A crop disease identification method, system, device and storage medium

Abstract

The application discloses a crop disease identification method, system and device and a storage medium, wherein the method comprises the following steps: obtaining a crop disease data set, preprocessing and data amplification are performed on the data set to obtain a preprocessed data set; a disease identification model is constructed, the disease identification model is trained by using the obtained data set, and the trained model is used for crop disease identification; wherein the disease identification model takes SCResNet50 as a backbone network, and uses a recursive feature pyramid to extract features on each pyramid level to capture object information of different sizes and resolutions. The application uses the recursive feature pyramid to extract features on each pyramid level to capture object information of different sizes and resolutions, combines low-layer features with high resolution and high-layer features rich in semantics, and significantly improves the semantic richness and accuracy of feature representation. The application can be widely applied to the technical field of crop disease image identification.

Description

Technical Field

[0001] This invention relates to the field of crop disease image recognition technology, and in particular to a crop disease recognition method, system, device and storage medium. Background Technology

[0002] Crop diseases and pests are among the most significant disasters restricting agricultural production, directly impacting healthy economic development and hindering farmers' income growth. Currently, internationally, pests and diseases reduce potential crop yields by an average of 40%, while in many developing countries, farmers suffer yield losses as high as 100%. Losses caused by crop diseases and pests account for approximately one-tenth of the world's total food production. Traditional methods of pest and disease detection rely primarily on field surveys by local agricultural technicians, making judgments based on crop pathological characteristics. However, this approach suffers from several drawbacks: firstly, the sheer variety of pests and diseases makes it difficult to disseminate relevant knowledge and there is a lack of experts in the field; secondly, the scarcity of professional personnel in rural areas hinders timely diagnosis of pest and disease conditions.

[0003] Early research on automatic identification of crop diseases and pests primarily relied on digital image processing and traditional machine learning techniques. This approach has several limitations: First, while it performs well in simple backgrounds, its effectiveness is less than satisfactory in complex real-world farmland environments. Second, the manually constructed models require continuous parameter adjustments based on performance, increasing both the workload and the time required for model execution. Finally, poor scalability is a significant drawback, making it difficult to adapt to different geographical features and conditions for diagnosing crop diseases and pests.

[0004] In recent years, with the significant improvement of hardware computing power, the artificial intelligence boom has emerged, making deep learning a popular method among researchers. Deep learning has significant advantages in the identification of crop diseases and pests, mainly in the following aspects: (1) Images and classification results can be input and output in an end-to-end manner. Images can be directly used as network input without preprocessing, and the network results can directly obtain the classification answer; (2) The model parameters are obtained by computer calculation, without the need for continuous optimization of model parameters by humans, saving manpower; (3) It has strong generalization ability and low dependence on background environment. The input training data can be adjusted according to different task needs, thereby expanding and optimizing the model's recognition ability for different recognition tasks. For image recognition tasks, deep learning methods represented by Convolutional Neural Networks (CNN) have performed well in many problems. CNN can use a simple network to automatically calculate and fit complex functions. At the same time, multiple hidden layers can automatically extract high-level features of images very well and perform recognition and classification.

[0005] In conclusion, it is necessary to propose a new crop disease identification network that fully utilizes features across multiple scales and integrates contextual information from different scales to improve the identification accuracy of the network model. Summary of the Invention

[0006] In order to at least partially solve one of the technical problems existing in the prior art, the present invention aims to provide a method, system, device and storage medium for identifying crop diseases based on recursive feature pyramid and self-calibrating convolution mechanism.

[0007] The technical solution adopted in this invention is:

[0008] A method for identifying crop diseases includes the following steps:

[0009] Obtain a crop disease dataset, preprocess and augment the dataset to obtain a preprocessed dataset;

[0010] A disease identification model is constructed, and the obtained dataset is used to train the disease identification model. The trained model is then used for crop disease identification.

[0011] The disease identification model uses SCResNet50 as the backbone network and uses a recursive feature pyramid to extract features at each pyramid level in order to capture information about objects of different sizes and resolutions.

[0012] Furthermore, the SCResNet50 network replaces the 3×3 convolutional modules in the ResNet50 network with self-calibrating convolutional modules.

[0013] Furthermore, the self-calibrating convolution module can simultaneously consider both the original scale space and the downsampled latent space. In the original scale space, the resolution of the feature maps remains consistent with the input data. However, in the downsampled latent space, although the resolution of the feature maps is reduced, their receptive field is correspondingly increased. This allows the model to capture a wider range of contextual information, which effectively influences and optimizes the feature transformation process in the original feature space when used as calibration parameters. The specific calculation process expression is shown below:

[0014] T1 = AvgPool r (X1)

[0015] X1′=Up(T1*K2)

[0016] Y1′=X1*K3·σ(X1+X1′)

[0017] Y1=Y1′*K4

[0018] Y2=X2*K1

[0019] Y = Concat(Y1, Y2)

[0020] In the formula, X1 and X2 represent dividing the input feature map into two groups evenly according to the channel dimension; * represents convolution, and · represents dot product; AvgPool r This indicates that the features are downsampled by a factor of r using average pooling; σ is the non-linear activation function Sigmoid; Concat indicates feature concatenation; Up indicates feature upsampling; K1, K2, K3, and K4 are the convolution kernels used.

[0021] Furthermore, the recursive feature pyramid (RFP) recursively constructs an image pyramid and extracts features at each pyramid level to capture object information of different sizes and resolutions. The RFP then merges the features output from each stage of the backbone network into the bottom-up backbone layer, forming a target detector backbone capable of multiple image observations. By combining high-resolution low-level features with semantically rich high-level features, the semantic richness and accuracy of the feature representation are significantly improved. The specific calculation process is expressed as follows:

[0022]

[0023] k i =2(i+1)+1

[0024]

[0025]

[0026]

[0027] In the formula, t = 0, 1, 2, ..., T represents the number of iterations. Let i represent the input feature map after the t-th iteration, where i = 0. The original input image; k i G represents the size of the i-th convolutional kernel. i For a convolution kernel size of k i The number of convolutional groups; Indicates a size of k i The number of groups is G i The convolution kernel; This indicates that after the t-th iteration, the kernel size is k. i The feature map after computation; Concat represents the feature concatenation operation; F∈R C×H×W This represents the output feature map after recursive feature pyramid convolution.

[0028] Furthermore, the preprocessing and augmentation of the dataset includes:

[0029] Randomly crop, randomly flip horizontally, randomly rotate, or randomly erase the images in the dataset;

[0030] The images in the dataset are normalized, and the size of the input images is uniformly adjusted to 224×224 during training.

[0031] Furthermore, training the disease identification model using the obtained dataset includes:

[0032] The network parameters were updated using the Adam optimization algorithm, with the initial learning rate set to 0.001, β1 = 0.9, β2 = 0.999, and eps = 1 × 10⁻⁶. -8 We employ the cross-entropy loss function and use a cosine annealing learning rate update strategy to train for 200 epochs. After each training iteration, we test the model on the validation set and save the model with the highest accuracy on the validation set.

[0033] Another technical solution adopted in this invention is:

[0034] A crop disease identification system includes:

[0035] The data acquisition module is used to acquire crop disease datasets, preprocess and augment the datasets to obtain preprocessed datasets.

[0036] The model training module is used to build a disease identification model. The obtained dataset is used to train the disease identification model, and the trained model is used for crop disease identification.

[0037] The disease identification model uses SCResNet50 as the backbone network and uses a recursive feature pyramid to extract features at each pyramid level in order to capture information about objects of different sizes and resolutions.

[0038] Another technical solution adopted in this invention is:

[0039] A crop disease identification device, comprising:

[0040] At least one processor;

[0041] At least one memory for storing at least one program;

[0042] When the at least one program is executed by the at least one processor, the at least one processor implements the method described above.

[0043] Another technical solution adopted in this invention is:

[0044] A computer-readable storage medium storing a processor-executable program, which, when executed by a processor, performs the method described above.

[0045] The beneficial effects of this invention are as follows: By replacing the 3×3 convolutional module in the ResNet50 network with a self-calibrating convolutional module, the receptive field is increased, enhancing the backbone network's ability to capture a wider range of contexts. Furthermore, by capturing information about objects of different sizes and resolutions through a recursive feature pyramid, and by combining high-resolution low-level features with semantically rich high-level features, the semantic richness and accuracy of feature representation are significantly improved, thereby enhancing the model's recognition accuracy.

[0046] Existing technical solutions mainly enhance the focus on the main diseased areas of plants. Compared with existing technical solutions, the core of the algorithm used in this invention is to enhance the model's ability to connect high-level and low-level features. Its main advantage is that it improves the accuracy of analyzing and judging the severity of crop diseases based on the distribution and specific shape of diseased crop leaves. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of the present invention or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram of the self-calibrating convolution used in the embodiments of the present invention;

[0049] Figure 2 This is a schematic diagram of the SCResNet50 model structure used in this embodiment of the invention;

[0050] Figure 3 This is a schematic diagram of the recursive feature pyramid in an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of the structure of adding the recursive feature pyramid to the SCResNet50 network in an embodiment of the present invention. Detailed Implementation

[0052] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0053] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0054] In the description of this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. If "first" or "second" is used, it is only for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features. Furthermore, "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0055] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0056] Terminology Explanation:

[0057] RFP: Recursive Feature Pyramid.

[0058] like Figure 1 As shown, this embodiment provides a method for identifying crop diseases based on recursive feature pyramid convolution and a self-calibrating attention mechanism, including the following steps:

[0059] Step 1: Obtain the crop disease dataset, and perform preprocessing and data augmentation on the dataset to obtain the preprocessed dataset. Specifically, the dataset is preprocessed and augmented, including random cropping, random horizontal flipping, random rotation, random erasure, and normalization. Finally, during training, the input images are uniformly adjusted to 224×224.

[0060] The disease dataset in this embodiment comes from the publicly available AI Challenger 2018 dataset, which contains 61 different diseases across 10 crops. The crops include apples, cherries, corn, grapes, citrus, peaches, peppers, potatoes, strawberries, and tomatoes. The diseases include black spot, gray spot, powdery mildew, and mosaic virus, with severity levels ranging from mild to severe. After cleaning and removing two categories with insufficient data and duplicate images, the dataset contains 36,020 images. These images are randomly divided into a training set (25,258 images, 70% of the total), a validation set (7,182 images, 20% of the total), and a test set (3,580 images, 10% of the total).

[0061] Step 2, see Figure 1 and Figure 2 The SCResNet50 network is constructed by replacing the 3×3 convolutional modules in the ResNet50 network with self-calibrating convolutions. The self-calibrating convolutional modules can simultaneously consider both the original scale space and the downsampled latent space. In the original scale space, the resolution of the feature maps remains consistent with the input data. However, in the downsampled latent space, although the resolution of the feature maps is reduced, their receptive field is correspondingly increased. This allows the model to capture a wider range of contextual information, which effectively influences and optimizes the feature transformation process in the original feature space when used as calibration parameters. The specific calculation process is expressed as follows:

[0062] T1 = AvgPool r (X1)

[0063] X1′=Up(T1*K2)

[0064] Y1′=X1*K3·σ(X1+X1′)

[0065] Y1=Y1′*K4

[0066] Y2=X2*K1

[0067] Y = Concat(Y1, Y2)

[0068] In the formula, X1 and X2 represent dividing the input feature map X into two groups evenly according to the channel dimension, denoted as X1 and X2 respectively; * represents convolution, and · represents dot product; AvgPool rThis indicates that the features are downsampled by a factor of r using average pooling; σ is the non-linear activation function Sigmoid; Concat indicates feature concatenation; Up indicates feature upsampling; K1, K2, K3, and K4 are the convolution kernels used.

[0069] Step 3, see Figure 3 and Figure 4 Based on the backbone network, a recursive feature pyramid (RFP) is used to extract features at each pyramid level to capture object information of different sizes and resolutions. The features output from each stage of the backbone network are then concatenated and merged into the bottom-up backbone layer, forming a target detector backbone capable of multiple image observations. By combining high-resolution low-level features with semantically rich high-level features, the semantic richness and accuracy of the feature representation are significantly improved. The specific calculation process is expressed as follows:

[0070]

[0071] k i =2(i+1)+1

[0072]

[0073]

[0074]

[0075] In the formula, t = 0, 1, 2, ..., T represents the number of iterations. Let i represent the input feature map after the t-th iteration, where i = 0. The original input image; k i G represents the size of the i-th convolutional kernel. i For a convolution kernel size of k i The number of convolutional groups; Indicates a size of k i The number of groups is G i convolution kernel; f i t This indicates that after the t-th iteration, the kernel size is k. i The processed feature map; Concat represents the feature concatenation operation; F represents the output feature map after recursive feature pyramid convolution.

[0076] Step 4: Train the model processed in Steps 2-3 using the dataset obtained in Step 1, and save the trained model. The specific training method is as follows: Update the network parameters using the Adam optimization algorithm, initializing the learning rate to 0.001, β1 = 0.9, β2 = 0.999, and eps = 1 × 10⁻⁶. -8We employ the cross-entropy loss function and use a cosine annealing learning rate update strategy to train for 200 epochs. After each training iteration, we test the model on the validation set and save the model with the highest accuracy on the validation set.

[0077] This embodiment also provides a crop disease identification system, including:

[0078] Obtain a crop disease dataset, preprocess and augment the dataset to obtain a preprocessed dataset;

[0079] A disease identification model is constructed, and the obtained dataset is used to train the disease identification model. The trained model is then used for crop disease identification.

[0080] The disease identification model uses SCResNet50 as its backbone network, replacing the 3×3 convolutional modules in the ResNet50 network with self-calibrating convolutions to form the SCResNet50 network. The self-calibrating convolutional modules can simultaneously consider both the original scale space and the downsampled latent space. In the original scale space, the resolution of the feature maps remains consistent with the input data. However, in the downsampled latent space, although the resolution of the feature maps is reduced, their receptive field is correspondingly increased. This allows the model to capture a wider range of contextual information, which effectively influences and optimizes the feature transformation process in the original feature space when used as calibration parameters. A recursive feature pyramid is used to extract features at each pyramid level to capture information about objects of different sizes and resolutions. By combining high-resolution low-level features with semantically rich high-level features, the semantic richness and accuracy of the feature representation are significantly improved.

[0081] This embodiment of the crop disease identification system can execute the crop disease identification method provided in the method embodiment of the present invention, and can execute any combination of the implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0082] This embodiment also provides a crop disease identification device, including:

[0083] At least one processor;

[0084] At least one memory for storing at least one program;

[0085] When the at least one program is executed by the at least one processor, the at least one processor implements the method described above.

[0086] This embodiment of a crop disease identification device can execute a crop disease identification method provided in the method embodiment of the present invention, and can execute any combination of the implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0087] This embodiment also provides a storage medium storing instructions or programs that can execute the crop disease identification method provided in the method embodiment of the present invention. When the instructions or programs are run, any combination of implementation steps of the method embodiment can be executed, and the method has the corresponding functions and beneficial effects.

[0088] This embodiment also provides a storage medium storing instructions or programs that can execute the crop disease identification method provided in the method embodiment of the present invention. When the instructions or programs are run, any combination of implementation steps of the method embodiment can be executed, and the method has the corresponding functions and beneficial effects.

[0089] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.

[0090] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0091] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0092] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0093] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0094] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0095] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0096] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0097] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for identifying crop diseases, characterized in that, Includes the following steps: Obtain a crop disease dataset, preprocess and augment the dataset to obtain a preprocessed dataset; A disease identification model is constructed, and the obtained dataset is used to train the disease identification model. The trained model is then used for crop disease identification. The disease identification model uses SCResNet50 as the backbone network and uses a recursive feature pyramid to extract features at each pyramid level in order to capture information about objects of different sizes and resolutions. The SCResNet50 network replaces the 3×3 convolutional modules in the ResNet50 network with self-calibrating convolutional modules. The self-calibrating convolution module can simultaneously consider both the original scale space and the downsampled latent space. In the original scale space, the resolution of the feature map remains consistent with the input data. However, in the downsampled latent space, although the resolution of the feature map is reduced, its receptive field is correspondingly increased. This allows the model to capture a wider range of contextual information. The specific calculation process is expressed as follows: T1=AvgPool r (X1) X′1=Up(T1*K2) Y′1=X1*K3·σ(X1+X′1) Y1=Y′1*K4 Y2=X2*K1 Y = Concat(Y1, Y2) In the formula, X1 and X2 represent dividing the input feature map X into two groups evenly according to the channel dimension, denoted as X1 and X2 respectively; * represents convolution, and · represents dot product; AvgPool r This indicates that the features are downsampled by a factor of r using average pooling; σ is the non-linear activation function Sigmoid; Concat represents the feature concatenation operation; Up represents the feature upsampling; K1, K2, K3, and K4 are the convolution kernels used. The recursive feature pyramid (RFP) recursively constructs an image pyramid and extracts features at each pyramid level to capture object information of different sizes and resolutions. The RFP then merges the features output from each stage of the backbone network into the bottom-up backbone layer, forming a target detector backbone capable of multiple image observations. By combining high-resolution low-level features with semantically rich high-level features, the semantic richness and accuracy of the feature representation are significantly improved. The specific calculation process is expressed as follows: k i =2(i+1)+1 In the formula, t = 0, 1, 2, ..., T represents the number of iterations. Let i represent the input feature map after the t-th iteration, where i = 0. The original input image; k i G represents the size of the i-th convolutional kernel. i For a convolution kernel size of k i The number of convolutional groups; Indicates a size of k i The number of groups is G i convolution kernel; f i t This indicates that after the t-th iteration, the kernel size is k. i The feature map after computation; Concat represents the feature concatenation operation; F represents the output feature map after recursive feature pyramid convolution.

2. The method for identifying crop diseases according to claim 1, characterized in that, The preprocessing and augmentation of the dataset includes: Randomly crop, randomly flip horizontally, randomly rotate, or randomly erase the images in the dataset; The images in the dataset are normalized, and the size of the input images is uniformly adjusted to 224×224 during training.

3. The method for identifying crop diseases according to claim 1, characterized in that, The step of training the disease identification model using the obtained dataset includes: The network parameters were updated using the Adam optimization algorithm, with the initial learning rate set to 0.001, β1 = 0.9, β2 = 0.999, and eps = 1 × 10⁻⁶. -8 We employ the cross-entropy loss function and use a cosine annealing learning rate update strategy to train for 200 epochs. After each training iteration, we test the model on the validation set and save the model with the highest accuracy on the validation set.

4. A crop disease identification system, applied to the method described in any one of claims 1-3, characterized in that, include: The data acquisition module is used to acquire crop disease datasets, preprocess and augment the datasets to obtain preprocessed datasets. The model training module is used to build a disease identification model. The obtained dataset is used to train the disease identification model, and the trained model is used for crop disease identification. The disease identification model uses SCResNet50 as the backbone network and uses a recursive feature pyramid to extract features at each pyramid level in order to capture information about objects of different sizes and resolutions.

5. A crop disease identification device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method of any one of claims 1-3.

6. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to perform the method as described in any one of claims 1-3.

Images