Deep learning-based pomegranate fruit full-growth-cycle multi-class disease identification method
A multi-category disease identification model for pomegranate fruit throughout its entire growth cycle, constructed using deep learning, solves the problems of low accuracy and poor adaptability in identifying pomegranate fruit diseases. It enables accurate identification and early warning of diseases, generates control strategies, and is applicable to field terminal equipment.
Patent Information
- Application Number
- CN202511421472.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-30
- Publication Date
- 2026-01-13
AI Technical Summary
Existing pomegranate fruit disease identification technologies suffer from low identification accuracy and poor adaptability in identifying diseases throughout the entire growth cycle and across multiple categories. They are unable to meet the needs of early warning and precise control, especially in complex environments where disease area characteristics are blurred and disease characteristics are similar and overlapping, making identification difficult.
A deep learning-based approach was used to construct a multi-category disease identification model for the entire growth cycle of pomegranate fruit. By using an improved YOLOv8 network structure, combined with a global attention mechanism and lightweight design, a disease knowledge base was built to achieve accurate disease identification and generation of prevention and control strategies.
It significantly improves the accuracy and sensitivity of disease identification, enables precise capture and proactive early warning of early diseases, generates timely prevention and control strategies, forms a decision-making closed loop of immediate detection and treatment, and is suitable for deployment in field terminal equipment.
Smart Images
Figure CN121330489A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of data processing, and in particular to a pomegranate fruit full growth cycle multi-class disease identification method based on deep learning. BACKGROUND
[0002] Pomegranate fruits are susceptible to multiple disease complex infections during the entire growth cycle, including anthracnose, bacterial wilt, alternaria fruit spot, etc. In the natural orchard environment, there are many challenges in carrying out disease image recognition. Due to the complex factors such as light changes, branch and leaf shading, fruit overlapping, and background interference in the environment, the image acquisition quality is disturbed, the disease area features are blurred, and the model recognition performance is affected. The significant differences in color, shape, texture, etc. of different diseases, as well as the changes in the representation of the same disease at different growth stages of the fruit, put higher requirements on the feature extraction and discrimination ability of the model. In addition, some diseases have certain feature similarity and spatial overlap at the image level, which further increases the difficulty of accurate identification of multi-class and cross-stage diseases.
[0003] With the rapid development of agricultural informatization and intelligent sensing technology, image-based fruit tree disease identification has become one of the core technical directions in the research and application of smart orchards. Existing disease identification technologies have promoted the development of agricultural disease intelligent prevention and control to some extent. However, due to the limitations of traditional methods such as insufficient modeling ability, high dependence on artificial image features, poor adaptability, and lack of systematic application support, there are still significant deficiencies in pomegranate fruit full growth cycle and multi-class disease identification, which cannot meet the actual needs of orchards for early warning, stage tracking, and precise prevention and control of diseases. SUMMARY
[0004] In order to accurately perform early warning on fruits at different stages, the present application provides a pomegranate fruit full growth cycle multi-class disease identification method based on deep learning.
[0005] In the first aspect, the present application provides a pomegranate fruit full growth cycle multi-class disease identification method based on deep learning, which adopts the following technical solution:
[0006] A pomegranate fruit full growth cycle multi-class disease identification method based on deep learning, comprising:
[0007] Obtaining image samples of pomegranate fruits in different periods and different diseases in a target pomegranate orchard, and generating a disease image dataset;
[0008] Constructing a pomegranate fruit full cycle disease identification model based on the disease image dataset;
[0009] Obtaining disease-related information and prevention and control information;
[0010] A pomegranate fruit full-cycle disease knowledge base is constructed based on the disease image dataset, the disease-related information, and the prevention and control information.
[0011] Obtain current fruit image data from the target pomegranate orchard;
[0012] Based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, the current fruit image data is identified and analyzed to generate identification and analysis results.
[0013] By adopting the above technical solutions, a pomegranate disease control system integrating full-cycle monitoring, intelligent diagnosis, and precision management was constructed. By using an improved model and training it with a full-cycle disease image dataset, a full-cycle pomegranate fruit disease identification model was built, which significantly improved the identification accuracy and sensitivity of early-stage highly similar diseases, enabling early detection of minor lesions at different stages. At the same time, by combining the identification model with a multimodal pomegranate fruit full-cycle disease knowledge base containing prevention and control information, it can not only accurately determine the disease type and occurrence stage, but also immediately generate timely prevention and control strategies, forming a decision-making closed loop of discovery and treatment. It takes into account both high performance and lightweight design, making it easy to deploy on field terminal equipment. Ultimately, it has achieved a breakthrough in the management of pomegranate diseases from passive response to active early warning, and from fuzzy judgment to precision management, thereby accurately providing early warning for fruits at different stages.
[0014] Optionally, the generated disease image dataset includes:
[0015] To obtain information on the growth stages and disease types of pomegranate fruits;
[0016] The image samples are labeled based on the growth stage and the disease type to generate labeled image sample data.
[0017] The image sample data is subjected to image data augmentation processing to generate enhanced labeled image samples;
[0018] The enhanced image samples are organized according to the growth stage and the disease type to generate a disease image dataset.
[0019] Optionally, constructing a pomegranate fruit full-cycle disease identification model based on the disease image dataset includes:
[0020] Obtain the target recognition model and the model structure of the target model;
[0021] Based on preset recognition requirements, the model structure is adjusted to generate an adjusted recognition model;
[0022] The adjustment recognition model is trained based on the disease image dataset to construct a full-cycle disease recognition model for pomegranate fruits.
[0023] Optionally, the step of adjusting the model structure based on preset recognition requirements to generate an adjusted recognition model includes:
[0024] Based on the model structure, the RepVGG module is used to replace the downsampling module in the backbone network of the target recognition model, and a global attention mechanism (GAM) is introduced in the backbone stage. The Ghost module is used to replace the C2f structure in the backbone.
[0025] The upsampling stage of the Neck layer in the model structure uses the DySample lightweight dynamic sampling operator to replace the original upsampling module.
[0026] An SCConv structured attention module is added before the head of the detection head in the model structure;
[0027] The loss function of the aforementioned model structure is replaced with SlideLoss to generate an adjusted recognition model.
[0028] Optionally, the construction of a pomegranate fruit full-cycle disease knowledge base based on the disease image dataset, the disease-related information, and the prevention and control information includes:
[0029] Key information related to the disease was extracted to identify typical symptoms.
[0030] Based on the typical symptoms, key content of the prevention and control information is extracted to determine the prevention and control plan and drug use plan corresponding to each typical symptom.
[0031] The disease image dataset is filtered to determine representative images of the disease.
[0032] By binding and storing the representative images of the diseases with the corresponding typical symptoms, prevention and control measures, and drug application plans, a knowledge base for pomegranate fruit diseases throughout their entire life cycle is constructed.
[0033] Optionally, the identification and analysis of the current fruit image data based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, and the generation of identification and analysis results, include:
[0034] Based on the pomegranate fruit full-cycle disease identification model, stage disease analysis is performed on the current fruit image data to generate stage disease analysis results.
[0035] The disease analysis results are used to search for treatment plans in the pomegranate fruit full-cycle disease knowledge base to generate prevention and control strategies.
[0036] The disease analysis results at each stage are linked with the prevention and control strategies to generate identification and analysis results.
[0037] Optionally, after binding the stage disease analysis results and the prevention and control strategy to generate identification analysis results, the method further includes:
[0038] Obtain the staff's mobile terminal and storage requirements, and send the identification and analysis results to the mobile terminal;
[0039] The analysis results are stored based on the aforementioned storage requirements.
[0040] Secondly, this application provides a deep learning-based device for identifying multiple types of diseases throughout the entire growth cycle of pomegranate fruits, employing the following technical solution:
[0041] A deep learning-based device for identifying multiple diseases throughout the entire growth cycle of pomegranate fruit, comprising:
[0042] The data sample acquisition module is used to acquire image samples of pomegranate fruits at different times and with different diseases in the target pomegranate orchard, and generate a disease image dataset.
[0043] The identification model construction module is used to construct a pomegranate fruit full-cycle disease identification model based on the disease image dataset.
[0044] The relevant information acquisition module is used to acquire disease-related information and prevention and control information;
[0045] The disease knowledge construction module is used to construct a pomegranate fruit full-cycle disease knowledge base based on the disease image dataset, the disease-related information and the prevention and control information.
[0046] The current data acquisition module is used to acquire the current fruit image data of the target pomegranate orchard;
[0047] The analysis result generation module is used to identify and analyze the current fruit image data based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, and generate identification and analysis results.
[0048] By adopting the above technical solutions, a pomegranate disease control system integrating full-cycle monitoring, intelligent diagnosis, and precision management was constructed. By using an improved model and training it with a full-cycle disease image dataset, a full-cycle pomegranate fruit disease identification model was built, which significantly improved the identification accuracy and sensitivity of early-stage highly similar diseases, enabling early detection of minor lesions at different stages. At the same time, by combining the identification model with a multimodal pomegranate fruit full-cycle disease knowledge base containing prevention and control information, it can not only accurately determine the disease type and occurrence stage, but also immediately generate timely prevention and control strategies, forming a decision-making closed loop of discovery and treatment. It takes into account both high performance and lightweight design, making it easy to deploy on field terminal equipment. Ultimately, it has achieved a breakthrough in the management of pomegranate diseases from passive response to active early warning, and from fuzzy judgment to precision management, thereby accurately providing early warning for fruits at different stages.
[0049] Thirdly, this application provides an electronic device that adopts the following technical solution:
[0050] An electronic device includes a processor coupled to a memory;
[0051] The processor is used to execute a computer program stored in the memory, so that the electronic device executes the computer program of the deep learning-based method for identifying multiple diseases throughout the entire growth cycle of pomegranate fruit as described in any of the first aspects.
[0052] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution:
[0053] A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the deep learning-based method for identifying multiple diseases throughout the entire growth cycle of pomegranate fruit as described in any of the first aspects.
[0054] In summary, this application includes at least one of the following beneficial technical effects:
[0055] A pomegranate disease control system integrating full-cycle monitoring, intelligent diagnosis, and precision management was constructed. By adopting an improved model and training it with a full-cycle disease image dataset, a full-cycle pomegranate fruit disease identification model was built, significantly improving the identification accuracy and sensitivity of early-stage highly similar diseases. This enabled the early capture of minor lesions at different stages. Furthermore, by integrating the identification model with a multimodal pomegranate fruit full-cycle disease knowledge base containing prevention and control information, it can not only accurately determine the disease type and stage of occurrence but also immediately generate timely prevention and control strategies, forming a closed-loop decision-making system of discovery and treatment. The system combines high performance with a lightweight design, making it easy to deploy on field terminal equipment. Ultimately, this system represents a breakthrough in pomegranate disease control, moving from passive response to proactive early warning and from fuzzy judgment to precision management, thereby accurately providing early warnings for fruits at different stages. Attached Figure Description
[0056] Figure 1 This is a flowchart illustrating a deep learning-based method for identifying multiple diseases throughout the entire growth cycle of pomegranate fruit, as provided in an embodiment of this application.
[0057] Figure 2 This is a schematic diagram of enhanced images of pomegranate fruit diseases provided in an embodiment of this application.
[0058] Figure 3 This is a structural block diagram of the YOLOv8 network structure provided in the embodiments of this application.
[0059] Figure 4 This is a structural block diagram of the improved network structure provided in the embodiments of this application.
[0060] Figure 5 This is a structural block diagram of a deep learning-based device for identifying multiple types of diseases throughout the entire growth cycle of pomegranate fruit, provided in an embodiment of this application.
[0061] Figure 6 This is a structural block diagram of the electronic device provided in the embodiments of this application. Detailed Implementation
[0062] The present application will be further described in detail below with reference to the accompanying drawings.
[0063] This application provides a deep learning-based method for identifying multiple diseases throughout the entire growth cycle of pomegranate fruit. This method can be executed by an electronic device, which can be a server or a terminal device. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smartphone, tablet computer, desktop computer, etc., but is not limited to these.
[0064] Figure 1 This is a flowchart illustrating a deep learning-based method for identifying multiple diseases throughout the entire growth cycle of pomegranate fruit, as provided in an embodiment of this application.
[0065] like Figure 1 As shown, the main process of this method is described below (steps S101 to S106):
[0066] Step S101: Obtain image samples of pomegranate fruits with different diseases at different times in the target pomegranate orchard, and generate a disease image dataset.
[0067] For step S101, the growth stages and disease types of pomegranate fruits are obtained; the image samples are labeled based on the growth stages and disease types to generate labeled image sample data; the image sample data is enhanced to generate enhanced labeled image samples; the enhanced image samples are organized according to the growth stages and disease types to generate a disease image dataset.
[0068] In this embodiment, image data of pomegranate fruits at different growth stages were collected in the target pomegranate orchard environment. The growth stages include the young fruit stage, the expansion stage, and the ripening stage, covering a variety of typical disease types, including anthracnose, bacterial wilt, and Alternaria fruit spot disease. A high-quality image dataset of pomegranate fruits with multiple diseases was established, and data annotation and preprocessing were completed to provide standard data support for subsequent deep learning model training.
[0069] This invention collected image samples of pomegranate diseases at different stages in pomegranate orchards in Shaanxi Province, constructing a pomegranate fruit image dataset containing multiple types of diseases. The collection covered four common pomegranate diseases and healthy fruit samples. Image collection was concentrated in two time periods daily: 8:00–11:00 and 13:00–18:00. The collection process covered multiple angles, distances, different lighting conditions, and various weather changes to ensure data diversity and representativeness. To improve the model's generalization ability, a large number of additional samples were supplemented from publicly available pomegranate fruit disease datasets. Finally, a high-quality image dataset containing images of each type of disease and healthy fruit was constructed.
[0070] Due to factors such as complex backgrounds, fruit occlusion, uneven lighting, and similar features among some disease images in natural scenes, the original image data contains noise interference, affecting the training effect of multi-category disease recognition models. To improve the model's recognition robustness and detection accuracy, the dataset needs to be augmented. Histogram Equalization (HE) algorithm is used to enhance the images, adjusting the grayscale distribution in the HSI color space, enhancing image contrast, making the distinction between pomegranate fruit disease areas and background clearer, and improving feature extraction. A comparison of the effects before and after image enhancement is provided. Figure 2 As shown, image samples with blurred targets or insufficient information are removed to ensure the quality of training data and the robustness of the recognition model.
[0071] Step S102: Construct a full-cycle disease identification model for pomegranate fruits based on the disease image dataset.
[0072] For step S102, the target recognition model and the model structure of the target model are obtained; the model structure is adjusted based on the preset recognition requirements to generate an adjusted recognition model; the adjusted recognition model is trained based on the disease image dataset to construct a full-cycle disease recognition model for pomegranate fruit.
[0073] In this embodiment, reference is made to... Figure 3 and Figure 4 To address the issue of numerous pomegranate fruit diseases in natural pomegranate orchard settings, and the high similarity of different diseases in their early stages, which easily leads to false positives and false negatives, a dual design combining structural enhancement and lightweight optimization significantly improves the model's disease recognition accuracy, recall, and deployment adaptability in complex environments. The target recognition model uses a YOLOv8 backbone network. The structure of the YOLOv8 backbone network was determined, and the model structure was adjusted and modified according to actual needs to obtain an adjusted recognition model for training. This adjusted recognition model was then trained using a disease image dataset to obtain a full-cycle pomegranate fruit disease recognition model.
[0074] Furthermore, based on the preset recognition requirements, the model structure is adjusted to generate an adjusted recognition model, including: based on the model structure, replacing the downsampling module in the backbone network of the target recognition model with the RepVGG module, introducing the global attention mechanism GAM in the backbone stage, and replacing the C2f structure in the backbone with the Ghost module; replacing the original upsampling module with the DySample lightweight dynamic sampling operator in the upsampling stage of the Neck layer of the model structure; adding the SCConv structured attention module before the detection head of the model structure; and replacing the loss function of the model structure with SlideLoss to generate the adjusted recognition model.
[0075] In terms of model structure design, the convolution stacking concept of the VGG network is adopted, and the downsampling module in the YOLOv8 backbone network is replaced with the RepVGG module to enhance the ability to extract complex image features of pomegranate fruit diseases. Simultaneously, a Global Attention Mechanism (GAM) is introduced in the backbone stage to enhance the model's ability to model the global dependencies of key disease regions and improve the perception of salient disease areas. Furthermore, to effectively address the problem of imbalanced pomegranate disease categories, SlideLoss is used instead of the original YOLOv8 loss function to dynamically focus on easily confused disease samples and edge samples, improving the model's ability to distinguish difficult-to-identify disease categories and ensuring accurate identification of multiple disease categories covering the entire growth cycle of pomegranate fruit.
[0076] Regarding model lightweighting, to meet the deployment requirements of embedded devices, the Ghost module replaces the C2f structure in the Backbone, significantly reducing the number of model parameters and computational complexity. In the Neck layer upsampling stage, the DySample lightweight dynamic sampling operator is introduced to replace the traditional upsampling module, reducing data transmission and computational costs. To compensate for the performance loss caused by lightweighting, the SCConv structured attention module is added before the detection head to reduce feature redundancy and enhance the model's ability to focus on lesion areas, ensuring improved deployment efficiency while maintaining high recognition performance. Test results on a self-built pomegranate disease dataset show an accuracy of 94.3%, a 2.4 percentage point improvement over the original YOLOv8, with a 28% reduction in memory usage, effectively achieving a dual improvement in disease recognition accuracy and lightweight deployment.
[0077] Step S103: Obtain disease-related information and prevention and control information.
[0078] In this embodiment, disease-related information refers to the typical characteristics of the disease, such as typical symptoms, susceptible periods, and susceptible environments. Prevention and control information refers to the prevention and treatment measures for different diseases, including prevention and control plans, drug application plans, and treatment effects. It should be noted that specific disease-related and prevention and control information may be added and configured according to actual needs, but no specific limitations are imposed here.
[0079] Step S104: Construct a pomegranate fruit full-cycle disease knowledge base based on disease image datasets, disease-related information, and prevention and control information.
[0080] For step S104, key content of disease-related information is extracted to determine typical disease symptoms; based on the typical disease symptoms, key content of prevention and control information is extracted to determine the prevention and control plan and drug application plan corresponding to each typical disease symptom; the disease image dataset is filtered to determine representative disease images; the representative disease images are bound and stored with the corresponding typical disease symptoms, prevention and control plans and drug application plans to construct a pomegranate fruit full-cycle disease knowledge base.
[0081] In this embodiment, a multi-category disease knowledge base covering the entire growth cycle of pomegranate fruit is constructed to provide professional and actionable prevention and control strategies for pomegranate disease identification results in orchards. The knowledge base systematically manages image samples, typical symptoms, control strategies, and pesticide information for multiple categories of pomegranate diseases throughout their life cycle, achieving structured storage and efficient retrieval of disease information. The system front-end uses Webpack for static resource compression to improve access efficiency, while the back-end utilizes Redis caching technology to accelerate high-frequency data calls and introduces a fast image upload mechanism to optimize image processing and loading speed. It integrates core functions such as disease image library management, multi-category disease identification, control strategy recommendation, and pesticide management, enabling linked access between identification results and the knowledge base. This automatically matches diagnostic information and precise control strategies for corresponding disease types, providing fruit growers with scientific, efficient, and accurate prevention and control guidance and decision support.
[0082] Step S105: Obtain the current fruit image data of the target pomegranate orchard.
[0083] In this embodiment, an image acquisition device is installed in the target pomegranate orchard to acquire images according to actual needs. The acquired images containing pomegranate fruits are used as the current fruit image data. It should be noted that multiple image acquisition devices are installed in a single target pomegranate orchard; therefore, the number of current fruit image data is not unique.
[0084] Step S106: Based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, the current fruit image data is identified and analyzed to generate identification and analysis results.
[0085] For step S106, based on the pomegranate fruit full-cycle disease identification model, stage disease analysis is performed on the current fruit image data to generate stage disease analysis results; the disease analysis results are used to search for treatment plans in the pomegranate fruit full-cycle disease knowledge base to generate prevention and control strategies; the stage disease analysis results and prevention and control strategies are bound together to generate identification analysis results.
[0086] In this embodiment, the trained pomegranate fruit full-cycle disease identification model and pomegranate fruit full-cycle disease knowledge base are deployed on an edge-embedded smart terminal. The current fruit image data is input into the terminal. First, the pomegranate fruit full-cycle disease identification model is used to identify and analyze the current fruit image data to determine the growth stage and disease condition corresponding to the current fruit image data, thus obtaining the stage disease analysis result. The stage disease analysis result is then used to search and compare in the pomegranate fruit full-cycle disease knowledge base to determine the treatment plan that needs to be implemented under the current growth stage and disease condition. The treatment plan includes the control measures to be used and the corresponding pesticide usage. Classic cases and successful treatment cases are also identified. The pesticide usage, classic cases, and successful treatment cases are bound together to obtain the control strategy. Finally, the stage disease analysis result and the control strategy are bound together to obtain the final identification and analysis result.
[0087] In this embodiment, the mobile terminal and storage requirements of the staff are obtained, and the identification and analysis results are sent to the mobile terminal; the analysis results are stored based on the storage requirements.
[0088] After generating the recognition and analysis results, feedback is required. The mobile terminal used by the staff is determined, and the recognition and analysis results are sent to the staff's mobile terminal. It is then checked whether the recognition and analysis results need to be stored. If storage is required, the actual storage location is determined and the recognition and analysis results are stored in the storage location. If storage is not required, the process ends after sending the results to the mobile terminal.
[0089] Figure 5 The structural block diagram of a pomegranate fruit multi-category disease identification device 200 based on deep learning throughout the entire growth cycle is provided for the application embodiment.
[0090] like Figure 5 As shown, the deep learning-based pomegranate fruit full growth cycle multi-category disease identification device 200 mainly includes:
[0091] The data sample acquisition module 201 is used to acquire image samples of pomegranate fruits with different diseases at different times in the target pomegranate orchard and generate a disease image dataset.
[0092] The identification model building module 202 is used to build a pomegranate fruit full-cycle disease identification model based on the disease image dataset;
[0093] The relevant information acquisition module 203 is used to acquire disease-related information and prevention and control information;
[0094] The disease knowledge construction module 204 is used to construct a pomegranate fruit full-cycle disease knowledge base based on disease image datasets, disease-related information and prevention and control information.
[0095] The current data acquisition module 205 is used to acquire the current fruit image data of the target pomegranate orchard;
[0096] The analysis result generation module 206 is used to identify and analyze the current fruit image data based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, and generate identification and analysis results.
[0097] As an optional implementation of this embodiment, the data sample acquisition module 201 is specifically used to acquire the growth stages and disease types of pomegranate fruits; to label image samples based on the growth stages and disease types to generate labeled image sample data; to perform image data enhancement processing on the image sample data to generate enhanced labeled image samples; and to organize the enhanced image samples according to the growth stages and disease types to generate a disease image dataset.
[0098] As an optional implementation of this embodiment, the recognition model construction module 202 includes:
[0099] The model information acquisition module is used to acquire the target recognition model and the model structure of the target model;
[0100] The model generation module is used to adjust the model structure based on preset recognition requirements and generate an adjusted recognition model.
[0101] The disease model building module is used to train the adjustment recognition model based on the disease image dataset to build a full-cycle disease recognition model for pomegranate fruits.
[0102] In this optional embodiment, the adjusted model generation module is specifically used to, based on the model structure, replace the downsampling module in the backbone network of the target recognition model with the RepVGG module, introduce a global attention mechanism (GAM) in the backbone stage, and replace the C2f structure in the backbone with the Ghost module; in the upsampling stage of the Neck layer of the model structure, the DySample lightweight dynamic sampling operator is used to replace the original upsampling module; the SCConv structured attention module is added before the detection head of the model structure; and the SlideLoss loss function of the model structure is used to generate the adjusted recognition model.
[0103] As an optional implementation of this embodiment, the disease knowledge construction module 204 is specifically used to extract key content of disease-related information and determine typical disease symptoms; extract key content of prevention and control information based on typical disease symptoms and determine the prevention and control plan and drug use plan corresponding to each typical disease symptom; perform image screening on the disease image dataset to determine representative disease images; bind and store the representative disease images with the corresponding typical disease symptoms, prevention and control plans and drug use plans to construct a pomegranate fruit full-cycle disease knowledge base.
[0104] As an optional implementation of this embodiment, the analysis result generation module 206 is specifically used to perform stage disease analysis on the current fruit image data based on the pomegranate fruit full-cycle disease identification model, and generate stage disease analysis results; search for treatment plans in the pomegranate fruit full-cycle disease knowledge base based on the disease analysis results, and generate prevention and control strategies; and bind the stage disease analysis results and prevention and control strategies to generate identification analysis results.
[0105] As an optional implementation of this embodiment, the deep learning-based pomegranate fruit full growth cycle multi-category disease identification device 200 further includes:
[0106] The personnel needs acquisition module is used to acquire the mobile terminals and storage needs of staff, and send the identification and analysis results to the mobile terminals.
[0107] The analysis results storage module is used to store analysis results based on storage requirements.
[0108] In one example, the module in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.
[0109] For example, when modules in a device can be implemented via a processing element scheduler, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling programs. Alternatively, these modules can be integrated together as a system-on-a-chip (SOC).
[0110] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0111] Figure 6 This is a structural block diagram of the electronic device 300 provided in an embodiment of this application.
[0112] like Figure 6 As shown, the electronic device 300 includes a processor 301 and a memory 302, and may further include one or more of an information input / output (I / O) interface 303, a communication component 304, and a communication bus 305.
[0113] The processor 301 controls the overall operation of the electronic device 300 to complete all or part of the steps of the deep learning-based method for identifying multiple diseases throughout the entire growth cycle of pomegranate fruit. The memory 302 stores various types of data to support the operation of the electronic device 300. This data may include, for example, instructions for any application or method operating on the electronic device 300, as well as application-related data. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as one or more of Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0114] I / O interface 303 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 304 is used for wired or wireless communication between electronic device 300 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 304 may include a Wi-Fi component, a Bluetooth component, and an NFC component.
[0115] The electronic device 300 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the deep learning-based multi-category disease identification method for the entire growth cycle of pomegranate fruit given in the above embodiments.
[0116] The communication bus 305 may include a path for transmitting information between the aforementioned components. The communication bus 305 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 305 may be divided into an address bus, a data bus, a control bus, etc.
[0117] Electronic device 300 may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers, and may also be servers.
[0118] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described deep learning-based method for identifying multiple types of diseases throughout the entire growth cycle of pomegranate fruit.
[0119] The computer-readable storage medium may include 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.
[0120] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0121] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.
Claims
1. A method for identifying multiple types of diseases throughout the entire growth cycle of pomegranate fruit based on deep learning, characterized in that, include: Acquire image samples of pomegranate fruits at different stages and with different diseases in the target pomegranate orchard, and generate a disease image dataset; A full-cycle disease identification model for pomegranate fruits was constructed based on the aforementioned disease image dataset. Obtain disease-related and prevention information; A pomegranate fruit full-cycle disease knowledge base is constructed based on the disease image dataset, the disease-related information, and the prevention and control information. Obtain current fruit image data from the target pomegranate orchard; Based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, the current fruit image data is identified and analyzed to generate identification and analysis results.
2. The method according to claim 1, characterized in that, The generated disease image dataset includes: To obtain information on the growth stages and disease types of pomegranate fruits; The image samples are labeled based on the growth stage and the disease type to generate labeled image sample data. The image sample data is subjected to image data augmentation processing to generate enhanced labeled image samples; The enhanced image samples are organized according to the growth stage and the disease type to generate a disease image dataset.
3. The method according to claim 1, characterized in that, The construction of a pomegranate fruit full-cycle disease identification model based on the disease image dataset includes: Obtain the target recognition model and the model structure of the target model; Based on preset recognition requirements, the model structure is adjusted to generate an adjusted recognition model; The adjustment recognition model is trained based on the disease image dataset to construct a full-cycle disease recognition model for pomegranate fruits.
4. The method according to claim 3, characterized in that, The step of adjusting the model structure based on preset recognition requirements to generate an adjusted recognition model includes: Based on the model structure, the RepVGG module is used to replace the downsampling module in the backbone network of the target recognition model, and a global attention mechanism (GAM) is introduced in the backbone stage. The Ghost module is used to replace the C2f structure in the backbone. The upsampling stage of the Neck layer in the model structure uses the DySample lightweight dynamic sampling operator to replace the original upsampling module. An SCConv structured attention module is added before the head of the detection head in the model structure; The loss function of the aforementioned model structure is replaced with SlideLoss to generate an adjusted recognition model.
5. The method according to claim 1, characterized in that, The construction of a pomegranate fruit full-cycle disease knowledge base based on the disease image dataset, the disease-related information, and the prevention and control information includes: Key information related to the disease was extracted to identify typical symptoms. Based on the typical symptoms, key content of the prevention and control information is extracted to determine the prevention and control plan and drug use plan corresponding to each typical symptom. The disease image dataset is filtered to determine representative images of the disease. By binding and storing the representative images of the diseases with the corresponding typical symptoms, prevention and control measures, and drug application plans, a knowledge base for pomegranate fruit diseases throughout their entire life cycle is constructed.
6. The method according to claim 1, characterized in that, The identification and analysis of the current fruit image data based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, generating identification and analysis results including: Based on the pomegranate fruit full-cycle disease identification model, stage disease analysis is performed on the current fruit image data to generate stage disease analysis results. The disease analysis results are used to search for treatment plans in the pomegranate fruit full-cycle disease knowledge base to generate prevention and control strategies. The disease analysis results at each stage are linked with the prevention and control strategies to generate identification and analysis results.
7. The method according to claim 6, characterized in that, After binding the stage disease analysis results and the prevention and control strategies to generate identification analysis results, the method further includes: Obtain the staff's mobile terminal and storage requirements, and send the identification and analysis results to the mobile terminal; The analysis results are stored based on the aforementioned storage requirements.
8. A deep learning-based device for identifying multiple types of diseases throughout the entire growth cycle of pomegranate fruit, characterized in that, include: The data sample acquisition module is used to acquire image samples of pomegranate fruits at different times and with different diseases in the target pomegranate orchard, and generate a disease image dataset. The identification model construction module is used to construct a pomegranate fruit full-cycle disease identification model based on the disease image dataset. The relevant information acquisition module is used to acquire disease-related information and prevention and control information; The disease knowledge construction module is used to construct a pomegranate fruit full-cycle disease knowledge base based on the disease image dataset, the disease-related information and the prevention and control information. The current data acquisition module is used to acquire the current fruit image data of the target pomegranate orchard; The analysis result generation module is used to identify and analyze the current fruit image data based on the pomegranate fruit full-cycle disease identification model and the pomegranate fruit full-cycle disease knowledge base, and generate identification and analysis results.
9. An electronic device, characterized in that, Includes a processor, which is coupled to a memory; The processor is configured to execute a computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, It includes a computer program or instructions that, when run on a computer, cause the computer to perform the method as described in any one of claims 1 to 7.