A method and device for identifying fish diseases in marine ranches based on improved YOLOv8

Abstract

The present invention provides a method and device for identifying fish diseases in a marine ranch based on improved YOLOv8, the method comprising: collecting first underwater images of a marine ranch at different positions at the same time and inputting them into a pre-built neural network model for fish disease identification, and outputting an initial fish disease identification result; performing a hue test on the first identification area of ​​the initial fish disease identification result and splicing the initial fish disease identification results that have undergone the hue test to obtain a first fish disease identification result; performing a disease cause analysis on the first fish disease identification result according to environmental data to generate a fish disease cause analysis report; constructing a fish movement trajectory based on the first fish disease identification results at different times, performing an activity analysis on fish with fish diseases according to fish activity data and movement trajectories, generating fish health assessment indicators, and generating a fish disease identification result in combination with the fish disease cause analysis report and the fish health assessment indicators. The present invention can improve the reliability and convenience of fish disease identification.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a method and device for identifying fish diseases in marine ranches based on improved YOLOv8. Background Art

[0002] Marine ranching refers to an area in the marine environment where artificial facilities and management measures are used to cultivate, breed and protect marine organisms. Through scientific breeding and management, marine ranching can optimize the structure of the marine ecosystem, increase biodiversity and reduce the impact of natural environmental changes on organisms. In the development of marine ranching, fish diseases are prone to occur due to environmental factors, poor management and food quality. For example, environmental factors such as water quality deterioration, temperature fluctuations and abuse of antibiotics may cause fish health problems, while poor feed quality, high stocking density and unsanitary facilities may also cause fish diseases. The outbreak of fish diseases may lead to large-scale fish deaths, resulting in direct economic losses. Antibiotics or other chemicals may be needed during the occurrence and control of diseases. If these drugs are not handled properly, they may have negative effects on the marine environment, such as drug residues and water pollution. Furthermore, if the drugs and chemicals used in the control of fish diseases are not properly handled, they will also pose a potential risk to human health.

[0003] In order to solve these problems, existing technologies usually use microscopic observation, laboratory testing (such as bacteriology, PCR testing), water quality chemical testing and other means to identify and predict possible aquatic product diseases in marine ranching. However, these traditional methods require a long detection time and are highly dependent on professional equipment, making it difficult to carry out large-scale monitoring in the marine ranch environment. Summary of the invention

[0004] The present invention aims to provide a method and device for identifying fish diseases in marine ranches based on improved YOLOv8, so as to solve the above-mentioned technical problems and improve the reliability and convenience of fish disease identification.

[0005] In order to solve the above technical problems, the present invention provides a method for identifying fish diseases in marine ranches based on improved YOLOv8, comprising:

[0006] Collecting first underwater images of the ocean ranch at different locations at the same time, and inputting the first underwater images into a pre-built neural network model for fish disease recognition, and outputting initial fish disease recognition results for each of the underwater images; wherein the neural network model is built based on the YOLOv8 neural network;

[0007] Performing a hue test on the initial fish disease recognition results in a first recognition area of ​​the first underwater image, eliminating initial fish disease recognition results that do not conform to the hue test, and splicing the initial fish disease recognition results that have undergone the hue test to generate a first fish disease recognition result of the marine ranch;

[0008] Acquire environmental data of the marine ranch, perform disease cause analysis on the first fish disease identification result according to the environmental data, and generate a fish disease cause analysis report;

[0009] Based on the first fish disease identification results of the ocean ranch at different times, the movement trajectory of the fish in the ocean ranch is constructed, and the activity analysis of the fish with fish diseases in the ocean ranch is performed according to the fish activity data of the ocean ranch and the movement trajectory to generate fish health assessment indicators, and the fish disease identification results of the ocean ranch are generated in combination with the fish disease cause analysis report and the fish health assessment indicators.

[0010] In the above scheme, by collecting underwater images at different locations of the marine ranch at the same time and inputting the improved YOLOv8 neural network model for fish disease identification, a large amount of image data can be quickly processed to achieve real-time detection. Compared with the traditional method that requires a long time for sample processing, this step greatly improves the detection efficiency. The hue test of the initial fish disease identification results can eliminate the identification results that do not meet the standards, thereby improving the accuracy of identification. The identification results that have been spliced ​​after the hue test form a comprehensive disease identification image, which reduces the possibility of false detection and missed detection. This refining process makes the final result more accurate and reliable. Combining the environmental data of the marine ranch for disease cause analysis can provide a deeper understanding of the causes of fish diseases. This analysis can identify the impact of environmental factors on fish diseases, thereby providing a scientific cause analysis report and providing a basis for formulating targeted management measures. Based on the disease identification results of fish at different time points, the motion trajectory is constructed, and the activity analysis is combined with the fish activity data to track the changes in the health status of the fish, and the activity and health level of the fish can be evaluated to form a comprehensive health assessment index. Compared with traditional offline detection, this dynamic analysis can more timely reflect the health status of the fish. The disease cause analysis report and fish health assessment indicators are combined to generate the final fish disease identification results. This comprehensive result not only takes into account the actual identification of the disease, but also incorporates environmental and activity analysis data to provide a comprehensive and accurate fish disease assessment. This comprehensive analysis can better support management decisions and optimize marine ranch health management and disease prevention and control strategies.

[0011] In one implementation, the neural network model is constructed based on the YOLOv8 neural network, specifically including:

[0012] Obtaining an initial network structure of the YOLOv8 neural network; wherein the initial network structure includes a backbone network, a feature pyramid network, and a detection head;

[0013] A fuzzy layer is set at the output end of the backbone network; wherein the fuzzy layer is used to fuzzy the initial feature map output by the backbone network and output a first feature map after fuzzy processing;

[0014] Inputting the initial feature map and the first feature map into the feature pyramid network so that the feature pyramid network performs multi-scale feature fusion on the received initial feature map and the first feature map to obtain a plurality of first feature fusion maps;

[0015] The first feature map is input into the detection head so that the detection head predicts fish diseases on the first feature fusion map to generate the initial fish disease recognition result.

[0016] In one implementation, the detection head performs fish disease prediction on the first feature fusion image to generate the initial fish disease recognition result, specifically including:

[0017] Obtaining the types of marine fish diseases, and adjusting the output dimension of the fully connected layer in the detection head according to the number of the types;

[0018] The initial feature map, the first feature map and the first feature fusion map are input into the fully connected layer, so that the fully connected layer performs fish disease prediction based on a preset activation function to generate the initial fish disease recognition result; wherein the initial fish disease recognition result includes a fish disease category label, a bounding box and a confidence level.

[0019] In one implementation, performing a hue test on the initial fish disease recognition result in the first recognition area of ​​the first underwater image and discarding the initial fish disease recognition result that does not meet the hue test specifically includes:

[0020] Acquire the first recognition area where each of the initial fish disease recognition results is located in the first underwater image one by one, and extract the pixel value of each pixel point in each channel in the first recognition area;

[0021] The pixel value is input into a preset database for traversal query, and when there is at least one pixel point in the first recognition area whose pixel value in each channel meets the preset pixel value range, it is determined that the initial fish disease recognition result passes the hue test; wherein the preset database is divided into pixel value ranges corresponding to several types of fish disease category labels;

[0022] Eliminating initial fish disease identification results that fail the hue test, and generating fish disease identification results for each of the first underwater images;

[0023] The fish disease recognition results of the first underwater images at different positions are spliced ​​to generate the first fish disease recognition result of the marine ranch.

[0024] In one implementation, performing disease cause analysis on the first fish disease identification result according to the environmental data to generate a fish disease cause analysis report specifically includes:

[0025] Acquire the marine ranch environmental data at the time when the first fish disease identification result is collected; wherein the environmental data includes water quality data and fish body surface feature data;

[0026] Constructing environmental characteristic variables based on the marine ranch environmental data and inputting the environmental characteristic variables into a preset regression model so that the regression model performs disease cause analysis and generates a predicted value of the degree of influence of each environmental characteristic variable relative to the first fish disease identification result;

[0027] The fish disease cause analysis report is generated based on the predicted value of the impact degree.

[0028] In one implementation, constructing the movement trajectory of fish in the marine ranch based on the first fish disease identification results at different times of the marine ranch specifically includes:

[0029] Obtaining the first fish disease identification result of the ocean ranch at each moment and extracting bounding box information in the fish disease identification result;

[0030] Generate fish position information in the marine ranch at each moment based on the bounding box information;

[0031] The fish position information at different times is sequentially spliced ​​to generate the motion trajectory.

[0032] In one implementation, the activity analysis of fish with fish diseases in the marine ranch is performed based on the fish activity data and the movement trajectory of the marine ranch to generate fish health assessment indicators, which specifically includes:

[0033] Obtaining fish activity data of the marine ranch at the different times; wherein the fish activity data includes swimming patterns, group activities and feeding behaviors;

[0034] Constructing a number of activity evaluation indicators based on the fish activity data and the movement trajectory;

[0035] Normalizing each of the activity evaluation indicators to obtain a normalized activity evaluation indicator;

[0036] A weighted operation is performed on each of the normalized activity assessment indicators based on a preset weight to generate the fish health assessment indicator.

[0037] In a second aspect, the present application also provides a marine ranch fish disease identification device based on improved YOLOv8, including: a fish disease identification module, a test splicing module, a first analysis module and a second analysis module;

[0038] The fish disease identification module is used to collect the first underwater images of the marine ranch at different locations at the same time, and input the first underwater images into a pre-built neural network model for fish disease identification, and output the initial fish disease identification result of each underwater image; wherein the neural network model is built based on the YOLOv8 neural network;

[0039] The inspection and splicing module is used to perform a hue inspection on the initial fish disease identification results in the first identification area of ​​the first underwater image, eliminate the initial fish disease identification results that do not meet the hue inspection, and splice the initial fish disease identification results that have undergone the hue inspection to generate the first fish disease identification result of the marine ranch;

[0040] The first analysis module is used to obtain environmental data of the marine ranch, perform disease cause analysis on the first fish disease identification result according to the environmental data, and generate a fish disease cause analysis report;

[0041] The second analysis module is used to construct the movement trajectory of fish in the ocean ranch based on the first fish disease identification results of the ocean ranch at different times, perform activity analysis on fish with fish diseases in the ocean ranch according to the fish activity data of the ocean ranch and the movement trajectory, generate fish health assessment indicators, and generate fish disease identification results of the ocean ranch in combination with the fish disease cause analysis report and the fish health assessment indicators.

[0042] In the above scheme, by collecting underwater images at different locations of the marine ranch at the same time and inputting the improved YOLOv8 neural network model for fish disease identification, a large amount of image data can be quickly processed to achieve real-time detection. Compared with the traditional method that requires a long time for sample processing, this step greatly improves the detection efficiency. The hue test of the initial fish disease identification results can eliminate the identification results that do not meet the standards, thereby improving the accuracy of identification. The identification results that have been spliced ​​after the hue test form a comprehensive disease identification image, which reduces the possibility of false detection and missed detection. This refining process makes the final result more accurate and reliable. Combining the environmental data of the marine ranch for disease cause analysis can provide a deeper understanding of the causes of fish diseases. This analysis can identify the impact of environmental factors on fish diseases, thereby providing a scientific cause analysis report and providing a basis for formulating targeted management measures. Based on the disease identification results of fish at different time points, the motion trajectory is constructed, and the activity analysis is combined with the fish activity data to track the changes in the health status of the fish, and the activity and health level of the fish can be evaluated to form a comprehensive health assessment index. Compared with traditional offline detection, this dynamic analysis can more timely reflect the health status of the fish. The disease cause analysis report and fish health assessment indicators are combined to generate the final fish disease identification results. This comprehensive result not only takes into account the actual identification of the disease, but also incorporates environmental and activity analysis data to provide a comprehensive and accurate fish disease assessment. This comprehensive analysis can better support management decisions and optimize marine ranch health management and disease prevention and control strategies.

[0043] In one implementation, the neural network model is constructed based on the YOLOv8 neural network, specifically including:

[0044] Obtaining an initial network structure of the YOLOv8 neural network; wherein the initial network structure includes a backbone network, a feature pyramid network, and a detection head;

[0045] A fuzzy layer is set at the output end of the backbone network; wherein the fuzzy layer is used to fuzzy the initial feature map output by the backbone network and output a first feature map after fuzzy processing;

[0046] Inputting the initial feature map and the first feature map into the feature pyramid network so that the feature pyramid network performs multi-scale feature fusion on the received initial feature map and the first feature map to obtain a plurality of first feature fusion maps;

[0047] The first feature map is input into the detection head so that the detection head predicts fish diseases on the first feature fusion map to generate the initial fish disease recognition result.

[0048] In one implementation, the detection head performs fish disease prediction on the first feature fusion image to generate the initial fish disease recognition result, specifically including:

[0049] Obtaining the types of marine fish diseases, and adjusting the output dimension of the fully connected layer in the detection head according to the number of the types;

[0050] The initial feature map, the first feature map and the first feature fusion map are input into the fully connected layer, so that the fully connected layer performs fish disease prediction based on a preset activation function to generate the initial fish disease recognition result; wherein the initial fish disease recognition result includes a fish disease category label, a bounding box and a confidence level.

[0051] In one implementation, performing a hue test on the initial fish disease recognition result in the first recognition area of ​​the first underwater image and discarding the initial fish disease recognition result that does not meet the hue test specifically includes:

[0052] Acquire the first recognition area where each of the initial fish disease recognition results is located in the first underwater image one by one, and extract the pixel value of each pixel point in each channel in the first recognition area;

[0053] The pixel value is input into a preset database for traversal query, and when there is at least one pixel point in the first recognition area whose pixel value in each channel meets the preset pixel value range, it is determined that the initial fish disease recognition result passes the hue test; wherein the preset database is divided into pixel value ranges corresponding to several types of fish disease category labels;

[0054] Eliminating initial fish disease identification results that fail the hue test, and generating fish disease identification results for each of the first underwater images;

[0055] The fish disease recognition results of the first underwater images at different positions are spliced ​​to generate the first fish disease recognition result of the marine ranch.

[0056] In one implementation, performing disease cause analysis on the first fish disease identification result according to the environmental data to generate a fish disease cause analysis report specifically includes:

[0057] Acquire the marine ranch environmental data at the time when the first fish disease identification result is collected; wherein the environmental data includes water quality data and fish body surface feature data;

[0058] Constructing environmental characteristic variables based on the marine ranch environmental data and inputting the environmental characteristic variables into a preset regression model so that the regression model performs disease cause analysis and generates a predicted value of the degree of influence of each environmental characteristic variable relative to the first fish disease identification result;

[0059] The fish disease cause analysis report is generated based on the predicted value of the impact degree.

[0060] In one implementation, constructing the movement trajectory of fish in the marine ranch based on the first fish disease identification results at different times of the marine ranch specifically includes:

[0061] Obtaining the first fish disease identification result of the ocean ranch at each moment and extracting bounding box information in the fish disease identification result;

[0062] Generate fish position information in the marine ranch at each moment based on the bounding box information;

[0063] The fish position information at different times is sequentially spliced ​​to generate the motion trajectory.

[0064] In one implementation, the activity analysis of fish with fish diseases in the marine ranch is performed based on the fish activity data and the movement trajectory of the marine ranch to generate fish health assessment indicators, which specifically includes:

[0065] Obtaining fish activity data of the marine ranch at the different times; wherein the fish activity data includes swimming patterns, group activities and feeding behaviors;

[0066] Constructing a number of activity evaluation indicators based on the fish activity data and the movement trajectory;

[0067] Normalizing each of the activity evaluation indicators to obtain a normalized activity evaluation indicator;

[0068] A weighted operation is performed on each of the normalized activity assessment indicators based on a preset weight to generate the fish health assessment indicator.

[0069] In a third aspect, the present application also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the marine ranch fish disease identification method based on the improved YOLOv8 as described above.

[0070] In a fourth aspect, the present application also provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the marine ranch fish disease identification method based on the improved YOLOv8 as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic flow chart of a method for identifying fish diseases in marine ranches based on improved YOLOv8 provided in one embodiment of the present invention;

[0072] Figure 2 The present invention is a schematic diagram of a module of a marine ranch fish disease identification device based on improved YOLOv8 provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0073] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0074] The terms "first" and "second" and the like in the specification, claims and drawings of this application are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0075] 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.

[0076] Example 1

[0077] See also Figure 1 , Figure 1 The flowchart of a method for identifying fish diseases in marine ranches based on improved YOLOv8 is provided in one embodiment of the present invention. The embodiment of the present invention provides a method for identifying fish diseases in marine ranches based on improved YOLOv8, including steps 101 to 104, each of which is as follows:

[0078] Step 101: Collect first underwater images of the ocean ranch at different locations at the same time, input the first underwater images into a pre-built neural network model for fish disease recognition, and output an initial fish disease recognition result for each of the underwater images; wherein the neural network model is built based on the YOLOv8 neural network.

[0079] In an embodiment of the present invention, high-resolution underwater cameras are pre-installed at different locations of the ocean ranch, and pictures of the ocean ranch at different locations are collected at regular intervals by the installed underwater cameras to obtain different perspectives and states of fish at the same time.

[0080] In one embodiment, the neural network model is constructed based on the YOLOv8 neural network, specifically including: obtaining the initial network structure of the YOLOv8 neural network; wherein the initial network structure includes a backbone network, a feature pyramid network and a detection head; setting a fuzzy layer at the output end of the backbone network; wherein the fuzzy layer is used to fuzzy the initial feature map output by the backbone network and output a first feature map after fuzzy processing; inputting the initial feature map and the first feature map into the feature pyramid network so that the feature pyramid network performs multi-scale feature fusion on the received initial feature map and the first feature map to obtain several first feature fusion maps; inputting the first feature map into the detection head so that the detection head performs fish disease prediction on the first feature fusion map to generate the initial fish disease identification result.

[0081] In the initial network structure of the YOLOv8 neural network, the backbone network is used to extract features from the input image, the feature pyramid network is used to combine feature maps from different levels for feature fusion, and the detection head is used to generate predictions, including bounding box regression, category prediction, and confidence score. It should be noted that the above initial network structure is a conventional network structure of the YOLOv8 neural network, and its specific network structure composition can be adjusted according to the specific application scenario requirements, and the specific composition of each structure is not limited here. In the embodiment of the present invention, a fuzzy layer is introduced after the backbone network, i.e., the feature extraction part, to ensure that the high-level features extracted from the original image have been fuzzy processed before entering the feature fusion stage. By introducing fuzzy algorithms and fuzzy layers, the model can better handle fuzziness and uncertainty in the image. This is particularly important for identifying lesion areas, because lesion features may become blurred or unclear due to image quality, lighting conditions, or the nature of the lesion itself. Specifically, a fuzzy convolution kernel is set in the fuzzy layer, and its kernel weight can be optimized through training. The convolution kernel will be applied to the feature map output by the backbone network, and the size and shape of the fuzzy convolution kernel should be adjusted according to the size and shape of the fish disease features. In the blur layer, the feature map is convolved with the blur convolution kernel to obtain the blurred first feature map. This feature map emphasizes the blurred features of the lesion area. The blurred feature map is fused with the original feature map output by the backbone network. This can be achieved by simple weighted averaging, splicing or other feature fusion techniques. The fused feature map will contain both the original features and the blurred lesion features. In the feature pyramid network (FPN) of YOLOv8, the fused feature map is used for further feature fusion. This ensures that the blurred features can participate in the final detection task. By fusing the blurred features with the original features, the model can use the information of both features at the same time. This fusion increases the richness and diversity of the features, allowing the model to capture and identify lesion features more comprehensively. The detection head is used to predict fish diseases based on the output first feature fusion map of the feature pyramid network to generate fish disease prediction results.

[0082] In one embodiment, the detection head predicts fish diseases on the first feature fusion map to generate the initial fish disease recognition result, specifically including: obtaining the types of marine fish diseases, and adjusting the output dimension of the fully connected layer in the detection head according to the number of the types; inputting the initial feature map, the first feature map and the first feature fusion map into the fully connected layer, so that the fully connected layer predicts fish diseases based on a preset activation function to generate the initial fish disease recognition result; wherein the initial fish disease recognition result includes a fish disease category label, a bounding box and a confidence level.

[0083] The detection head in the YOLOv8 neural network is generally set with a fully connected layer for classification. If the detection head part of the YOLOv8 neural network does not have a dedicated classification fully connected layer, a new fully connected layer can be added for classification. According to the diversity of marine fish diseases, determine the number of categories output by the classification head, that is, the fully connected layer. For example, if there are 10 disease types and 1 health state, the total number of categories is 11. The feature map after feature extraction and fusion will be flattened into a one-dimensional vector as the input of the fully connected layer. The purpose of the fully connected layer is to map these high-dimensional feature vectors to the final output space. Then, the activation function Softmax function is used after the fully connected layer. The output of the fully connected layer is converted into a category probability chalk through the activation function, so that the sum of the probabilities of each category is 1, and finally a complete initial fish disease recognition result is obtained. Among them, the bounding box is used to predict the position of the object bounding box, indicating the position of the fish disease area in the image; the category label is used to identify the category label of each detected area, indicating the specific type of fish disease; the confidence score is the confidence score of each detected area, which is used to indicate the confidence of the model in the detection result. Preferably, when generating the initial fish disease recognition results, the bounding box, category label and confidence score can be superimposed on the original image to intuitively display the fish disease recognition results of the neural network model. By adjusting the detection head of the YOLOv8 model to adapt to the diversity of marine fish diseases, the model's ability to recognize various disease features can be significantly improved.

[0084] Step 102: Perform a hue test on the initial fish disease identification results in the first identification area of ​​the first underwater image, eliminate the initial fish disease identification results that do not conform to the hue test, and splice the initial fish disease identification results that have undergone the hue test to generate the first fish disease identification results of the marine ranch.

[0085] In one embodiment, the hue test is performed on the initial fish disease identification result in the first identification area of ​​the first underwater image, specifically including: obtaining the first identification area where each initial fish disease identification result is located in the first underwater image one by one, and extracting the pixel value of each pixel in the first identification area in each channel; inputting the pixel value into a preset database for traversal query, and when there is at least one pixel in the first identification area whose pixel value in each channel meets the preset pixel value range, it is determined that the initial fish disease identification result passes the hue test; wherein the preset database is divided into pixel value ranges corresponding to several types of fish disease category labels; the initial fish disease identification results that fail the hue test are eliminated to generate the fish disease identification result of each first underwater image; the fish disease identification results of the first underwater images at different positions are spliced ​​to generate the first fish disease identification result of the marine ranch.

[0086] The pixel value range of a healthy fish is relatively stable, usually expressed as a specific RGB value, while common fish disease changes include white spot disease, red tide disease, skin disease, etc. In the white spot disease state, white spots may appear on the surface of the fish, resulting in a significant increase in the RGB value of the corresponding area, especially the red and blue components; in the red tide disease state, the fish may appear red or brown, resulting in an increase in the R value of the relevant pixels; in the skin disease state, the surface of the fish may become dark or gray, and the average RGB value may drop significantly. In an embodiment of the present invention, a preset database is constructed according to the pixel value range of the fish in a healthy state and different fish disease states, which is used to perform hue inspection on different underwater color photos to determine whether there is a pixel value range of fish disease in the picture. Exemplarily, the pixel value range of a healthy fish is R: 100-255, G: 100-255, B: 100-255. The pixel value range of white spot disease is R: 200-255, G: 150-255, B: 150-255; the pixel value range of red tide disease is R: 180-255, G: 50-100, B: 30-80; the pixel value range of skin disease is R: 50-150, G: 50-150, B: 50-150. The pixel values ​​of the pixels in the first identification area in each channel are input into the pre-built database for traversal query. As long as there is any pixel in the first identification area whose pixel value in each channel meets the pixel value range corresponding to the fish disease category label in the initial fish disease identification result, it can be judged that there are indeed diseased fish in the area and pass the hue test. For the initial fish disease identification results that do not pass the hue test, it is considered that the initial fish disease identification results output by the neural network model are wrong and are eliminated. The hue test is performed on the first identification area where each initial fish disease identification result is located, and the fish disease identification result of each first underwater image at different positions is generated. Then, according to the shooting position of each first underwater image, the fish disease identification results corresponding to each first underwater image are spliced ​​to obtain the first fish disease identification result of the entire marine ranch.

[0087] Step 103: Acquire environmental data of the marine ranch, perform disease cause analysis on the first fish disease identification result according to the environmental data, and generate a fish disease cause analysis report.

[0088] The method of performing disease cause analysis on the first fish disease identification result according to the environmental data and generating a fish disease cause analysis report specifically includes: acquiring the marine ranch environmental data at the time when the first fish disease identification result is collected; wherein the environmental data includes water quality data and fish surface characteristic data; constructing environmental characteristic variables based on the marine ranch environmental data and inputting the environmental characteristic variables into a preset regression model so that the regression model performs disease cause analysis and generates a predicted value of the degree of influence of each of the environmental characteristic variables relative to the first fish disease identification result; and generating the fish disease cause analysis report based on the predicted value of the degree of influence.

[0089] In an embodiment of the present invention, water quality sensors are installed at various locations of the marine ranch to monitor chemical and physical parameters in the water, such as pH value, dissolved oxygen (DO), ammonia nitrogen, etc. At the same time, a camera is used to capture the surface features of the fish, and the appearance of the fish, such as scales, skin lesions, etc., is analyzed by image processing technology. The marine ranch water quality data and fish surface feature data at different times can be obtained by the above means. Relevant environmental characteristic variables are constructed according to the acquired environmental data, and a target variable is constructed based on the first fish disease identification result. The environmental characteristic variables include pH value, dissolved oxygen, ammonia nitrogen, lesion area, and lesion type; the target variable is whether there is a fish disease. The environmental characteristic variables are used as the output of the regression model, and the first fish disease identification result, that is, whether there is a fish disease, is used as the target variable. Preferably, the embodiment of the present invention uses a linear regression model to analyze the cause of the disease, and checks the coefficients of the linear regression model to understand the degree of influence of each environmental characteristic variable on the disease. The coefficient value reflects the degree of influence of environmental factors on the disease. A larger coefficient value indicates that the factor has a greater influence on the severity of the disease. Furthermore, the influence of the change of each environmental characteristic variable on the prediction result can also be calculated, so as to further understand the sensitivity of each environmental characteristic variable to the severity of the disease. As an optimization scheme of an embodiment of the present invention, management establishment can also be provided based on the fish disease cause analysis report. For example, if ammonia nitrogen has a significant effect on the severity of the disease, it is recommended to strengthen water quality monitoring and ammonia nitrogen control.

[0090] Step 104: construct movement trajectories of fish in the ocean ranch based on the first fish disease identification results at different times of the ocean ranch, perform activity analysis on fish with fish diseases in the ocean ranch according to the fish activity data of the ocean ranch and the movement trajectories, generate fish health assessment indicators, and generate fish disease identification results of the ocean ranch in combination with the fish disease cause analysis report and the fish health assessment indicators.

[0091] In one embodiment, the movement trajectory of fish in the ocean ranch is constructed based on the first fish disease identification result of the ocean ranch at different times, specifically including: obtaining the first fish disease identification result of the ocean ranch at each time and extracting the bounding box information in the fish disease identification result; generating the fish position information in the ocean ranch at each time based on the bounding box information; and sequentially splicing the fish position information at different times to generate the movement trajectory.

[0092] In an embodiment of the present invention, the fish disease recognition results corresponding to the first underwater images at different positions are spliced ​​by using image registration technology to generate a complete fish disease recognition result of the marine ranch at a moment, that is, the first fish disease recognition result. The first fish disease recognition result at each moment is obtained and the bounding box information therein is extracted, and the bounding box information includes the bounding box coordinates and center position of the fish with the fish disease. The first fish disease recognition result is converted into position information through the extracted bounding box information. Then, the fish positions at different moments are spliced ​​to generate the fish movement trajectories at different moments in the marine ranch.

[0093] In one embodiment, the activity analysis of fish with fish diseases in the marine ranch is performed based on the fish activity data and the movement trajectory of the marine ranch to generate a fish health assessment index, which specifically includes: obtaining the fish activity data of the marine ranch at the different times; wherein the fish activity data includes swimming patterns, group activities and feeding behaviors; constructing a number of activity assessment indicators based on the fish activity data and the movement trajectory; normalizing each of the activity assessment indicators to obtain a normalized activity assessment index; and performing a weighted operation on each of the normalized activity assessment indicators based on preset weights to generate the fish health assessment index.

[0094] In an embodiment of the present invention, fish activity data at different times of the marine ranch are collected in advance. Among them, the swimming mode includes the swimming speed, direction change, etc. of the fish; the group activity is the clustering behavior and fish density of the fish school; the feeding behavior includes the number of times the fish eats, the feeding time, the feeding duration, etc. Several activity evaluation indicators are constructed based on the collected fish activity data and motion trajectory. Preferably, in an embodiment of the present invention, the activity evaluation indicators include swimming mode indicators, group activity indicators, feeding behavior indicators and motion trajectory indicators. Specifically, when calculating the swimming mode index of the fish, the swimming speed and direction change of the fish in different time periods can be calculated to obtain the average swimming speed, directional stability (for example, the standard deviation of the direction change) and other indicators as the swimming mode index. When calculating the group activity index, the fish school density in each time period is calculated to obtain the average value and volatility (for example, standard deviation) of the fish school density. When calculating the feeding behavior, the number of feedings and the duration can be counted to obtain the feeding frequency, average feeding time and other indicators. When calculating the motion trajectory index, the average movement distance and volatility (for example, standard deviation) can be obtained by calculating the position change distance of the fish per unit time. Each activity evaluation index is normalized. Specifically, the Z-Score normalization method is used for normalization in the embodiment of the present invention. The normalized activity evaluation index is obtained by performing normalization operation through the average value and standard deviation of the index at different times. After all the indicators are normalized, the fish health evaluation index is obtained by weighted average method. Among them, the weight of each activity evaluation index can be allocated according to the importance of the fish activity data corresponding to each activity evaluation index in the specific marine biological activity to the fish health, which is not limited here. The obtained fish disease cause analysis report and fish health evaluation index are used as the fish disease identification result of the entire marine ranch.

[0095] In an embodiment of the present invention, a marine ranch fish disease identification device based on improved YOLOv8 is also provided, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the above-mentioned marine ranch fish disease identification method based on improved YOLOv8 is implemented.

[0096] In an embodiment of the present invention, a computer-readable storage medium is also provided, and the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned marine ranch fish disease identification method based on improved YOLOv8.

[0097] Exemplarily, the computer program can be divided into one or more modules, one or more modules are stored in a memory and executed by a processor to complete the present invention. One or more modules can be a series of computer program instruction segments that can perform specific functions, and the instruction segments are used to describe the execution process of the computer program in the marine ranch fish disease identification device based on the improved YOLOv8.

[0098] The marine ranch fish disease identification device based on improved YOLOv8 can be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The marine ranch fish disease identification device based on improved YOLOv8 may include, but is not limited to, a processor, a memory, and a display. Those skilled in the art will understand that the above-mentioned components are merely examples of marine ranch fish disease identification devices based on improved YOLOv8, and do not constitute a limitation on marine ranch fish disease identification devices based on improved YOLOv8. More or fewer components may be included than the components, or some components may be combined, or different components may be included. For example, the marine ranch fish disease identification device based on improved YOLOv8 may also include input and output devices, network access devices, buses, etc.

[0099] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the marine ranch fish disease identification device based on the improved YOLOv8, and uses various interfaces and lines to connect various parts of the marine ranch fish disease identification device based on the improved YOLOv8.

[0100] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the marine ranch fish disease identification device based on the improved YOLOv8 by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, a text conversion function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text message data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0101] Among them, if the module for identifying fish diseases in marine ranches based on improved YOLOv8 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0102] The embodiment of the present invention provides a method for identifying fish diseases in marine ranches based on improved YOLOv8. By collecting underwater images at different locations in the marine ranch at the same time and inputting the improved YOLOv8 neural network model for fish disease identification, a large amount of image data can be quickly processed to achieve real-time detection. Compared with the traditional method that requires a long time for sample processing, this step greatly improves the detection efficiency. By performing a hue test on the initial fish disease identification results, the identification results that do not meet the standards can be eliminated, thereby improving the accuracy of identification. The identification results that have been spliced ​​after the hue test are formed to form a comprehensive disease identification image, reducing the possibility of false detection and missed detection. This refining process makes the final result more accurate and reliable. By combining the environmental data of the marine ranch to analyze the causes of the disease, a deeper understanding of the causes of the fish disease can be obtained. This analysis can identify the impact of environmental factors on fish diseases, thereby providing a scientific cause analysis report and providing a basis for formulating targeted management measures. Based on the disease identification results of fish at different time points, the motion trajectory is constructed, and the activity analysis is performed in combination with the fish activity data, which can track the changes in the health status of the fish, and can also evaluate the activity and health level of the fish to form a comprehensive health assessment index. Compared with traditional offline detection, this dynamic analysis can reflect the health status of fish more timely. The disease cause analysis report and fish health assessment indicators are combined to generate the final fish disease identification results. This comprehensive result not only takes into account the actual identification of the disease, but also incorporates environmental and activity analysis data to provide a comprehensive and accurate fish disease assessment. This comprehensive analysis can better support management decisions and optimize the health management and disease prevention and control strategies of marine ranches.

[0103] Example 2

[0104] See also Figure 2 , Figure 2 The module diagram of a marine ranch fish disease identification device based on improved YOLOv8 provided in one embodiment of the present invention. The embodiment of the present invention provides a marine ranch fish disease identification device based on improved YOLOv8, including: a fish disease identification module 201, a test splicing module 202, a first analysis module 203 and a second analysis module 204;

[0105] The fish disease identification module 201 is used to collect the first underwater images of the marine ranch at different locations at the same time, and input the first underwater images into a pre-built neural network model for fish disease identification, and output the initial fish disease identification result of each underwater image; wherein the neural network model is built based on the YOLOv8 neural network;

[0106] The inspection and splicing module 202 is used to perform a hue inspection on the initial fish disease identification results in the first identification area of ​​the first underwater image, remove the initial fish disease identification results that do not meet the hue inspection, and splice the initial fish disease identification results that have undergone the hue inspection to generate the first fish disease identification results of the marine ranch;

[0107] The first analysis module 203 is used to obtain environmental data of the marine ranch, perform disease cause analysis on the first fish disease identification result according to the environmental data, and generate a fish disease cause analysis report;

[0108] The second analysis module 204 is used to construct the movement trajectory of fish in the ocean ranch based on the first fish disease identification results of the ocean ranch at different times, perform activity analysis on fish with fish diseases in the ocean ranch according to the fish activity data of the ocean ranch and the movement trajectory, generate fish health assessment indicators, and generate fish disease identification results of the ocean ranch in combination with the fish disease cause analysis report and the fish health assessment indicators.

[0109] In one embodiment, the neural network model is constructed based on the YOLOv8 neural network, specifically including: obtaining the initial network structure of the YOLOv8 neural network; wherein the initial network structure includes a backbone network, a feature pyramid network and a detection head; setting a fuzzy layer at the output end of the backbone network; wherein the fuzzy layer is used to fuzzy the initial feature map output by the backbone network and output a first feature map after fuzzy processing; inputting the initial feature map and the first feature map into the feature pyramid network so that the feature pyramid network performs multi-scale feature fusion on the received initial feature map and the first feature map to obtain several first feature fusion maps; inputting the first feature map into the detection head so that the detection head performs fish disease prediction on the first feature fusion map to generate the initial fish disease identification result.

[0110] In one embodiment, the detection head predicts fish diseases on the first feature fusion map to generate the initial fish disease recognition result, specifically including: obtaining the types of marine fish diseases, and adjusting the output dimension of the fully connected layer in the detection head according to the number of the types; inputting the initial feature map, the first feature map and the first feature fusion map into the fully connected layer, so that the fully connected layer predicts fish diseases based on a preset activation function to generate the initial fish disease recognition result; wherein the initial fish disease recognition result includes a fish disease category label, a bounding box and a confidence level.

[0111] In one embodiment, the hue test is performed on the initial fish disease recognition result in the first recognition area of ​​the first underwater image, and the initial fish disease recognition result that does not meet the hue test is eliminated, which specifically includes: obtaining the first recognition area where each initial fish disease recognition result is located in the first underwater image one by one, and extracting the pixel value of each pixel in each channel in the first recognition area; inputting the pixel value into a preset database for traversal query, and when there is at least one pixel in the first recognition area whose pixel value in each channel meets the preset pixel value range, it is determined that the initial fish disease recognition result passes the hue test; wherein the preset database is divided into pixel value ranges corresponding to several types of fish disease category labels; eliminating the initial fish disease recognition results that do not pass the hue test, and generating the fish disease recognition result of each first underwater image; splicing the fish disease recognition results of the first underwater images at different positions to generate the first fish disease recognition result of the marine ranch.

[0112] In one embodiment, the first fish disease identification result is analyzed for the cause of the disease based on the environmental data to generate a fish disease cause analysis report, which specifically includes: obtaining the marine ranch environmental data at the time when the first fish disease identification result is collected; wherein the environmental data includes water quality data and fish surface characteristic data; constructing environmental characteristic variables based on the marine ranch environmental data and inputting the environmental characteristic variables into a preset regression model so that the regression model performs disease cause analysis and generates a predicted value of the degree of influence of each of the environmental characteristic variables relative to the first fish disease identification result; and generating the fish disease cause analysis report based on the predicted value of the degree of influence.

[0113] In one embodiment, the movement trajectory of fish in the ocean ranch is constructed based on the first fish disease identification result of the ocean ranch at different times, specifically including: obtaining the first fish disease identification result of the ocean ranch at each time and extracting the bounding box information in the fish disease identification result; generating the fish position information in the ocean ranch at each time based on the bounding box information; and sequentially splicing the fish position information at different times to generate the movement trajectory.

[0114] In one implementation, in one embodiment, the activity analysis of fish with fish diseases in the marine ranch is performed based on the fish activity data and the movement trajectory of the marine ranch to generate fish health assessment indicators, which specifically includes: obtaining the fish activity data of the marine ranch at the different times; wherein the fish activity data includes swimming patterns, group activities and feeding behaviors; constructing a number of activity assessment indicators based on the fish activity data and the movement trajectory; normalizing each of the activity assessment indicators to obtain a normalized activity assessment indicator; and performing a weighted operation on each of the normalized activity assessment indicators based on preset weights to generate the fish health assessment index.

[0115] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0116] The embodiment of the present invention provides a marine ranch fish disease identification device based on improved YOLOv8. By collecting underwater images at different locations in the marine ranch at the same time and inputting the improved YOLOv8 neural network model for fish disease identification, a large amount of image data can be quickly processed to achieve real-time detection. Compared with the traditional method that requires a long time for sample processing, this step greatly improves the detection efficiency. The initial fish disease identification results are subjected to hue inspection, and the identification results that do not meet the standards can be eliminated, thereby improving the accuracy of identification. The identification results that have been spliced ​​after hue inspection are formed to form a comprehensive disease identification image, which reduces the possibility of false detection and missed detection. This refining process makes the final result more accurate and reliable. The cause of the disease can be analyzed in combination with the environmental data of the marine ranch, and the cause of the fish disease can be more deeply understood. This analysis can identify the impact of environmental factors on fish diseases, thereby providing a scientific cause analysis report, which provides a basis for formulating targeted management measures. Based on the disease identification results of fish at different time points, the motion trajectory is constructed, and the activity analysis is performed in combination with the fish activity data, which can track the changes in the health status of the fish, and can also evaluate the activity and health level of the fish to form a comprehensive health assessment index. Compared with traditional offline detection, this dynamic analysis can reflect the health status of fish more timely. The disease cause analysis report and fish health assessment indicators are combined to generate the final fish disease identification results. This comprehensive result not only takes into account the actual identification of the disease, but also incorporates environmental and activity analysis data to provide a comprehensive and accurate fish disease assessment. This comprehensive analysis can better support management decisions and optimize the health management and disease prevention and control strategies of marine ranches.

[0117] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention. These improvements and substitutions should also be regarded as the scope of protection of the present invention.

Claims

1. A method for identifying fish diseases in marine ranches based on improved YOLOv8, characterized in that: include: Collecting first underwater images of the ocean ranch at different locations at the same time, and inputting the first underwater images into a pre-built neural network model for fish disease recognition, and outputting initial fish disease recognition results for each of the underwater images; wherein the neural network model is built based on the YOLOv8 neural network; Performing a hue test on the initial fish disease recognition results in a first recognition area of ​​the first underwater image, eliminating initial fish disease recognition results that do not conform to the hue test, and splicing the initial fish disease recognition results that have undergone the hue test to generate a first fish disease recognition result of the marine ranch; Acquire environmental data of the marine ranch, perform disease cause analysis on the first fish disease identification result according to the environmental data, and generate a fish disease cause analysis report; constructing the movement trajectory of the fish in the marine ranch based on the first fish disease identification results of the marine ranch at different times, performing activity analysis on the fish with fish diseases in the marine ranch according to the fish activity data of the marine ranch and the movement trajectory, generating fish health assessment indicators, and generating the fish disease identification result of the marine ranch in combination with the fish disease cause analysis report and the fish health assessment indicators; The neural network model is built based on the YOLOv8 neural network, specifically including: Obtaining an initial network structure of the YOLOv8 neural network; wherein the initial network structure includes a backbone network, a feature pyramid network, and a detection head; A fuzzy layer is set at the output end of the backbone network; wherein the fuzzy layer is used to fuzzy the initial feature map output by the backbone network and output a first feature map after fuzzy processing; Inputting the initial feature map and the first feature map into the feature pyramid network so that the feature pyramid network performs multi-scale feature fusion on the received initial feature map and the first feature map to obtain a plurality of first feature fusion maps; Inputting the first feature map to the detection head so that the detection head performs fish disease prediction on the first feature fusion map to generate the initial fish disease recognition result; The step of performing disease cause analysis on the first fish disease identification result according to the environmental data and generating a fish disease cause analysis report specifically includes: Acquire the marine ranch environmental data at the time when the first fish disease identification result is collected; wherein the environmental data includes water quality data and fish body surface feature data; Constructing environmental characteristic variables based on the marine ranch environmental data and inputting the environmental characteristic variables into a preset regression model so that the regression model performs disease cause analysis and generates a predicted value of the degree of influence of each environmental characteristic variable relative to the first fish disease identification result; Generate the fish disease cause analysis report based on the impact degree prediction value; The activity analysis of fish with fish diseases in the marine ranch is performed based on the fish activity data and the movement trajectory of the marine ranch to generate fish health assessment indicators, specifically including: Obtaining fish activity data of the marine ranch at the different times; wherein the fish activity data includes swimming patterns, group activities and feeding behaviors; Constructing a number of activity evaluation indicators based on the fish activity data and the movement trajectory; Normalizing each of the activity evaluation indicators to obtain a normalized activity evaluation indicator; A weighted operation is performed on each of the normalized activity assessment indicators based on a preset weight to generate the fish health assessment indicator.

2. A method for identifying fish diseases in marine ranches based on improved YOLOv8 as claimed in claim 1, characterized in that: The detection head performs fish disease prediction on the first feature fusion graph to generate the initial fish disease recognition result, specifically including: Obtaining the types of marine fish diseases, and adjusting the output dimension of the fully connected layer in the detection head according to the number of the types; The initial feature map, the first feature map and the first feature fusion map are input into the fully connected layer, so that the fully connected layer performs fish disease prediction based on a preset activation function to generate the initial fish disease recognition result; wherein the initial fish disease recognition result includes a fish disease category label, a bounding box and a confidence level.

3. A method for identifying fish diseases in marine ranches based on improved YOLOv8 as claimed in claim 2, characterized in that: The step of performing a hue test on the initial fish disease recognition result in the first recognition area of ​​the first underwater image and removing the initial fish disease recognition result that does not meet the hue test specifically includes: Acquire the first recognition area where each of the initial fish disease recognition results is located in the first underwater image one by one, and extract the pixel value of each pixel point in each channel in the first recognition area; The pixel value is input into a preset database for traversal query, and when there is at least one pixel point in the first recognition area whose pixel value in each channel meets the preset pixel value range, it is determined that the initial fish disease recognition result passes the hue test; wherein the preset database is divided into pixel value ranges corresponding to several types of fish disease category labels; Eliminating initial fish disease identification results that fail the hue test, and generating fish disease identification results for each of the first underwater images; The fish disease recognition results of the first underwater images at different positions are spliced ​​to generate the first fish disease recognition result of the marine ranch.

4. A method for identifying fish diseases in marine ranches based on improved YOLOv8 as claimed in claim 1, characterized in that: The step of constructing the movement trajectory of the fish in the marine ranch based on the first fish disease identification results at different times of the marine ranch specifically includes: Obtaining the first fish disease identification result of the ocean ranch at each moment and extracting bounding box information in the fish disease identification result; Generate fish position information in the marine ranch at each moment based on the bounding box information; The fish position information at different times is sequentially spliced ​​to generate the motion trajectory.

5. A marine ranch fish disease identification device based on improved YOLOv8, characterized in that: include: Fish disease identification module, inspection and splicing module, first analysis module and second analysis module; The fish disease identification module is used to collect the first underwater images of the marine ranch at different locations at the same time, and input the first underwater images into a pre-built neural network model for fish disease identification, and output the initial fish disease identification result of each underwater image; wherein the neural network model is built based on the YOLOv8 neural network; The inspection and splicing module is used to perform a hue inspection on the initial fish disease identification results in the first identification area of ​​the first underwater image, eliminate the initial fish disease identification results that do not meet the hue inspection, and splice the initial fish disease identification results that have undergone the hue inspection to generate the first fish disease identification result of the marine ranch; The first analysis module is used to obtain environmental data of the marine ranch, perform disease cause analysis on the first fish disease identification result according to the environmental data, and generate a fish disease cause analysis report; The second analysis module is used to construct the movement trajectory of the fish in the marine ranch based on the first fish disease identification results of the marine ranch at different times, perform activity analysis on the fish with fish diseases in the marine ranch according to the fish activity data of the marine ranch and the movement trajectory, generate fish health assessment indicators, and generate the fish disease identification result of the marine ranch in combination with the fish disease cause analysis report and the fish health assessment indicators; The neural network model is built based on the YOLOv8 neural network, specifically including: Obtaining an initial network structure of the YOLOv8 neural network; wherein the initial network structure includes a backbone network, a feature pyramid network, and a detection head; A fuzzy layer is set at the output end of the backbone network; wherein the fuzzy layer is used to fuzzy the initial feature map output by the backbone network and output a first feature map after fuzzy processing; Inputting the initial feature map and the first feature map into the feature pyramid network so that the feature pyramid network performs multi-scale feature fusion on the received initial feature map and the first feature map to obtain a plurality of first feature fusion maps; Inputting the first feature map to the detection head so that the detection head performs fish disease prediction on the first feature fusion map to generate the initial fish disease recognition result; The step of performing disease cause analysis on the first fish disease identification result according to the environmental data and generating a fish disease cause analysis report specifically includes: Acquire the marine ranch environmental data at the time when the first fish disease identification result is collected; wherein the environmental data includes water quality data and fish body surface feature data; Constructing environmental characteristic variables based on the marine ranch environmental data and inputting the environmental characteristic variables into a preset regression model so that the regression model performs disease cause analysis and generates a predicted value of the degree of influence of each environmental characteristic variable relative to the first fish disease identification result; Generate the fish disease cause analysis report based on the impact degree prediction value; The activity analysis of fish with fish diseases in the marine ranch is performed based on the fish activity data and the movement trajectory of the marine ranch to generate fish health assessment indicators, specifically including: Obtaining fish activity data of the marine ranch at the different times; wherein the fish activity data includes swimming patterns, group activities and feeding behaviors; Constructing a number of activity evaluation indicators based on the fish activity data and the movement trajectory; Normalizing each of the activity evaluation indicators to obtain a normalized activity evaluation indicator; A weighted operation is performed on each of the normalized activity assessment indicators based on a preset weight to generate the fish health assessment indicator.

6. A terminal device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for identifying fish diseases in marine ranches based on improved YOLOv8 as described in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the marine ranch fish disease identification method based on improved YOLOv8 as described in any one of claims 1 to 4.

Images