This hierarchical design allows the model to capture image features at different scales, thereby improving the ability to understand image content. The model usually contains four processing stages, each stage processes images of different resolutions, and as the network goes deeper, the number of channels gradually increases to capture more complex feature representations.

[0060] The fish disease images input to the hierarchical Mamba model are divided into 4x4 non-overlapping patches and embedded into a feature vector, and the fish disease images are converted into feature maps composed of feature vectors. This process helps the model process image data more efficiently and lays the foundation for subsequent feature extraction and classification tasks.

[0061] The feature map is processed by multiple Mamba blocks, each of which includes a Mamba operation layer, a channel similarity modulation operator, and a feed-forward network.

[0062] Among them, the Mamba operation layer can process the input feature map through linear transformation and selective mechanism based on the selective state space model (SSM). It can intelligently choose which information to pay attention to or ignore, thereby improving the efficiency and accuracy of the model.

[0063] The channel similarity modulation operator is used to enhance cross-channel communication and improve feature aggregation. By dynamically adjusting the feature weights between different channels, it can improve the model's ability to capture and aggregate complex features. This helps the model more accurately identify and segment these features when processing fish images with multiple disease features.

[0064] The feed-forward network is responsible for further processing the feature map after the Mamba operation layer and the channel similarity modulation operator. It generates the output feature map of the Mamba block and provides input for the next stage of processing.

[0065] In a preferred embodiment of the present invention, the hierarchical Mamba model adopts a multi-layer architecture to process fish disease images of different resolutions through four stages. The specific process is as follows:

[0066] Reference Figure 2 In the first stage, the fish disease images input to the hierarchical Mamba model are embedded into a feature vector of dimension 1 through two 3x3 convolutional layers, and the feature map passes through a downsampling layer, which merges the blocks within the 2×2 area to generate a feature map of size H / 4×W / 4×C1H / 4×W / 4×C1. In the first stage, the image size is reduced to 1 / 4 of the original image, generating a lower resolution feature map to capture the low-level features of the image and provide a basis for subsequent processing.

[0067] In the second stage, the feature map passes through a downsampling layer again to further reduce the size. The feature map is then fed into the Mamba block for processing. Each Mamba block contains a Mamba operation layer, a channel similarity modulation operator, and a feedforward network. The feature map size output by this stage is H / 8×W / 8×C2H / 8×W / 8×C2. The second stage achieves capturing image features at a more abstract level and enhances the extraction and aggregation capabilities of features through special operation layers in the Mamba block.

[0068] In the third stage, the process of the second stage is repeated, and the downsampling layer is applied again, followed by the Mamba block. The feature map size output in this stage is H / 16×W / 16×C3H / 16×W / 16×C3; the size of the feature map is further reduced to 1 / 16 of the original image. In this embodiment, the image features are further abstracted and aggregated in the third stage to capture higher-level semantic information.

[0069] In the fourth stage, the feature map is downsampled again, and then, through the Mamba block, the final feature map size is output as H / 32×W / 32×C4.

[0070] In the embodiment of the present invention, the powerful feature extraction and context information aggregation capabilities of the hierarchical Mamba model are used to accurately identify and extract disease feature areas from preprocessed fish images. This provides a solid foundation for subsequent disease classification and helps to achieve efficient and accurate fish disease detection.

[0071] Step S3: Use the MobileNet model to extract features from the segmented disease feature areas, and classify the diseases through the fully connected layer to obtain the probability distribution of the disease categories.

[0072] Specifically, in step S3, the image classification process includes:

[0073] S31, use the MobileNet model to extract features from the segmented disease feature areas.

[0074] The MobileNet model is a lightweight deep learning architecture, which is particularly suitable for efficient image classification tasks in resource-constrained environments. In an embodiment of the present invention, the MobileNet model is used to further extract features from the fish disease feature areas segmented by the hierarchical Mamba model.

[0075] In this embodiment, the MobileNet model achieves efficient feature extraction and calculation through depthwise separable convolution.

[0076] S32, input the extracted features into the fully connected layer for disease classification.

[0077] After being processed by the deep separable convolutional layer of the MobileNet model, the extracted features are input into the fully connected layer. The function of the fully connected layer is to map the extracted features to the probability space of disease categories.

[0078] S33, the output of the fully connected layer is converted into the probability distribution of disease categories through the Softmax activation function.

[0079] In this embodiment, the Softmax activation function is used to convert the output of the fully connected layer into the probability distribution of disease categories. The Softmax function compresses the output value of each category between 0 and 1, and the sum of the output values ​​of all categories is 1. In this way, each output value can be interpreted as the probability of belonging to the corresponding category.

[0080] Finally, the output after processing by the Softmax activation function will be interpreted as the classification result of fish diseases. The disease category with the highest probability will be selected as the final disease diagnosis result.

[0081] In this embodiment, the MobileNet model is used to efficiently extract and classify the segmented disease feature areas. The computational optimization is achieved through deep separable convolution, the fully connected layer maps features to categories, and the Softmax activation function obtains the final classification probability distribution, thus realizing the automatic detection of fish diseases.

[0082] The fish disease detection method based on deep learning image processing technology provided by the present invention uses a hierarchical Mamba model for image segmentation and a MobileNet model for disease classification, which can effectively improve the efficiency and accuracy of fish disease detection, and can cope with the challenges of poor image quality, limited training data and diversity of fish diseases, and achieve high-precision segmentation results and recognition under limited training data conditions. The present invention realizes an automated and efficient fish disease detection system that can monitor and warn the occurrence of fish diseases in real time, improves the automation level of fish disease detection, reduces the workload of breeders, and improves production efficiency.

[0083] In a preferred embodiment of the present invention, before using the MobileNet model to extract features from the segmented disease feature area in step S3, the fish disease detection method based on deep learning image processing technology provided by the present invention further includes:

[0084] The segmented fish disease feature areas are scaled, cropped, and normalized to fit the input size and range of the MobileNet model.

[0085] It is understandable that the MobileNet model has specific requirements for the size and range of the input image. Typically, these requirements include the width, height, and range of pixel values ​​of the image. In order to meet these requirements, the segmented disease feature regions need to be scaled, cropped, and normalized.

[0086] In this embodiment, through preprocessing steps such as scaling, cropping and normalization, it can be ensured that the segmented feature area matches the input requirements of the MobileNet model, thereby improving the classification performance and accuracy of the model.

[0087] In a preferred embodiment of the present invention, after obtaining the probability distribution of disease categories in step S3, the fish disease detection method based on deep learning image processing technology provided by the present invention further includes:

[0088] In step S4, a variety of evaluation indicators are used to evaluate the classification performance of the fish disease detection model, and a confusion matrix is ​​used to analyze the performance of the model on different disease categories.

[0089] Specifically, in order to comprehensively and accurately evaluate the classification performance of the model, the present invention uses a variety of indicators, including accuracy, recall, F1 score, receiver operating characteristic curve (AUC-ROC), and uses confusion matrix to deeply analyze the specific performance of the model on different disease categories. The following is a detailed explanation of these evaluation indicators:

[0090] Accuracy is a basic indicator for evaluating model classification performance. It indicates the proportion of samples correctly classified by the model to the total number of samples. In fish disease detection, accuracy can reflect the overall classification ability of the model, that is, how many fish disease images the model can correctly identify.

[0091] Recall, also known as sensitivity, indicates the proportion of samples predicted by the model to be positive that are actually positive. In fish disease detection, recall can reflect the model's ability to identify positive samples (i.e., images of fish with diseases).

[0092] The F1 score is the harmonic mean of precision and recall, which takes into account the model's precision and recall, and is used to evaluate the model's performance in balancing precision and recall. In fish disease detection, the F1 score can reflect the model's ability to simultaneously consider accurately identifying diseased fish and reducing misdiagnosis. The higher the F1 score, the better the model performs in balancing these two aspects.

[0093] Receiver Operating Characteristic Curve (AUC-ROC): The ROC curve is a curve that reflects the relationship between the true positive rate (TPR) and the false positive rate (FPR) of the model at different thresholds. AUC (Area Under Curve) is the area under the ROC curve, which is used to quantify the classification performance of the model. In fish disease detection, ROC-AUC can reflect the ability of the model to distinguish between diseased fish and healthy fish. The closer the AUC value is to 1, the better the classification performance of the model.

[0094] The confusion matrix is ​​a matrix used to display the classification results of the model. It cross-compares the actual category with the predicted category to obtain four basic indicators: true positive (TP), false positive (FP), true negative (TN) and false negative (FN). Through the confusion matrix, we can intuitively see the performance of the model on different disease categories and identify possible deviations or deficiencies in the model. For example, if the false negative rate of the model in a certain disease category is high, it means that the model has a weak ability to identify this type of disease, and it may be necessary to further optimize the model or increase the training data of samples in this category.

[0095] In this embodiment, the accuracy, recall, F1 score, ROC-AUC index and confusion matrix are used to evaluate the classification performance of the model, which can comprehensively and accurately reflect the performance of the model in fish disease detection. These evaluation indicators not only help to understand the overall performance of the model, but also help technicians identify possible problems and deficiencies in the model, thereby providing support for the optimization and improvement of the model.

[0096] In a preferred embodiment of the present invention, after the performance evaluation of the fish disease detection model is completed in step S4, the method further comprises:

[0097] Step S5, converting the fish disease detection model that has passed the performance evaluation into a format suitable for deployment, and deploying it on a server or a mobile device to provide fish disease detection services.

[0098] Specifically, after the performance evaluation of the fish disease detection model is passed, the fish disease detection model can be put into use. The fish disease detection model needs to be converted into a format suitable for deployment. This is because different deployment environments (such as servers, mobile devices) may support different model formats and computing frameworks. In order to ensure that the model can run efficiently in these environments, it needs to be converted into a compatible format.

[0099] In actual operation, the format of the fish disease detection model can be converted to ONNX or TensorRT. Among them, ONNX (OpenNeural Network Exchange) is an open neural network model format that supports model conversion between multiple deep learning frameworks (such as PyTorch, TensorFlow, etc.). TensorRT is a library launched by NVIDIA for high-performance deep learning inference. It optimizes deep learning models into a format that can run efficiently on NVIDIA GPUs. Converting the model to TensorRT format can significantly improve the inference speed of the model on the GPU.

[0100] In this embodiment, the model that has passed the performance evaluation is converted into a format suitable for deployment and deployed on a server or mobile device, so that an efficient and accurate fish disease detection service can be achieved. At the same time, optimization for different deployment environments can ensure that the model can perform at its best in various scenarios.

[0101] Figure 3 This is a structural block diagram of the fish disease detection system based on deep learning image processing technology provided by the present invention, refer to Figure 3 , the fish disease detection system 300 based on deep learning image processing technology includes:

[0102] An image preprocessing module 301 is used to collect fish disease images and perform image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations;

[0103] An image segmentation module 302 is used to segment the pre-processed image using a hierarchical Mamba model to extract fish disease characteristic regions;

[0104] The image classification module 303 is used to extract features from the segmented disease feature areas using the MobileNet model, and classify the diseases through a fully connected layer to obtain a probability distribution of the disease categories.

[0105] The fish disease detection system based on deep learning image processing technology provided by the present invention executes the fish disease detection method based on deep learning image processing technology provided by the above-mentioned embodiments through the above-mentioned modules. The fish disease detection method based on deep learning image processing technology has been described in detail in the above-mentioned embodiments, and this embodiment will not be repeated here.

[0106] Figure 4 The structural block diagram of the electronic device provided by the present invention is as follows: Figure 4 As shown, the present invention also provides an electronic device, and the electronic device 400 can be a computing device such as a mobile terminal, a desktop computer, a notebook, a PDA, and a server. The electronic device 400 includes a processor 401 and a memory 402, wherein the memory 402 stores a fish disease detection program 403 based on deep learning image processing technology.

[0107] In some embodiments, the memory 402 may be an internal storage unit of a computer device, such as a hard disk or memory of the computer device. In other embodiments, the memory 402 may also be an external storage device of a computer device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the computer device. Furthermore, the memory 402 may also include both an internal storage unit of a computer device and an external storage device. The memory 402 is used to store application software and various types of data installed on the computer device, such as program codes for installing the computer device. The memory 402 may also be used to temporarily store data that has been output or is to be output. In one embodiment, when the fish disease detection program 403 based on deep learning image processing technology is executed by the processor 401, the following steps are implemented:

[0108] Step S1, collecting fish disease images and performing image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations;

[0109] Step S2, using a hierarchical Mamba model to segment the preprocessed image to extract fish disease characteristic areas;

[0110] Step S3: Use the MobileNet model to extract features from the segmented disease feature areas, and classify the diseases through the fully connected layer to obtain the probability distribution of the disease categories.

[0111] In some embodiments, the processor 401 can be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 402, such as executing a fish disease detection program based on deep learning image processing technology.

[0112] This embodiment also provides a computer-readable storage medium, on which a fish disease detection program based on deep learning image processing technology is stored. When the fish disease detection program based on deep learning image processing technology is executed by a processor, the following steps are implemented:

[0113] Step S1, collecting fish disease images and performing image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations;

[0114] Step S2, using a hierarchical Mamba model to segment the preprocessed image to extract fish disease characteristic areas;

[0115] Step S3: Use the MobileNet model to extract features from the segmented disease feature areas, and classify the diseases through the fully connected layer to obtain the probability distribution of the disease categories.

[0116] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fish disease detection method based on deep learning image processing technology, characterized in that: include: Step S1, collecting fish disease images and performing image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations; Step S2, using a hierarchical Mamba model to segment the preprocessed image to extract fish disease characteristic areas; Step S3: Use the MobileNet model to extract features from the segmented disease feature areas, and classify the diseases through the fully connected layer to obtain the probability distribution of the disease categories.

2. The fish disease detection method based on deep learning image processing technology according to claim 1 is characterized in that: In step S2, the use of the hierarchical Mamba model to segment the preprocessed image includes: Input the preprocessed image into the hierarchical Mamba model; The hierarchical Mamba model extracts image features layer by layer and aggregates contextual information; Output the segmentation mask map of the fish disease characteristic area to identify the disease characteristic area in the image.

3. The fish disease detection method based on deep learning image processing technology according to claim 1 is characterized in that: The hierarchical Mamba model adopts a multi-layer architecture to process fish disease images of different resolutions through four stages, gradually extracting and aggregating image features; The fish disease images input to the hierarchical Mamba model are divided into 4x4 non-overlapping patches and embedded into a feature vector, and the fish disease images are converted into feature maps composed of feature vectors; The feature map is processed by multiple Mamba blocks, each of which includes a Mamba operation layer, a channel similarity modulation operator, and a feed-forward network.

4. The fish disease detection method based on deep learning image processing technology according to claim 1 is characterized in that: In step S3, the MobileNet model is used to extract features from the segmented disease feature areas, and disease classification is performed through a fully connected layer, specifically including: Use the MobileNet model to extract features from the segmented disease feature areas; The extracted features are input into the fully connected layer for disease classification; The output of the fully connected layer is converted into the probability distribution of disease categories through the Softmax activation function.

5. The fish disease detection method based on deep learning image processing technology according to claim 1 is characterized in that: Before using the MobileNet model to extract features from the segmented disease feature areas in step S3, the method further includes: The segmented fish disease feature areas are scaled, cropped, and normalized to fit the input size and range of the MobileNet model.

6. The fish disease detection method based on deep learning image processing technology according to claim 5 is characterized in that: After obtaining the probability distribution of the disease categories in step S3, the method further includes: Step S4, using multiple evaluation indicators to evaluate the classification performance of the fish disease detection model, and using a confusion matrix to analyze the performance of the model on different disease categories; The fish disease detection model is composed of the hierarchical Mamba model and the MobileNet model; the evaluation indicators include accuracy, recall, F1 score and AUC-ROC; the F1 score is the harmonic mean of precision and recall, and AUC-ROC is the area under the receiver operating characteristic curve.

7. The fish disease detection method based on deep learning image processing technology according to claim 6 is characterized in that: After the performance evaluation of the fish disease detection model is completed in step S4, the method further comprises: Step S5, converting the fish disease detection model that has passed the performance evaluation into a format suitable for deployment, and deploying it on a server or a mobile device to provide fish disease detection services.

8. A fish disease detection system based on deep learning image processing technology, characterized in that: include: An image preprocessing module, used for collecting fish disease images and performing image preprocessing, wherein the image preprocessing includes denoising, image enhancement and size standardization operations; An image segmentation module is used to segment the preprocessed image using a hierarchical Mamba model to extract fish disease characteristic areas; The image classification module is used to extract features from the segmented disease feature areas using the MobileNet model, and classify diseases through the fully connected layer to obtain the probability distribution of disease categories.

9. An electronic device, It is characterized in that comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the fish disease detection method based on deep learning image processing technology as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the fish disease detection method based on deep learning image processing technology as described in any one of claims 1 to 7.

Images