Tea garden disease and pest identification method and system based on deep learning
Through deep learning technology, the identification of pests and diseases in tea gardens has been solved, and the problems of inaccurate positioning and untimely identification of traditional monitoring methods have been achieved, real-time monitoring and fine distribution maps of tea gardens have been achieved, scientific prevention and control suggestions have been provided, and tea garden production and quality assurance have been improved. .
Patent Information
- Application Number
- CN202411211043.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-12-06
AI Technical Summary
The prior art relies on manual observation and traditional equipment in the monitoring of pests and diseases in tea gardens, resulting in inaccurate positioning and untimely identification, and the inability to realize real-time monitoring of the entire region, making it difficult to provide a fine pest distribution map, resulting in lagging prevention and control measures.
The tea garden pest and disease identification method based on deep learning is adopted to locate the concentrated areas of pest and disease through the pest and disease target detection model, and use the regional distribution analysis model to perform more refined segmentation, generate pest and disease distribution maps, and formulate prevention and control suggestions based on environmental information.
It improves the accuracy of pest positioning and identification accuracy, realizes real-time monitoring of tea gardens throughout the entire region, provides a detailed pest distribution map, scientific and reasonable decision-making suggestions, and improves pesticide utilization and environmental protection.
Smart Images

Figure CN119091380A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to a method and system for identifying tea garden pests and diseases based on deep learning. Background Art
[0002] As the tea industry becomes increasingly prosperous, tea plantation pests and diseases become more prominent. Pests and diseases have always been an important factor affecting tea production and quality. They not only lead to a decrease in tea production, but also may affect the appearance, flavor and quality of tea. In severe cases, they may even lead to the death of tea trees. Early detection and accurate identification of pests and diseases are crucial for the health management of tea plantations and the timely implementation of prevention and control measures.
[0003] At present, the monitoring of pests and diseases in tea gardens mainly relies on manual observation and simple traditional monitoring equipment. Manual observation is not only time-consuming and laborious, but also due to the different experience levels of observers, the accuracy and timeliness of pest and disease identification are difficult to guarantee. Traditional monitoring equipment can often only cover part of the tea garden, and cannot achieve real-time monitoring of the entire tea garden. The distribution of pests and diseases cannot be fed back in time, resulting in delayed prevention and control measures. In addition, traditional methods are difficult to provide detailed pest and disease distribution maps, lack of accurate analysis of the specific distribution areas and severity of pests and diseases, and managers are prone to making mistakes in decision-making. Summary of the invention
[0004] In order to solve the above problems, the present invention proposes a tea garden pest and disease identification method and system based on deep learning. The method first uses a pest and disease target detection model to locate the concentrated area of pests and diseases, and then performs more refined segmentation through a regional distribution analysis model to obtain a pest and disease distribution map, thereby greatly improving the positioning accuracy and recognition precision.
[0005] In order to achieve the above object, the present invention adopts the following technical solution:
[0006] In a first aspect, the present invention provides a method for identifying tea garden pests and diseases based on deep learning, comprising:
[0007] Acquire tea leaves images and environmental information in tea gardens; pre-process tea leaves images in tea gardens;
[0008] The pest target detection model is used to perform preliminary pest identification on the preprocessed tea images in the tea garden, and the pest concentrated area images and pest types are obtained;
[0009] Input the pest and disease concentrated area images and pest and disease types into the regional distribution analysis model to perform in-depth pest and disease identification, obtain pest and disease target images, and generate pest and disease distribution maps;
[0010] Formulate prevention and control recommendations based on environmental information, images of areas where pests and diseases are concentrated, pest and disease types, and pest and disease distribution maps.
[0011] Preferably, the environmental information includes the temperature, humidity, soil pH value, wind speed and wind direction information of the tea garden.
[0012] Preferably, the preprocessing includes: performing uniform brightness correction, histogram equalization, image size adjustment and normalization processing on the tea garden tea image.
[0013] Preferably, the pest and disease target detection model includes a CSPDarknet network, a multi-scale feature fusion network and an Anchor-free based prediction module;
[0014] The CSPDarknet network includes multiple convolution modules, C2F modules and SPPF modules; the convolution module is used to extract low-level features, the C2F module is used to fuse the low-level features output by the convolution module; the SPPF module is used to perform multi-level interaction on the fused features output by the C2F module;
[0015] The multi-scale feature fusion network consists of a feature pyramid network and a path aggregation network, and is used to perform multi-scale fusion of the interactive features output by the SPPF module;
[0016] The anchor-free based prediction module receives the multi-scale features output by the multi-scale feature fusion network and outputs the category loss and prediction box loss of the target.
[0017] Preferably, the method of inputting the pest and disease concentrated area image and pest and disease type into the regional distribution analysis model to perform deep pest and disease identification to obtain the pest and disease target image specifically includes:
[0018] The image of the pest and disease concentrated area is passed through continuous convolutional layers and activation layers to extract the initial feature map; the initial feature map is extracted through the maximum pooling layer to extract the high-level feature map;
[0019] The high-level feature maps are gradually restored to the resolution of the original image to obtain the restored image, and the restored images are spliced and fused through jump connections;
[0020] According to the pest type, a classifier trained based on the pest database corresponding to the pest type is called to classify the spliced and fused features, extract the pest target foreground, and obtain the pest target image.
[0021] Preferably, the pest and disease distribution map is obtained by splicing all the tea garden tea images obtained in the tea garden, marking the pest and disease target images belonging to the same pest and disease type with the same color based on the position of the pest and disease target image, marking different pest and disease types with different colors, and not processing the background pixels to obtain the pest and disease distribution map.
[0022] Preferably, the method of formulating prevention and control suggestions by combining environmental information, images of pest and disease concentrated areas, pest and disease types and pest and disease distribution maps specifically includes:
[0023] According to the pest distribution map, locate the areas of various pests and diseases; by calculating the area of the pest concentration area image of each pest and disease, obtain the distribution ratio of each pest and disease, and estimate the severity of each pest and disease;
[0024] According to the type of pests and diseases, corresponding prevention and control suggestions are called from the agricultural pests and diseases database; the prevention and control suggestions include the type of medicine, the amount of medicine applied and the time of medicine application;
[0025] Adjust prevention and control recommendations based on environmental information and the severity of the pest or disease.
[0026] In a second aspect, the present invention provides a tea garden pest and disease identification system based on deep learning, comprising:
[0027] The data acquisition module is configured to acquire tea leaves images and environmental information in the tea garden; and pre-process the tea leaves images in the tea garden;
[0028] The preliminary identification module is configured to use the pest target detection model to perform preliminary pest identification on the pre-processed tea images of the tea garden to obtain pest concentrated area images and pest types;
[0029] The deep recognition module is configured to input the pest and disease concentrated area image and pest and disease type into the regional distribution analysis model, perform deep pest and disease recognition, obtain the pest and disease target image, and generate the pest and disease distribution map;
[0030] The analysis and feedback module is configured to formulate prevention and control recommendations based on environmental information, images of pest and disease concentrated areas, pest and disease types, and pest and disease distribution maps.
[0031] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a method for identifying tea garden pests and diseases based on deep learning as described in the first aspect.
[0032] In a fourth aspect, the present invention provides a computer 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, the steps in the method for identifying pests and diseases in a tea garden based on deep learning described in the first aspect are implemented.
[0033] Compared with the prior art, the present invention has the following beneficial effects:
[0034] (1) The present invention obtains RGB image information of the tea garden to locate the area where the pests and diseases are located. In order to analyze the distribution of pests and diseases, the image of the concentrated area of pests and diseases is input into the regional distribution analysis network to analyze the type and detailed distribution of pests and diseases. Finally, the pest and disease analysis report and the environmental information fed back by other environmental sensors are combined to give decision-making suggestions to managers. By quickly locating the pest and disease area, the pests and diseases in the tea garden can be monitored in real time, the information feedback speed can be improved, and the decision-making suggestions obtained by combining the sensor feedback information and the pest and disease analysis are more scientific and reasonable. Managers can accurately control the prevention and control measures and effectively ensure the yield and quality of the tea garden.
[0035] (2) The present invention proposes a two-stage pest and disease positioning and identification method, which first locates the pest and disease concentration area, and then obtains the pest and disease distribution map through more refined segmentation, which greatly improves the positioning accuracy and identification precision.
[0036] (3) Based on the information from multiple environmental sensors and pest and disease analysis reports, the present invention enables managers to have a more comprehensive understanding of the tea garden environment, and to accurately prevent and control different pests and diseases and their severity, thereby improving the utilization rate of pesticides and reducing environmental pollution.
[0037] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention but do not constitute a limitation of the present invention.
[0039] Figure 1 A main flow chart of a tea garden pest and disease identification method based on deep learning provided by an embodiment of the present invention;
[0040] Figure 2 A detailed flow chart of a tea garden pest and disease identification method based on deep learning provided by an embodiment of the present invention;
[0041] Figure 3 A schematic diagram of a pest and disease target detection module provided in an embodiment of the present invention;
[0042] Figure 4 A schematic diagram of a regional distribution analysis model provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0043] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0044] Embodiment 1
[0045] like Figure 1 As shown, this embodiment discloses a tea garden pest and disease identification method based on deep learning, comprising the following steps:
[0046] S1: Acquire tea images and environmental information of tea gardens; pre-process tea images of tea gardens;
[0047] S2: Use the pest target detection model to perform preliminary pest identification on the pre-processed tea images in the tea garden to obtain the pest concentrated area image and pest type;
[0048] S3: Input the pest and disease concentrated area image and pest and disease type into the regional distribution analysis model, perform deep pest and disease identification, obtain the pest and disease target image, and generate the pest and disease distribution map;
[0049] S4: Formulate prevention and control recommendations based on environmental information, images of pest and disease concentration areas, pest and disease types, and pest and disease distribution maps.
[0050] Next, combine Figure 2 , a tea garden pest and disease identification method based on deep learning disclosed in this embodiment is described in detail.
[0051] In S1, 360-degree high-definition cameras are set up at multiple locations in the tea garden to collect images of tea crop areas in the tea garden to obtain tea images in the tea garden.
[0052] As a specific implementation method, the purpose of collecting tea images in tea gardens is to obtain tea bud images suitable for training. The shooting angle needs to be as close to the top view as possible, and the horizontal angle of the camera should be greater than 45 degrees. Each camera should collect a part of the tea garden area. If a tea garden is too large, it is planned to use multiple cameras to divide the area for division of labor.
[0053] In particular, since the brightness of the image will change under different lighting conditions, it is difficult to distinguish the image details. Therefore, this embodiment uses the current ambient light brightness fed back by the light sensor to perform brightness correction, histogram equalization, image size adjustment and normalization on the image to ensure the validity of the image.
[0054] Temperature and humidity sensors, soil sensors, and wind speed and direction sensors are installed at multiple locations in the tea garden to collect environmental information such as temperature, humidity, soil pH value, wind speed, and wind direction in the tea garden.
[0055] As a specific implementation method, since tea fields are distributed in blocks and the terrain is complex in actual tea gardens, the temperature and humidity sensors and wind speed and direction sensors are installed on the pole where the camera is located. The pole is high enough, and the wind speed and direction sensor is installed at a higher position on the pole to avoid being affected by terrain obstruction; the temperature and humidity sensor is installed at a lower position and is shielded by an opaque shell to avoid direct sunlight, so as to more accurately measure the temperature and humidity of the tea field area; the soil sensor is randomly distributed and placed in units of tea fields to increase the robustness of soil pH detection. It should be understood that the specific installation method and position can be set by technicians in this field according to actual conditions.
[0056] As an optional implementation method, this embodiment makes the preprocessed tea garden tea images into a target detection sample data set for building a tea garden pest and disease target detection model; more than 6,000 collected samples are manually selected and adjusted, and 3,600 superior samples with obvious features are selected from them, and Labelme software is used to mark the tea garden image samples with rectangular frames of various common tea bud pests and diseases types, and geometric transformation (including flipping, scaling, and translation), noise injection (including Gaussian noise and salt and pepper noise), affine transformation, blurring and sharpening (including Gaussian blur, motion blur, and image sharpening) operations are used for data enhancement to enhance the target detection data set to more than 7,800 images, thereby improving the generalization ability of the data set; the data set obtained in the above steps is divided into a training set, a validation set, and a test set in a ratio of 8:1:1, and input into the pest and disease target detection model for sufficient training to obtain a pest and disease target detection model.
[0057] In S2, the pest and disease target detection model is used to perform preliminary pest and disease identification on the preprocessed tea images in the tea garden to obtain images of concentrated pest and disease areas and pest and disease types.
[0058] like Figure 3 As shown in the figure, the pest and disease target detection model is mainly composed of the CSPDarknet network, the multi-scale feature fusion network and the Anchor-free based prediction module.
[0059] The CSPDarknet network consists of multiple convolutional modules, a C2F module and an SPPF module.
[0060] The convolution module is used to extract low-level features of the image, such as edges, textures, etc. The convolution module includes a convolution layer, a batch normalization layer, and an activation function layer.
[0061] Among them, the convolution layer generates a series of feature maps by sliding filters on the input image; the batch normalization layer standardizes the feature maps to make their mean 0 and variance 1, thereby accelerating the training process; the activation function layer applies nonlinear functions to the feature maps to enhance the expressive power of the model.
[0062] Multiple convolution modules are used to further extract fine-grained feature information and enhance the model's ability to capture details.
[0063] The C2F module is a core component used to achieve deeper feature extraction and feature fusion. The C2F module consists of a convolutional module and several residual units. The residual unit directly transmits two paths of input information through normal convolution and short-circuit connection, alleviating the gradient vanishing problem in deep networks and improving training stability and model accuracy. The C2F module fuses low-level features of different paths through partial convolution and skip connections of residual units to obtain fused features, thereby reducing computational costs and parameters while maintaining the representation ability of the model. This structure helps to improve the diversity and expression ability of features.
[0064] The SPPF module consists of a convolution module, a connection layer, and a maximum pooling layer. Maximum pooling is a downsampling operation used to reduce the size of feature maps while retaining important information. It reduces the resolution and computation of feature maps by applying a fixed-size sliding window to the feature map and taking the maximum value in each window as the output. Maximum pooling helps extract significant features in the image, enhances the translation invariance of the model, and reduces the risk of overfitting. The connection layer is used to transfer and fuse feature information between different scales, and realizes multi-level interaction of features through skip connections and lateral connections to ensure that information is fully propagated in the network.
[0065] In the SPPF module, after the fusion features are convolved, the connection layer and the maximum pooling layer perform multi-scale pooling operations on the input feature map and splice the results together to obtain multi-scale context information and achieve multi-level interaction. This method can improve the model's ability to recognize objects of different sizes while maintaining high efficiency.
[0066] The multi-scale feature fusion network consists of a feature pyramid network and a path aggregation network. The feature pyramid network FPN consists of an upsampling module, a connection layer, and a C2F module. The path aggregation network PAN consists of a convolution module, a C2F module, and a connection layer. Among them, the upsampling module is used to restore the low-resolution feature map to a higher resolution, so that the high-level semantic features can be fused with the low-level detail features to enhance the detection ability of small targets; the role of the connection layer, C2F module, and convolution module has been explained and will not be repeated. FPN and PAN together form a multi-scale feature fusion network, which improves the model's ability to learn multi-scale feature information.
[0067] The anchor-free method means that the model directly predicts the center point and the size of the bounding box of the target without pre-defining the anchor box, reducing the reliance on hyperparameters and simplifying the network structure. The prediction module based on anchor-free consists of three prediction modules that output detection results from features at three scales. The prediction module consists of a convolution module and a convolution layer, receives the fused multi-scale features, and outputs the category loss and prediction box loss of the target through regression and classification heads to quantify the quality of the training model.
[0068] The CSPDarknet network in this embodiment belongs to a network structure design method, which improves the efficiency and performance of the network by dividing the network into multiple stages and only partially connecting them in each stage. CSPDarknet is a Darknet variant based on the CSPNet structure. It uses the design concept of CSPNet on the basis of Darknet to improve feature extraction capabilities and network performance, especially when dealing with complex target detection tasks.
[0069] In this embodiment, the multi-scale feature fusion network adopts a combined structure of a feature pyramid network (FPN) and a path aggregation network (PAN) to further fuse and enhance features at different scales, improve the detection capability of pest and disease areas of different sizes, and finally generate the final target detection result through the Anchor-free mechanism, output the pest and disease detection image, and obtain the concentrated areas and types of pests and diseases.
[0070] In this embodiment, in the combination of feature pyramid network (FPN) and path aggregation network (PAN), the FPN structure fuses feature maps of different scales together, while the PAN network further enhances the effect of feature fusion through path aggregation. This combined structure fully integrates feature information of different levels at different scales, improves the network's detection ability for pest and disease areas of different sizes, and especially significantly improves the detection effect for small targets and long-distance targets.
[0071] In this embodiment, the traditional Anchor-based detection method requires a set of pre-set prior frames, while the Anchor-free mechanism directly predicts the bounding box of the target, reducing the impact of the preset frame on the detection result and improving the accuracy and robustness of the detection. At the same time, this mechanism is more adaptable to the complex and changeable tea garden environment and can better cope with pest and disease areas of different shapes, sizes and colors.
[0072] As a specific implementation method, the pest target detection model learns the shallow features of pests and diseases, mainly to regionally locate pests and diseases and identify the types of pests and diseases. Since the model does not need to determine the exact location of the pests and diseases, it only needs to be able to predict the approximate area where they are located, so the model learning is biased towards the large-scale feature information of pests and diseases. Sending the tea image of the tea garden into the pest target detection model can actually obtain the pest area predicted by the model, that is, mark various types of pests and diseases with rectangular frames; these several rectangular frames are cut to obtain the concentrated area image of pests and diseases, and the pest type is obtained based on the preset label. The pest types include four types of diseases: "tea red leaf spot", "tea tree anthracnose", "tea cake disease", and "tea white star disease", and four types of pests: "tea short-palped mite", "small green leafhopper", "tea poisonous moth", and "black thorn whitefly".
[0073] In S3, the pest and disease concentrated area image and pest and disease type are input into the regional distribution analysis model to perform deep pest and disease identification and obtain the pest and disease target image. Figure 4 As shown, it includes downsampling operation and upsampling operation; specifically:
[0074] S301: The downsampling operation includes: extracting an initial feature map from the image of the pest and disease concentrated area through a continuous convolution layer and an activation layer;
[0075] S302: extracting a high-level feature map from the initial feature map through a maximum pooling layer;
[0076] S303: The upsampling operation includes: gradually restoring the high-level feature map to the resolution of the original image to obtain a restored image, and splicing and fusing the restored image through jump connections.
[0077] According to the pest type, a classifier trained based on the pest database corresponding to the pest type is called to classify the spliced and fused features, extract the pest target foreground, and obtain a pest target image at the pixel level.
[0078] In this embodiment, the regional distribution analysis model learns the deep features of pests and diseases. Relatively speaking, the model learns the feature information of pests and diseases on a small scale. The pest and disease concentrated area image and the pest and disease type obtained in the previous step are input into the regional distribution analysis model. The model will call the classifier (Softmax multi-classification activation function) generated by training the pest and disease database of this type according to the pest and disease type, perform pixel-level semantic segmentation on the pest and disease concentrated area image, extract the pest and disease target foreground, and obtain the pest and disease target image.
[0079] It should be understood that the pest and disease database contains various common pest and disease types and corresponding pest and disease characteristics, which can be set by those skilled in the art according to actual conditions.
[0080] All the tea leaves images obtained in the tea garden are stitched together. Based on the position of the pest target image, the pest target images belonging to the same pest type are marked with the same color, and different pest types are marked with different colors. The background pixels are not processed to obtain the pest distribution map.
[0081] In this embodiment, the downsampling process uses continuous convolutional layers and activation layers, and the model can learn the deep features of the pest and disease images. These features can better represent the essential attributes of pests and diseases than the original pixel values. The maximum pooling layer further extracts the key information in these features, removes redundancy and noise, and thus improves the accuracy of recognition; the upsampling process model not only restores the resolution of the image, but also splices and fuses feature maps at different levels through jump connections. This multi-scale feature fusion method enables the model to capture the local details and global structure of pests and diseases at the same time, enhances the model's ability to recognize pests and diseases of different types and severity, and improves the generalization performance of the model.
[0082] As an optional implementation method, this embodiment manually screens and collects 1,200 images with a large number and type of pests and diseases from the collected tea garden images, manually annotates the pest and disease areas in the images, creates a training data set with pixel-level labels, and also uses the above-mentioned data enhancement technology to expand the data set, and finally obtains a semantic segmentation data set of 2,600 images; uses a U-Net-based semantic segmentation network to train the data set, adjusts hyperparameters to optimize model performance, and obtains a regional distribution analysis model.
[0083] In S4, prevention and control recommendations are formulated by combining environmental information, images of areas where pests and diseases are concentrated, pest and disease types, and pest and disease distribution maps.
[0084] In this embodiment, environmental information, images of pest and disease concentrated areas, pest and disease types, and pest and disease distribution maps constitute a feedback report, and management personnel perform targeted prevention and control based on the feedback report.
[0085] As a specific implementation method, according to the pest and disease distribution map, the distribution proportion of each type of pest and disease is obtained by calculating the area of the image of the pest and disease concentrated area where each type of pest and disease is located, and the severity is estimated by the area of the pest and disease concentrated area. When the pest and disease area proportion is 0%-10%, 10%-25%, 25%-50%, and above 50%, it is divided into: mild, moderate, severe, and extremely severe levels.
[0086] According to the type of pests and diseases, the control measures corresponding to the type of pests and diseases are retrieved through the agricultural pests and diseases database, and the control suggestions include the type of medicine to be used, the amount of medicine to be applied and the time of application.
[0087] Managers can apply pesticides or other control measures in a targeted manner according to the prevention and control measures, determine the amount of pesticides to be applied according to the severity of the pests and diseases, and adjust the prevention and control measures based on environmental data.
[0088] As an optional implementation method, in a feedback report, an image of the area where pests and diseases are concentrated shows that there is a small green leafhopper pest in a part of the 4th tea ridge in the No. 5 tea field. The pest and disease distribution map shows more specifically the specific location and distribution of the small green leafhopper on the 4th tea ridge. The pixel ratio calculation shows that the severity of the small green leafhopper pest is medium. The prevention and control suggestions for the small green leafhopper are automatically retrieved from the agricultural pest and disease database, and the sensor of the tea field where the small green leafhopper is located is called to display the current temperature and humidity, soil pH, wind speed and direction of the tea field.
[0089] Managers make judgments and analyses on this information and suggestions. Assuming that the prevention and control recommendations in the agricultural pest database suggest the use of pesticide A for green leafhoppers, because the severity is moderate, the application amount should also match. Assuming that pesticide A is volatile at high temperatures, managers determine the application time based on the temperature and determine whether it is suitable for application based on wind speed and direction. It is not suitable to apply pesticides when the wind speed is too high. Assuming manual application, the applicator should be upwind of the tea field. Finally, managers combine this information to issue or implement specific prevention and control measures.
[0090] It should be understood that the agricultural pest database is a relational database, which contains common pest types and corresponding prevention and control measures, such as the type of medicine used, the amount of medicine applied and the time of application.
[0091] This specific embodiment obtains tea image information in a tea garden through a 360-degree camera, and obtains environmental information of the tea garden through an environmental sensor; uses the light intensity fed back by the light sensor to perform uniform brightness correction and histogram equalization on the collected image to ensure the validity of the image; uses the pest target recognition network to identify pests and diseases in tea garden images, and quickly locates pest and disease areas and their types; inputs pest and disease area images into the regional distribution analysis model to identify the areas and extents of pest and disease distribution in the tea garden; combines environmental information and pest and disease distribution results to generate pest and disease distribution maps and feedback reports, and gives decision-making suggestions based on corresponding prevention and control measures. Through the method, not only can tea garden pests and diseases be monitored in real time and the speed of information feedback be improved, but also prevention and control measures can be accurately controlled to effectively ensure the yield and quality of tea gardens, effectively solving the problems of untimely feedback of pest and disease information and unclear distribution of pests and diseases in tea gardens.
[0092] Embodiment 2
[0093] This embodiment provides a tea garden pest and disease identification system based on deep learning, including:
[0094] The data acquisition module is configured to acquire tea leaves images and environmental information in the tea garden; and pre-process the tea leaves images in the tea garden;
[0095] The preliminary identification module is configured to use the pest target detection model to perform preliminary pest identification on the pre-processed tea images of the tea garden to obtain pest concentrated area images and pest types;
[0096] The deep recognition module is configured to input the pest and disease concentrated area image and pest and disease type into the regional distribution analysis model, perform deep pest and disease recognition, obtain the pest and disease target image, and generate the pest and disease distribution map;
[0097] The analysis and feedback module is configured to formulate prevention and control recommendations based on environmental information, images of pest and disease concentrated areas, pest and disease types, and pest and disease distribution maps.
[0098] Embodiment 3
[0099] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the method for identifying pests and diseases in a tea garden based on deep learning as described in the first embodiment above are implemented.
[0100] Embodiment 4
[0101] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for identifying pests and diseases in a tea garden based on deep learning as described in the first embodiment are implemented.
[0102] The steps or modules involved in the above embodiments 2 to 4 correspond to those in embodiment 1. For the specific implementation, please refer to the relevant description of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.
[0103] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A tea garden pest and disease identification method based on deep learning, characterized in that: include: Obtain tea plantation tea images and environmental information; Preprocess the tea leaves images in the tea garden; The pest target detection model is used to perform preliminary pest identification on the preprocessed tea images in the tea garden, and the pest concentrated area images and pest types are obtained; Input the pest and disease concentrated area images and pest and disease types into the regional distribution analysis model to perform in-depth pest and disease identification, obtain pest and disease target images, and generate pest and disease distribution maps; Formulate prevention and control recommendations based on environmental information, images of areas where pests and diseases are concentrated, pest and disease types, and pest and disease distribution maps.
2. A tea garden pest and disease identification method based on deep learning as claimed in claim 1, characterized in that: The environmental information includes the temperature, humidity, soil pH value, wind speed and wind direction information of the tea garden.
3. A tea garden pest and disease identification method based on deep learning as claimed in claim 1, characterized in that: The preprocessing includes: performing uniform brightness correction, histogram equalization, image size adjustment and normalization processing on the tea garden tea image.
4. A tea garden pest and disease identification method based on deep learning as claimed in claim 1, characterized in that: The pest target detection model includes a CSPDarknet network, a multi-scale feature fusion network and an Anchor-free based prediction module; The CSPDarknet network includes multiple convolution modules, C2F modules and SPPF modules; the convolution module is used to extract low-level features, the C2F module is used to fuse the low-level features output by the convolution module; the SPPF module is used to perform multi-level interaction on the fused features output by the C2F module; The multi-scale feature fusion network consists of a feature pyramid network and a path aggregation network, and is used to perform multi-scale fusion of the interactive features output by the SPPF module; The anchor-free based prediction module receives the multi-scale features output by the multi-scale feature fusion network and outputs the category loss and prediction box loss of the target.
5. The method for identifying tea garden pests and diseases based on deep learning as claimed in claim 1, characterized in that: The image of the pest and disease concentrated area and the pest and disease type are input into the regional distribution analysis model to perform deep pest and disease identification to obtain the pest and disease target image, specifically including: The image of the pest and disease concentrated area is passed through continuous convolutional layers and activation layers to extract the initial feature map; the initial feature map is extracted through the maximum pooling layer to extract the high-level feature map; The high-level feature maps are gradually restored to the resolution of the original image to obtain the restored image, and the restored images are spliced and fused through jump connections; According to the pest type, a classifier trained based on the pest database corresponding to the pest type is called to classify the spliced and fused features, extract the pest target foreground, and obtain the pest target image.
6. A tea garden pest and disease identification method based on deep learning as claimed in claim 1, characterized in that: The pest distribution map is obtained by splicing all the tea garden tea images obtained in the tea garden, marking the pest target images belonging to the same pest type with the same color based on the position of the pest target image, marking different pest types with different colors, and not processing the background pixels to obtain the pest distribution map.
7. The method for identifying tea garden pests and diseases based on deep learning as claimed in claim 1, characterized in that: Formulate prevention and control recommendations by combining environmental information, images of pest and disease concentrated areas, pest and disease types, and pest and disease distribution maps; Specifically include: According to the pest distribution map, locate the areas of various pests and diseases; by calculating the area of the pest concentration area image of each pest and disease, obtain the distribution ratio of each pest and disease, and estimate the severity of each pest and disease; According to the type of pests and diseases, corresponding prevention and control suggestions are called from the agricultural pests and diseases database; the prevention and control suggestions include the type of medicine, the amount of medicine applied and the time of medicine application; Adjust prevention and control recommendations based on environmental information and the severity of the pest or disease.
8. A tea garden pest and disease identification system based on deep learning, characterized in that: include: The data acquisition module is configured to acquire tea leaves images and environmental information in the tea garden; Preprocess the tea leaves images in the tea garden; The preliminary identification module is configured to use the pest target detection model to perform preliminary pest identification on the pre-processed tea images of the tea garden to obtain pest concentrated area images and pest types; The deep recognition module is configured to input the pest and disease concentrated area image and pest and disease type into the regional distribution analysis model, perform deep pest and disease recognition, obtain the pest and disease target image, and generate the pest and disease distribution map; The analysis and feedback module is configured to formulate prevention and control recommendations based on environmental information, images of pest and disease concentrated areas, pest and disease types, and pest and disease distribution maps.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, the steps in the method for identifying pests and diseases in a tea garden based on deep learning as described in any one of claims 1 to 7 are implemented.
10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the method for identifying pests and diseases in a tea garden based on deep learning are implemented as described in any one of claims 1-7.
Citation Information
Cited By
Long tail disease identification and pesticide application integrated method based on disease pesticide knowledge graph
CN119359476A
Plant growth condition monitoring method, system and equipment and storage medium
CN119478708A
