Fish school disease degree identification method and device

Abstract

The invention provides a fish school disease degree identification method and device, and is applied to the technical field of aquaculture. The method comprises the steps of collecting fish school image data, performing data enhancement processing on the fish school image data, and constructing a fish disease recognition data set based on the processed fish school image data; training a fish disease recognition model through the fish disease recognition data set, wherein the fish disease recognition model is used for recognizing the disease degree grade of a fish school; wherein the fish disease identification model comprises a depth separable convolution module, a compression and excitation SE attention mechanism module and a cavity convolution DCLS module with learnable spacing, the depth separable convolution module is used for capturing fish body lesion area features, and the SE attention mechanism module is used for enhancing the fish body lesion area features. The DCLS module is used for dynamically adjusting a sampling interval to extract multi-scale features.

Description

Technical Field

[0001] The invention relates to the technical field of aquaculture, and in particular to a method and a device for identifying the disease degree of a fish school. Background Art

[0002] In aquaculture, the health of fish directly affects production efficiency and ecological balance. As the stocking density continues to increase, the risk of fish diseases is also increasing. If the degree of fish disease can be accurately identified in real time and in time, and reasonable measures can be taken according to the degree of disease, it can effectively prevent excessive treatment such as abuse of antibiotics, which is of great significance to ensuring the quality and safety of aquatic products and reducing aquaculture risks.

[0003] In the prior art, underwater fish image information can be collected in real time through a high-definition camera placed underwater. After the image processor pre-processes the fish image information, the Mali-T864 with the improved YOLOv5 algorithm embedded in the fish disease monitor is used to process the underwater image information and identify fish diseases, and feature extraction and classification are performed on images with fish diseases.

[0004] However, in the complex environment of a recirculating aquaculture system, due to factors such as large changes in target scale, fish occlusion, feature blur and background interference, the existing technical solutions perform poorly in terms of measurement accuracy. Summary of the invention

[0005] The present invention provides a method and device for identifying the disease degree of a fish school, so as to solve the problem that the existing technical solutions have poor performance in measurement accuracy.

[0006] The present invention provides a method for identifying the disease degree of a school of fish, comprising: collecting school of fish image data, performing data enhancement processing on the school of fish image data, and constructing a fish disease recognition data set based on the processed school of fish image data; training a fish disease recognition model through the fish disease recognition data set, the fish disease recognition model is used to identify the disease degree level of the school of fish; wherein the fish disease recognition model comprises a depthwise separable convolution module, a compression and excitation SE attention mechanism module and a dilated convolution DCLS module with a learnable spacing, the depthwise separable convolution module is used to capture the characteristics of the fish body lesion area, the SE attention mechanism module is used to enhance the characteristics of the fish body lesion area, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

[0007] According to a method for identifying the disease degree of fish schools provided by the present invention, the depthwise separable convolution module includes depthwise convolution and pointwise convolution; the depthwise convolution is used to perform an independent convolution operation on each input channel; and the pointwise convolution is used to fuse the features of all channels after depthwise convolution.

[0008] According to a method for identifying the disease degree of a fish school provided by the present invention, the output of the depthwise separable convolution module It is expressed as: ; in, is the output of the depthwise convolution, is the activation function, is the weight of the point-wise convolution.

[0009] According to a method for identifying the disease degree of a fish school provided by the present invention, the SE attention mechanism module is specifically used to: obtain a global feature representation of each channel through a global information compression operation, and adjust the weight of each channel according to the global feature representation.

[0010] According to a method for identifying the disease degree of a fish school provided by the present invention, the DCLS module includes a learnable expansion factor, and the learnable expansion factor is used to dynamically adjust the sampling interval.

[0011] The present invention also provides a device for identifying the disease degree of fish schools, comprising the following modules: a processing module and a training module; the processing module is used to collect fish school image data, perform data enhancement processing on the fish school image data, and construct a fish disease recognition data set based on the processed fish school image data; the training module is used to train a fish disease recognition model through the fish disease recognition data set, and the fish disease recognition model is used to identify the disease degree level of fish schools; wherein the fish disease recognition model includes a deep separable convolution module, a compression and excitation SE attention mechanism module and a void convolution DCLS module with a learnable spacing, the deep separable convolution module is used to capture the characteristics of the fish body lesion area, the SE attention mechanism module is used to enhance the characteristics of the fish body lesion area, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

[0012] According to a device for identifying the disease degree of fish schools provided by the present invention, the depthwise separable convolution module includes depthwise convolution and pointwise convolution; the depthwise convolution is used to perform an independent convolution operation on each input channel; and the pointwise convolution is used to fuse the features of all channels after depthwise convolution.

[0013] According to the device for identifying the disease degree of fish swarm provided by the present invention, the output of the depthwise separable convolution module It is expressed as: ; in, is the output of the depthwise convolution, is the activation function, is the weight of the point-wise convolution.

[0014] According to a device for identifying the disease degree of a fish school provided by the present invention, the SE attention mechanism module is specifically used to: obtain a global feature representation of each channel through a global information compression operation, and adjust the weight of each channel according to the global feature representation.

[0015] According to a device for identifying the disease degree of a school of fish provided by the present invention, the DCLS module includes a learnable expansion factor, and the learnable expansion factor is used to dynamically adjust the sampling interval.

[0016] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned methods for identifying the disease degree of a fish school is implemented.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for identifying the disease degree of a fish school as described in any one of the above is implemented.

[0018] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for identifying the disease degree of a fish group as described above is implemented.

[0019] The method and device for identifying the disease degree of fish swarms provided by the present invention can obtain the disease degree level output of fish swarms through a fish disease identification model. The fish disease identification model includes a deep separable convolution module, an SE attention mechanism module and a DCLS module. On the one hand, compared with traditional convolution, the deep separable convolution module can greatly reduce the amount of calculation, and efficiently extract image features at the same time, laying a solid foundation for subsequent analysis and processing; on the other hand, in the feature extraction process, the SE attention mechanism module can adaptively recalibrate the channel feature response through unique "squeezing" and "excitation" operations, so that the model focuses on those channel information that play a key role in fish disease identification, greatly improving the effectiveness of feature utilization; on the other hand, the DCLS module can dynamically adjust the sampling interval according to the characteristics of the input data during operation, cleverly introduce dilated convolution, and then flexibly expand the receptive field and accurately extract features of different scales. Through the complementary nature of the three, the recognition efficiency of the disease degree of fish swarms in the recirculating aquaculture system can be comprehensively improved, and the real-time and accurate identification of fish diseases can be successfully realized. In this way, it can provide an accurate data basis for farmers to formulate scientific and reasonable breeding strategies, and effectively promote the steady development of the recirculating aquaculture industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 It is a schematic diagram of the flow chart of the method for identifying the disease degree of fish swarms provided by the present invention; Figure 2 It is a schematic diagram of the structure of the device for identifying the disease degree of fish swarms provided by the present invention; Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0023] It should be noted that, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0024] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0025] In order to facilitate a clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and order of execution.

[0026] The embodiments of the present application describe some exemplary embodiments for the purpose of explanation. It should be understood that the present application can be implemented in other ways that are not specifically shown in the drawings.

[0027] like Figure 1 As shown, the embodiment of the present application provides a method for identifying the disease degree of a school of fish, and the method for identifying the disease degree of a school of fish can be applied to a device for identifying the disease degree of a school of fish. The method for identifying the disease degree of a school of fish can include S101-S102: S101. A device for identifying the disease degree of a school of fish collects image data of a school of fish, performs data enhancement processing on the image data of the school of fish, and constructs a fish disease identification data set based on the processed image data of the school of fish.

[0028] Specifically, the device for identifying the degree of disease of a school of fish can collect image data of a school of fish through a camera device, and then perform image conversion on the captured image data of the school of fish, that is, extract key frames from the video data and convert them into separate picture files. After that, the degree of disease of the fish in each picture is graded, and data enhancement processing such as random brightness, random noise, random scaling and horizontal flipping is performed. For example, the degree of disease of the fish can be divided into different levels such as disease-free, mild, moderate, and severe. The basis for the annotation can be the size of the lesion area on the surface of the fish body, color changes, the degree of abnormal behavior of the fish, etc. Finally, a fish disease recognition data set is constructed based on the annotated images.

[0029] It should be noted that by randomly changing the brightness value of the image, the model can adapt to fish school images under different lighting conditions. By adding random noise to the image, the interference factors that may occur in the actual image acquisition process, such as the noise of the camera sensor, impurities in the water body, etc., can be simulated. The model is exposed to noisy images during the training process, and can better learn the ability to remove noise interference and accurately identify fish disease characteristics, so that it has a certain tolerance for noise in practical applications and reduces the impact of noise on the recognition results. By flipping the image horizontally, the diversity of the image can be increased. For some symmetrical fish disease features, the horizontally flipped image can provide more training samples, so that the model can better learn these features. It also helps the model learn the performance of fish disease features in different directions and improve the generalization ability of the model.

[0030] Optionally, the device for identifying the disease degree of fish can reallocate the fish disease identification data set into a training set, a test set and a validation set in a certain proportion. Among them, the training set is used to train the model. The model learns a large number of images with fish disease degree grades in the training set and adjusts its own parameters to extract features and patterns that can accurately identify the degree of fish disease. The test set is used to test and evaluate the performance of the model after the model training is completed. The performance of the model on the test set can reflect the generalization ability and accuracy of the model in practical applications. By comparing with the training set, it can be found whether the model has problems such as overfitting. The validation set is used to verify the performance of the model at different training stages and adjust the hyperparameters of the model during the model training process. Through the evaluation results on the validation set, the training strategy and parameters of the model can be adjusted in time to improve the performance and generalization ability of the model.

[0031] S102: The fish disease degree identification device trains a fish disease identification model using the fish disease identification data set.

[0032] Among them, the above-mentioned fish disease recognition model is used to identify the disease degree of fish schools. The fish disease recognition model includes a deep separable convolution module, a Squeeze-and-Excitation (SE) attention mechanism module and a Dilated Convolution with Learnable Spacings (DCLS) module. The deep separable convolution module is used to capture the characteristics of the diseased area of ​​the fish body, the SE attention mechanism module is used to enhance the characteristics of the diseased area of ​​the fish body, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

[0033] Optionally, the fish disease degree identification device can use the fish disease identification data set as the input of the fish disease identification model, set the network initial parameters and use the AdamW optimizer to train the fish disease identification model to generate a trained fish disease identification model. Finally, the trained fish disease identification model is used to identify the fish disease degree, and finally the fish disease degree level is obtained. The fish disease degree level can include three categories: "normal", "mildly sick", and "severely sick".

[0034] Optionally, the depthwise separable convolution module includes depthwise convolution and pointwise convolution; the depthwise convolution is used to perform independent convolution operations on each input channel; the pointwise convolution is used to fuse the features of all channels after depthwise convolution. The output of the depthwise separable convolution module is expressed as: ;in, is the output of the depthwise convolution, is the activation function, is the weight of the point-wise convolution.

[0035] Specifically, the device for identifying the disease degree of fish can further divide the image in S101 into patches (local areas), and then use the deep separable convolution module to extract features from these patches. The deep separable convolution module can flexibly extract features from the data by combining independent deep convolution and point-by-point convolution. First, the deep convolution uses a convolution kernel to process each input channel separately, and only performs convolution operations on a single channel, so that the features of each channel can be mined separately; then the point-by-point convolution uses a 1×1 convolution kernel to linearly combine the information of each channel output by the deep convolution and integrate them, thereby achieving effective feature extraction.

[0036] When the input feature map is X When , the depth convolution applies a convolution kernel to each input channel, and the size of each convolution kernel is H × W , the convolution kernel only performs convolution operation on a single channel, and its output can be expressed as: (1); in, Indicates channel c The output, Indicates input channel c ,and Indicates that it is applied to the channel c The convolution kernel.

[0037] The point-by-point convolution part can use a 1×1 convolution kernel to linearly combine the output of the depthwise convolution to integrate the information of each channel. It is expressed as: (2); in, Represents the weights of the point-wise convolution.

[0038] Output of the final depthwise separable convolutional module It can be expressed as: (3); in, is the output of the depthwise convolution, is the activation function.

[0039] By replacing the original deep convolution with deep separable convolution, the model can perform more complex feature learning on each channel, further improving the feature extraction effect, enabling the model to better capture various features such as the health status of different fish in complex fish school images. It can also capture multi-scale features by combining different convolution kernel sizes, and more comprehensively analyze the manifestations of different lesions.

[0040] It should be noted that with the help of deep separable convolution combined with activation function, the model can better capture complex feature relationships, thereby improving the overall model performance. Moreover, when faced with the problem of fish classification with insufficient samples or unbalanced categories, due to its relatively few parameters and reduced model complexity, the risk of overfitting is reduced and it has better generalization ability. However, after replacing the deep convolution with the deep separable convolution, the model complexity will increase significantly, which may have an adverse effect on the speed of training and reasoning. Therefore, this application can reduce the network depth from 32 to 12, and cooperate with the deep separable convolution module. In this way, the number of parameters of the model can be reduced, thereby improving the running speed and balancing the relationship between model performance and running efficiency.

[0041] Optionally, the SE attention mechanism module is specifically used to: obtain a global feature representation of each channel through a global information compression operation, and adjust the weight of each channel according to the global feature representation.

[0042] Specifically, the specific operations of the SE attention mechanism module include Squeeze operation and Excitation operation. The Squeeze operation can process the channel dimension by compressing the spatial dimension information of each channel and converting the spatial information into a single global description to capture the global feature representation of each channel. The Excitation operation can selectively recalibrate the weights of each channel by modeling the relationship between channels, thereby determining the relative importance of each channel.

[0043] When processing fish school image data, the feature map Represents the height of the image H ,width W and number of channels C The Squeeze operation compresses the spatial information of each channel into a single global description through global average pooling, which is specifically expressed as: ; in, It is c The global average response of the channels, Representation feature map Middle c Channels at position For the task of classifying the disease level of fish, It reflects the average importance of the channel to a specific feature (such as color, texture, or lesion area) over the entire image. For example, some channels may be specifically used to capture the health status of fish (such as color changes or fin damage), and these features have a significant spatial distribution in the overall image.

[0044] The Excitation operation models the relationship between channels through a two-layer fully connected neural network to generate attention weights for each channel. C Compress to C / r ,in, r is the compression ratio. The second fully connected layer reduces the number of channels from C / r Restore to C , and the weights are limited to [0, 1] through the Sigmoid activation function. The generated weights represent the relative importance of each channel. Finally, the output feature map can be expressed as: ; in, Represents the first feature map after recalibration c channels, Represents the first c The attention weight of each channel, Represents the original feature map c This means that for features that are considered important in the fish classification task (such as obvious signs of lesions), their corresponding The value will be larger, causing these features to be enhanced in the subsequent network; while for unimportant features, Smaller values ​​suppress the influence of these features.

[0045] The SE attention mechanism module can improve the model's sensitivity to the channel-to-channel relationship of features. Especially in tasks such as fish disease level classification, it can enable the model to focus on the features of different channels more effectively, ultimately improving the classification accuracy and improving the overall performance of the model.

[0046] Optionally, the DCLS module includes a learnable expansion factor, and the learnable expansion factor is used to dynamically adjust the sampling interval.

[0047] Specifically, in order to enhance the model's ability to extract features of different scales, the DCLS module of the present application extracts features through dilated convolution with a learnable interval. Dilated convolution is a special convolution operation. It expands the receptive field by introducing a dilation factor between the elements of the convolution kernel, so that the network can capture a wider range of information at a shallower layer, thereby reducing the need for a large receptive field in subsequent layers. This results in the use of smaller convolution kernels or fewer channels in subsequent layers, further reducing the number of parameters and accelerating the reasoning process. The dilation factor of the dilated convolution is a learnable parameter. During the model training process, the model will dynamically adjust the dilation factor according to the characteristics of the input data to achieve the purpose of optimizing the model performance. The receptive field refers to the size of the input area that a neuron in a convolutional neural network can perceive. For example, in image recognition, the larger the receptive field, the wider the image range that the neuron can see. When calculating the loss function, the dilation rate parameter will also participate in the gradient calculation. At the beginning of training, the dilation factor of the convolution layer is 1, that is, ordinary convolution. As the training proceeds, the dilation factor will be dynamically adjusted, and the receptive field of the convolution kernel will gradually expand to meet the extraction needs of features of different scales.

[0048] During the forward propagation process, the input is first processed by a depthwise separable convolution, which is mainly used to extract the spatial features of each channel. Then it is processed by a dilated convolution, which dynamically adjusts the receptive field through a learnable dilation factor, thereby effectively extracting features of different scales.

[0049] The process of dilated convolution can be expressed as: ; in, is the pixel value on the output feature map, that is, the pixel at position after the dilated convolution operation The results obtained, X is the input feature map, W is the convolution kernel, k is the size of the convolution kernel, Represents the convolution kernel W The position index in the horizontal direction (row direction), its value range is from 0 to k-1 , Represents the convolution kernel W The position index in the vertical direction (column direction) also ranges from 0 to k-1 , d is the expansion factor. d=1 When , the dilated convolution degenerates into a normal convolution; when d>1 When , the convolution kernel will expand and the distance between its elements will increase, thus expanding the receptive field.

[0050] In the embodiment of the present application, the fish disease recognition model can be used to obtain the fish disease degree output, which includes a deep separable convolution module, an SE attention mechanism module and a DCLS module. On the one hand, compared with traditional convolution, the deep separable convolution module can greatly reduce the amount of calculation, and efficiently extract image features, laying a solid foundation for subsequent analysis and processing; on the other hand, in the feature extraction process, the SE attention mechanism module can adaptively recalibrate the channel feature response through unique "squeezing" and "excitation" operations, so that the model focuses on those channel information that play a key role in fish disease recognition, greatly improving the effectiveness of feature utilization; on the other hand, the DCLS module can dynamically adjust the sampling interval according to the characteristics of the input data during operation, cleverly introduce dilated convolution, and then flexibly expand the receptive field and accurately extract features of different scales. Through the complementary nature of the three, the recognition efficiency of the disease degree of fish in the recirculating aquaculture system can be comprehensively improved, and the real-time and accurate recognition of fish diseases can be successfully realized. In this way, it can provide an accurate data basis for farmers to formulate scientific and reasonable breeding strategies, and effectively promote the steady development of the recirculating aquaculture industry.

[0051] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the method. In order to realize the above functions, it includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiment disclosed herein, the embodiment of the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0052] It should be noted that the device in the embodiment of the present application includes a virtual device and a physical device. The virtual device can be a device for identifying the disease degree of fish, and the physical device can include an electronic device, a computer storage medium, and a computer program product.

[0053] The method for identifying the disease degree of a school of fish provided in the embodiment of the present application can be executed by a device for identifying the disease degree of a school of fish, or a control module for identifying the disease degree of a school of fish in the device for identifying the disease degree of a school of fish. In the embodiment of the present application, the device for identifying the disease degree of a school of fish provided in the embodiment of the present application is described by taking the device for identifying the disease degree of a school of fish as an example to execute the method for identifying the disease degree of a school of fish.

[0054] It should be noted that the embodiment of the present application can divide the functional modules of the fish disease degree identification device according to the above method example. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules. Optionally, the division of modules in the embodiment of the present application is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0055] like Figure 2 As shown, an embodiment of the present application provides a device 200 for identifying the disease degree of a school of fish. The device 200 for identifying the disease degree of a school of fish includes: a processing module 201 and a training module 202; the processing module 201 is used to collect image data of a school of fish, perform data enhancement processing on the image data of the school of fish, and construct a fish disease recognition data set based on the processed image data of the school of fish; the training module 202 is used to train a fish disease recognition model through the fish disease recognition data set, and the fish disease recognition model is used to identify the disease degree level of a school of fish; wherein the fish disease recognition model includes a depthwise separable convolution module, a compression and excitation SE attention mechanism module, and a dilated convolution DCLS module with a learnable spacing, the depthwise separable convolution module is used to capture the characteristics of the fish body lesion area, the SE attention mechanism module is used to enhance the characteristics of the fish body lesion area, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

[0056] Optionally, the depthwise separable convolution module includes depthwise convolution and pointwise convolution; the depthwise convolution is used to perform an independent convolution operation on each input channel; and the pointwise convolution is used to fuse features of all channels after depthwise convolution.

[0057] Optionally, the output of the depthwise separable convolution module It is expressed as: ; in, is the output of the depthwise convolution, is the activation function, is the weight of the point-wise convolution.

[0058] Optionally, the SE attention mechanism module is specifically used to: obtain a global feature representation of each channel through a global information compression operation, and adjust the weight of each channel according to the global feature representation.

[0059] Optionally, the DCLS module includes a learnable expansion factor, and the learnable expansion factor is used to dynamically adjust the sampling interval.

[0060] In the embodiment of the present application, the fish disease recognition model can be used to obtain the fish disease degree output, which includes a deep separable convolution module, an SE attention mechanism module and a DCLS module. On the one hand, compared with traditional convolution, the deep separable convolution module can greatly reduce the amount of calculation, and efficiently extract image features, laying a solid foundation for subsequent analysis and processing; on the other hand, in the feature extraction process, the SE attention mechanism module can adaptively recalibrate the channel feature response through unique "squeezing" and "excitation" operations, so that the model focuses on those channel information that play a key role in fish disease recognition, greatly improving the effectiveness of feature utilization; on the other hand, the DCLS module can dynamically adjust the sampling interval according to the characteristics of the input data during operation, cleverly introduce dilated convolution, and then flexibly expand the receptive field and accurately extract features of different scales. Through the complementary nature of the three, the recognition efficiency of the disease degree of fish in the recirculating aquaculture system can be comprehensively improved, and the real-time and accurate recognition of fish diseases can be successfully realized. In this way, it can provide an accurate data basis for farmers to formulate scientific and reasonable breeding strategies, and effectively promote the steady development of the recirculating aquaculture industry.

[0061] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320 and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute the method for identifying the disease degree of fish school, the method comprising: collecting fish school image data, performing data enhancement processing on the fish school image data, and constructing a fish disease recognition data set based on the processed fish school image data; training a fish disease recognition model through the fish disease recognition data set, the fish disease recognition model is used to identify the disease degree of fish school; wherein the fish disease recognition model comprises a depth separable convolution module, a compression and excitation SE attention mechanism module and a void convolution DCLS module with a learnable spacing, the depth separable convolution module is used to capture the characteristics of the fish body lesion area, the SE attention mechanism module is used to enhance the characteristics of the fish body lesion area, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

[0062] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0063] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the fish disease degree identification method provided by the above-mentioned methods, which includes: collecting fish image data, performing data enhancement processing on the fish image data, and constructing a fish disease identification data set based on the processed fish image data; training a fish disease identification model through the fish disease identification data set, and the fish disease identification model is used to identify the disease degree of the fish; wherein the fish disease identification model includes a deep separable convolution module, a compression and excitation SE attention mechanism module, and a void convolution DCLS module with a learnable spacing, the deep separable convolution module is used to capture the characteristics of the fish body disease area, the SE attention mechanism module is used to enhance the characteristics of the fish body disease area, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

[0064] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the method for identifying the disease degree of fish schools provided by the above-mentioned methods, the method comprising: collecting fish school image data, performing data enhancement processing on the fish school image data, and constructing a fish disease identification data set based on the processed fish school image data; training a fish disease identification model through the fish disease identification data set, the fish disease identification model being used to identify the disease degree level of fish schools; wherein the fish disease identification model comprises a depthwise separable convolution module, a compression and excitation SE attention mechanism module, and a dilated convolution DCLS module with a learnable spacing, the depthwise separable convolution module being used to capture the characteristics of the diseased area of ​​the fish body, the SE attention mechanism module being used to enhance the characteristics of the diseased area of ​​the fish body, and the DCLS module being used to dynamically adjust the sampling spacing to extract multi-scale features.

[0065] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0066] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0067] 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 method for identifying the disease degree of fish swarms, characterized in that: include: Collecting fish school image data, performing data enhancement processing on the fish school image data, and constructing a fish disease recognition data set based on the processed fish school image data; A fish disease recognition model is trained by using the fish disease recognition data set, wherein the fish disease recognition model is used to identify the disease degree of fish swarms; Among them, the fish disease recognition model includes a deep separable convolution module, a compression and excitation SE attention mechanism module and a void convolution DCLS module with learnable spacing. The deep separable convolution module is used to capture the characteristics of the fish disease area, the SE attention mechanism module is used to enhance the characteristics of the fish disease area, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

2. The method for identifying the disease degree of fish swarms according to claim 1, characterized in that: The depthwise separable convolution module includes depthwise convolution and pointwise convolution; the depthwise convolution is used to perform an independent convolution operation on each input channel; the pointwise convolution is used to fuse the features of all channels after depthwise convolution.

3. The method for identifying the disease degree of fish swarms according to claim 2, characterized in that: The output of the depthwise separable convolution module It is expressed as: ； in, is the output of the depthwise convolution, is the activation function, is the weight of the point-wise convolution.

4. The method for identifying the disease degree of fish swarms according to claim 1, characterized in that: The SE attention mechanism module is specifically used to obtain the global feature representation of each channel through a global information compression operation, and adjust the weight of each channel according to the global feature representation.

5. The method for identifying the disease degree of fish swarm according to claim 1, characterized in that: The DCLS module includes a learnable dilation factor, which is used to dynamically adjust the sampling spacing.

6. A device for identifying the disease degree of fish, characterized in that: include: Processing module and training module; The processing module is used to collect fish school image data, perform data enhancement processing on the fish school image data, and construct a fish disease recognition data set based on the processed fish school image data; The training module is used to train a fish disease recognition model using the fish disease recognition data set, and the fish disease recognition model is used to identify the disease degree of fish swarms; Among them, the fish disease recognition model includes a deep separable convolution module, a compression and excitation SE attention mechanism module and a void convolution DCLS module with learnable spacing. The deep separable convolution module is used to capture the characteristics of the fish disease area, the SE attention mechanism module is used to enhance the characteristics of the fish disease area, and the DCLS module is used to dynamically adjust the sampling spacing to extract multi-scale features.

7. The device for identifying the disease degree of fish schools according to claim 6, characterized in that: The depthwise separable convolution module includes depthwise convolution and pointwise convolution; the depthwise convolution is used to perform an independent convolution operation on each input channel; the pointwise convolution is used to fuse the features of all channels after depthwise convolution.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for identifying the disease degree of a fish group as described in any one of claims 1 to 5 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for identifying the disease degree of a fish population as claimed in any one of claims 1 to 5 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for identifying the disease degree of a fish population as claimed in any one of claims 1 to 5 is implemented.

Images