Green plant disease identification and early warning management system and method

Abstract

The invention relates to the technical field of green plant image recognition, in particular to a green plant disease recognition and early warning management system and method. The method comprises the following steps: acquiring a high-definition image of a green plant; performing image preprocessing on the green plant high-definition image to generate a standard green plant high-definition image; performing green plant part image segmentation and integration on the standard green plant high-definition image to obtain a standard green plant part image; performing part multi-dimensional disease feature extraction on the standard green plant part image to obtain green plant part multi-dimensional disease features; performing scab area identification on the standard green plant part image based on the multi-dimensional disease features of the green plant part to generate scab area identification data of the green plant part; and performing green plant disease type discrimination on the standard green plant part image according to the green plant part scab area identification data to generate green plant disease type discrimination data. According to the method, high-definition image processing, multi-dimensional feature extraction, propagation trend analysis and dynamic early warning are introduced, so that the accuracy and reaction time sequence of green plant disease recognition early warning are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of green plant image recognition, and particularly to a green plant disease recognition, warning and management system and method. Background Art

[0002] Early recognition of green plant diseases relied on manual inspection and traditional pesticide spraying methods, which had high labor costs and low efficiency. The technology at this stage mainly relied on the experience of agronomy experts, and judged the disease situation through on-site investigation and symptom recognition. However, this method had problems such as long diagnosis time and small coverage area, and was difficult to cope with large-scale green plant management. With the rise of computer vision technology, green plant disease recognition methods based on image processing gradually emerged. By collecting images of green plant leaves and using image analysis technology to identify disease symptoms, the diagnosis efficiency could be improved and manual intervention could be reduced. At this time, the introduction of artificial intelligence and machine learning algorithms provided more accurate recognition means for green plant disease recognition. In particular, the application of deep learning greatly improved the accuracy of disease recognition. However, current traditional disease management methods often ignore the dynamic changes of disease transmission, but focus on static recognition and treatment. At the same time, they often deal with the existing manifestations of diseases, lacking real-time monitoring and dynamic response, which leads to low accuracy and reaction timeliness of green plant disease recognition and warning. Summary of the Invention

[0003] Based on this, it is necessary to provide a green plant disease recognition, warning and management system and method to solve at least one of the above technical problems.

[0004] To achieve the above object, a green plant disease recognition, warning and management method, the method includes the following steps:

[0005] Step S1: Obtain a high-definition image of a green plant; perform image preprocessing on the high-definition image of the green plant to generate a standard high-definition image of the green plant; perform image segmentation and integration on the standard high-definition image of the green plant to obtain a standard image of the green plant part;

[0006] Step S2: Extract multi-dimensional disease features of the green plant part from the standard image of the green plant part to obtain multi-dimensional disease features of the green plant part; identify the lesion area of the standard image of the green plant part based on the multi-dimensional disease features of the green plant part to generate identification data of the lesion area of the green plant part; determine the type of green plant disease for the standard image of the green plant part according to the identification data of the lesion area of the green plant part to generate discrimination data of the type of green plant disease, where the discrimination data of the type of green plant disease includes minor disease types and similar disease types;

[0007] Step S3: Based on the minor disease types and similar disease types, analyze the transmission trend of plant diseases for the green plants with minor lesion types, generate plant disease transmission risk data, and convert the plant disease transmission risk data into a plant disease abnormal risk map through graph theory methods;

[0008] Step S4: Use the plant disease abnormal risk map to conduct a risk warning for the image of the diseased area of the plant part, generate plant disease risk warning data; construct a prevention and control plan based on the plant disease risk warning data to perform the operation of plant disease prevention and control management.

[0009] Through obtaining high-definition images and performing preprocessing, the present invention ensures the standardization of image quality and avoids the influence of low-quality images on the recognition effect. Image segmentation and integration further improve the accuracy of disease recognition, making subsequent processing more targeted and accurate. By extracting multi-dimensional disease characteristics of the plant part, it is possible to comprehensively analyze the disease manifestations and identify different types of diseased areas. The discrimination of disease types not only improves the accuracy of disease diagnosis but also distinguishes minor diseases and similar disease types, reducing the possibility of misdiagnosis and missed diagnosis. Through the analysis of the transmission trends of minor disease types and similar disease types, the spread of diseases can be predicted, and potential high-risk areas can be identified in advance. Graph theory methods convert the transmission risk data into a plant disease abnormal risk map, which helps to intuitively display the spread trend of diseases and provides a scientific basis for formulating prevention and control strategies. By using the disease abnormal risk map to warn of the risk of plant lesions, problems can be identified in a timely manner before the spread of diseases, avoiding further spread of diseases. Constructing a prevention and control plan based on the warning data can achieve precise prevention and control management, improve the efficiency and effect of disease control, and reduce resource waste. Therefore, the present invention improves the accuracy and response timeliness of plant disease recognition and warning by introducing high-definition image processing, multi-dimensional feature extraction, transmission trend analysis, and dynamic warning.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: Use a high-definition digital camera to obtain high-definition images of green plants;

[0012] Step S12: Perform image filtering on the high-definition images of green plants to generate high-definition filtered images of green plants;

[0013] Step S13: Perform image brightness enhancement on the high-definition filtered images of green plants to generate high-definition brightness-enhanced images of green plants;

[0014] Step S14: Perform image geometric transformation on the high-definition brightness-enhanced images of green plants to generate standard high-definition images of green plants; perform image segmentation and integration of the plant parts on the standard high-definition images of green plants to obtain standard images of plant parts.

[0015] The present invention obtains high-definition images of green plants through a high-definition digital camera, ensuring that the resolution and details of the images are clear enough to provide high-quality basic data for subsequent image processing and analysis. Image filtering helps to remove noise in the images and improve the clarity of the images. The filtered images are cleaner, which helps to improve the accuracy of subsequent processing and feature extraction. Image brightness enhancement can increase the contrast and brightness of the images, making the details of the green plants more prominent, which helps to improve the quality of images taken under different lighting conditions and ensure that the visual effects of the images better meet the analysis requirements. Geometric transformations such as rotation, scaling, or cropping can standardize the images, eliminate differences in shooting angles and perspectives, ensure that the images conform to specific geometric specifications, and provide consistency for subsequent operations such as analysis and model training. Through image segmentation technology, each part of the green plant can be extracted from the standard images, the effective areas can be extracted, and the background noise can be removed, which helps to accurately identify and analyze the structure, features, etc. of the green plants and improve the accuracy of subsequent calculations and automated analysis.

[0016] Preferably, step S2 includes the following steps:

[0017] Step S21: Extract multi-dimensional disease characteristics of the plant parts from the standard plant part images to obtain multi-dimensional disease characteristics of the plant parts;

[0018] Step S22: Identify the lesion areas in the standard plant part images based on the multi-dimensional disease characteristics of the plant parts to generate plant part lesion area identification data;

[0019] Step S23: Separate the lesion backgrounds from the standard plant part images according to the plant part lesion area identification data to obtain plant part lesion area images, and extract the morphological characteristics of the plant part lesion area images to obtain lesion area morphological data;

[0020] Step S24: Discriminate the types of green plant diseases for the plant part lesion area images through the lesion area morphological data to generate green plant disease type discrimination data, where the green plant disease type discrimination data includes minor disease types and similar disease types.

[0021] By extracting the multi-dimensional disease characteristics of green plant parts, the present invention can identify the manifestation forms of diseases in different dimensions (such as color, texture, shape, etc.), providing a comprehensive feature basis for subsequent disease analysis and helping to accurately identify disease types. The identification of the diseased area helps to accurately locate the specific area of the disease on the green plant, ensuring that the image analysis only focuses on the diseased part, rather than the background or healthy area, which improves the accuracy and efficiency of diseased spot detection and avoids false alarms. After the background is separated, the diseased area is clearly focused, reducing the interference of external factors on the analysis results. Extracting the morphological characteristics (such as area, shape, boundary, etc.) of the diseased area provides key indicators for subsequent disease type judgment, making the diagnosis of the diseased spot more accurate. The morphological characteristics of the diseased spot combined with the classification model can effectively distinguish different types of green plant diseases, identify minor diseases and similar disease types, so as to timely discover the severity of the disease and provide specific diagnostic data, providing support for the formulation of disease prevention and control plans and avoiding over-intervention or omission of important diseases.

[0022] Preferably, step S21 includes the following steps:

[0023] Step S211: Construct a gray-level co-occurrence matrix for the standard green plant part image, and use the constructed gray-level co-occurrence matrix to analyze the texture features of the green plant leaf in the standard green plant part image, generating green plant leaf texture feature data;

[0024] Step S212: Reconstruct the standard green plant part image according to the color histogram, generating a color histogram of the green plant part; perform RGB channel analysis on the color histogram of the green plant part, generating RGB channel data; perform HSV color mapping on the standard green plant part image according to the RGB channel data, generating green plant color feature data;

[0025] Step S213: Perform edge detection on the standard green plant part image based on the Sobel operator algorithm, generating edge detection data of the green plant part; calculate the serration degree of the edge detection data of the green plant part to obtain the edge serration degree of the green plant part; extract the structural features of the standard green plant part image according to the edge serration degree of the green plant part, obtaining structural feature data of the green plant part;

[0026] Step S214: Integrate the green plant leaf texture feature data, the green plant color feature data and the structural feature data of the green plant part into the multi-dimensional disease characteristics of the green plant part.

[0027] The present invention can capture the spatial relationship between gray levels in an image through the gray-level co-occurrence matrix, providing detailed texture information, which is very important for identifying subtle texture changes on the leaf surface. Especially in the early stage of disease detection or for minor diseases, it can effectively help identify the diseased areas of the leaves and improve the sensitivity of disease detection. The color histogram provides detailed data on the color distribution in the image, reflecting the hue, saturation, and brightness in the green plant image. The reconstructed image helps to enhance the details of the image, making the color changes caused by diseases more obvious, thereby improving the accuracy of disease detection. Through further RGB channel analysis and HSV color mapping, the subtle differences between the healthy condition of the green plant and the disease characteristics can be more precisely distinguished. The Sobel operator can efficiently extract the edge information in the image, highlighting the boundaries of the diseased areas. Edge features are crucial for the accurate positioning of the diseased areas, especially in cases where the edges are blurred or the shape of the diseased spots is complex, which can help improve the accuracy of diseased spot recognition. The jaggedness calculation helps to evaluate the smoothness of the edge and further analyze whether the edge of the diseased spot has abnormal fluctuations or a jagged shape. The jaggedness analysis can identify the morphological characteristics of the diseased spot, which is particularly important for the classification of diseased spots and the early diagnosis of diseases. Combining with the extraction of structural features can more comprehensively reflect the morphological characteristics of the diseased spot, providing more dimensional data support for the subsequent identification of disease types. By integrating feature data from different dimensions (texture, color, structure), the disease manifestations of green plants can be comprehensively and accurately described. Such multi-dimensional disease characteristics help to improve the accuracy and robustness of disease classification. Especially when faced with multiple different disease types or variant manifestations of diseases, it can effectively reduce misjudgment and missed judgment.

[0028] Preferably, step S24 includes the following steps:

[0029] Step S241: Divide the image of the diseased area of the green plant part into image blocks to generate image blocks of the diseased area of the green plant part; perform wavelet transform on the image blocks of the diseased area of the green plant part through the morphological data of the diseased area to generate a transformed image of the diseased area of the green plant part;

[0030] Step S242: Extract the high-frequency details of the transformed image of the diseased area of the green plant part to obtain the detailed data of the diseased area of the green plant part; use the detailed data of the diseased area of the green plant part to identify minor diseases in the transformed image of the diseased area of the green plant part to generate identification data for minor disease types;

[0031] Step S243: Calculate the occurrence frequency of minor lesion types in the image blocks of the diseased area of the green plant part based on the minor disease type image to obtain the minor lesion frequency data;

[0032] Step S244: When the minor lesion frequency data is greater than 1, the diseased spot image blocks of the green plant part are subjected to hierarchical analysis according to the minor lesion frequency data to generate a hierarchical image set of the diseases of the green plant part; the hierarchical image set of the diseases of the green plant part is subjected to symptom similarity analysis to generate similar disease type recognition data;

[0033] Step S245: The disease type of the diseased spot area image of the green plant part is discriminated by using the minor disease type recognition data and the similar disease type recognition data to generate disease type discrimination data of the green plant part, where the disease type discrimination data of the green plant part includes the minor disease type and the similar disease type.

[0034] Through image block division, the present invention can subdivide the diseased spot area into smaller areas, facilitating the detection and analysis of local lesions. Wavelet transform helps analyze the local features of the image at different scales, especially high-frequency details (such as the edges and textures of diseased spots), and can effectively extract the detail information of the diseased area, improving the sensitivity of disease detection, especially for smaller or weaker diseased spots. The extraction of high-frequency details helps capture the subtle changes of diseased spots, which is crucial for the identification of minor diseases. Through the analysis of detail data, the type of minor disease can be judged more accurately, reducing misdiagnosis and improving the detection accuracy of minor diseases. The minor lesion frequency data helps quantify the occurrence frequency of minor diseases. Frequency analysis not only helps identify the spread trend of diseases but also can provide early warnings for the diseases of green plants. By calculating the occurrence frequency of lesions, it can be identified whether the disease is in the spreading stage, thus providing a timeliness reference for disease control. Hierarchical analysis can refine the different hierarchical features of the diseased area and provide more hierarchical disease information. In areas where minor lesions frequently occur, similar disease types can be identified through similarity analysis, which is very effective for the early diagnosis of diseases and the differentiation of different diseases, reducing misdiagnosis and improving the accuracy of judgment. Combining the recognition results of minor disease types and similar disease types can more accurately discriminate the disease type of the diseased spot area. This step helps accurately judge the type of disease, thus providing strong data support for subsequent disease control measures, avoiding misjudgment and ensuring the pertinence of disease treatment.

[0035] Preferably, step S244 includes the following steps:

[0036] Step S2441: When the minor lesion frequency data is greater than 1, the diseased spot image blocks of the green plant part are subjected to hierarchical analysis according to the minor lesion frequency data to generate a hierarchical image set of the diseases of the green plant part;

[0037] Step S2442: Similar feature extraction is performed on the hierarchical image set of the diseases of the green plant part to obtain similar feature data; according to the similar feature data, the Euclidean distance formula is used to calculate the texture and morphological similarity of the hierarchical image set of the diseases of the green plant part to obtain the first similarity data;

[0038] Step S2443: Calculate the hue and saturation similarity of the hierarchical image set of plant disease lesions according to the similar feature data by using the cosine similarity formula to obtain the second similarity data;

[0039] Step S2444: Construct a disease symptom similarity matrix based on the first similarity data and the second similarity data, and group the diseased spot image patches of the plant part according to the constructed disease symptom similarity matrix to generate grouped image patches of similar disease types;

[0040] Step S2445: Calibrate the disease types of the grouped image patches of similar disease types to generate recognition data of similar disease types.

[0041] Through hierarchical analysis, the present invention can reveal the different hierarchical features of diseased spots in detail, which helps to deeply understand the different development stages of diseases. This step lays a foundation for subsequent similarity analysis and disease classification, making the hierarchical structure of diseases clear, and then improving the accuracy of classification. By extracting similar features, the texture and morphological features of diseased spot images can be captured, especially details such as diseased spot shape and edge. Calculating the similarity of texture and morphology using the Euclidean distance can quantify the similarity between diseased spots and provide quantitative support for subsequent disease classification and recognition. This process helps to exclude irrelevant information and focus on the key features of diseases. The cosine similarity effectively evaluates the color feature differences in the diseased spot area by calculating the hue and saturation similarity of the diseased spot image, which is very useful for distinguishing different types of diseases, especially those with weak color changes. Calculating the hue and saturation similarity helps to enhance the color discrimination ability of diseases and further improve the accuracy of disease recognition. Combining the first and second similarity data to form a more comprehensive disease symptom model through a similarity matrix, this matrix can comprehensively consider multiple aspects such as the texture, morphology, and hue of diseased spots, accurately group diseased spot images with similar features, and thus effectively achieve the classification of diseases. The grouped diseased spot image patches provide a clear division for subsequent calibration of specific disease types, facilitating the rapid identification of disease types. By calibrating the grouped image patches of similar disease types, the specific disease types corresponding to each group can be finally determined. This process not only improves the accuracy of disease recognition but also avoids confusion of disease types, which helps to accurately formulate disease prevention and control measures. The calibration results can provide detailed disease recognition data for agricultural workers and then provide targeted suggestions during the disease prevention and control process.

[0042] Preferably, step S3 includes the following steps:

[0043] Step S31: Continuously analyze the image of the diseased spot area of the plant part based on the slightly diseased type to generate a continuous image comparison set of the diseased area;

[0044] Step S32: Conduct short-term extended anomaly monitoring on green plants with minor lesion types through the continuous image comparison set of the lesion area to generate short-term disease anomaly monitoring data;

[0045] Step S33: Based on similar disease types, conduct dynamic video monitoring on the image of the lesion area of the green plant part to generate long-term anomaly monitoring data for the lesion area;

[0046] Step S34: According to the preset environmental sensors, analyze the short-term disease anomaly monitoring data and the long-term anomaly monitoring data of the lesion area to generate green plant disease transmission trend data, and convert the green plant disease transmission trend data into a green plant disease anomaly risk map through graph theory methods.

[0047] Through continuous image comparison and analysis, the present invention can accurately track the changes in the minor lesion area, thereby identifying the expansion trend of the disease. This step can update the changes in the lesion area in real time, providing data support for subsequent monitoring and early warning. The continuous comparison and analysis can also reveal the subtle changes in the lesion, contributing to early diagnosis and intervention to avoid the spread of the disease. The short-term extended anomaly monitoring can promptly detect the rapid expansion or other abnormal phenomena in the lesion area. By generating short-term disease anomaly monitoring data, it can quickly respond to disease changes and provide early disease prevention and control plans, which is crucial for preventing the rapid spread of the disease, especially in the initial stage of the lesion, effectively reducing losses. Through dynamic video monitoring, the long-term changes and trends in the lesion area can be captured. Compared with short-term monitoring, long-term monitoring can comprehensively reflect the evolution process of the disease, helping to predict the performance of the disease under different seasons and environmental conditions. The long-term anomaly monitoring data provides more in-depth insights for formulating disease prevention and control strategies. Especially for seasonal or long-term developing diseases, dynamic monitoring provides the necessary information support. By comprehensively analyzing the short-term and long-term anomaly monitoring data, the transmission risk of the disease can be comprehensively evaluated. Using graph theory methods to generate a green plant disease anomaly risk map can intuitively display the transmission path and risk area of the disease, providing decision-making support for farm managers. This risk map provides a basis for precise prevention and control, helping to formulate personalized disease prevention and control strategies, reducing the risk of disease spread, and improving the efficiency and effectiveness of disease management.

[0048] Preferably, analyzing the short-term disease anomaly monitoring data and the long-term anomaly monitoring data of the lesion area according to the preset environmental sensors includes:

[0049] Collect the temperature and humidity data of the green plant lesion according to the preset temperature and humidity sensors for the short-term disease anomaly monitoring data and the long-term anomaly monitoring data of the lesion area to obtain the temperature and humidity data of the green plant lesion area;

[0050] Perform time series analysis on the temperature and humidity data of the diseased area of green plants to generate the temperature and humidity change data of the diseased area of green plants; confirm the temperature and humidity change area of the green plant disease based on the temperature and humidity change data of the diseased area of green plants, and use a multispectral imager to perform regional spectral imaging to generate a regional spectral imaging map;

[0051] Map the temperature and humidity data of the diseased area of green plants to the spectral band of the regional spectral imaging map to generate the spectral band data of the diseased area of green plants; perform directional change analysis on the spectral band data of the diseased area of green plants to generate the propagation wind direction data of the diseased area of green plants;

[0052] Use the propagation wind direction data of the diseased area of green plants to predict the propagation trend of short-term disease anomaly monitoring data and long-term anomaly monitoring data of the diseased area, and generate the green plant disease propagation risk data.

[0053] The present invention can help accurately understand the growth environment of diseases by collecting and monitoring the temperature and humidity data of the diseased area. Through the collection of temperature and humidity data, the environmental conditions of the diseases can be understood in real time, providing basic data for subsequent propagation trend analysis. Time series analysis can reveal the changing trend of temperature and humidity over time, helping to identify the seasonal characteristics of disease occurrence and the fluctuation law of temperature and humidity. Through this analysis, the growth pattern of diseases under specific temperature and humidity conditions can be predicted, providing a scientific basis for further propagation trend prediction. The multispectral imager can capture the reflection characteristics of plants in different spectral bands, helping to identify the subtle changes in the diseases of green plants under different temperature and humidity conditions. Through regional spectral imaging, a high-resolution image of the diseased area can be obtained, revealing the spatial distribution characteristics of the diseases. This technology provides a visual basis for further analysis of the diseased area. Combining the temperature and humidity data with the spectral imaging map can provide more detailed environmental information for the diseased area. Through the spectral band data, the health status of plant leaves can be analyzed in depth, and the spectral characteristics of different diseases can be identified. This step helps to comprehensively understand the propagation and development process of diseases from two dimensions of the environment and the diseased state. Through the directional change analysis of the spectral band data, the directional characteristics of disease propagation can be captured. Wind direction and climatic conditions are key factors affecting disease propagation. By obtaining wind direction data, the propagation path of diseases can be predicted, providing strong support for precise prevention and control. Using the propagation wind direction data to predict the disease propagation trend can accurately judge the expansion direction and risk area of the diseases. Through the combined analysis of short-term and long-term monitoring data, the propagation speed and range of diseases can be dynamically evaluated, generating predictive data, and providing timely warnings and prevention and control measures for agricultural producers.

[0054] Preferably, step S4 includes the following steps:

[0055] Step S41: Use the green plant disease abnormal risk map to conduct early warning of the green plant lesion risk for the image of the lesion area of the green plant part, and generate green plant lesion risk early warning data;

[0056] Step S42: Construct a prevention and control plan based on the green plant lesion risk early warning data to generate a green plant lesion prevention and control plan; visualize the data of the green plant lesion prevention and control plan to execute the green plant disease prevention and control management operation.

[0057] Through the early warning of the risk of the image of the lesion area, the present invention can monitor the risk level of the disease in real time. This early warning mechanism helps to identify potential disease threats in advance and prevent the further spread of the disease. The timely lesion risk early warning data provides effective decision-making support for farmers and managers, enabling them to take appropriate countermeasures at the initial stage of the disease development. Constructing a prevention and control plan based on the risk early warning data can adopt personalized prevention and control measures for different types of diseases. This plan combines real-time monitoring data, the types of diseases, the distribution range and the risk degree, providing an accurate plan for the prevention and control of green plant diseases. Through this data-driven prevention and control strategy, the prevention and control efficiency can be greatly improved and the losses caused by diseases can be reduced. Data visualization can transform the complex prevention and control plan into intuitive and easy-to-understand charts or maps, enabling farmers and managers to quickly understand and apply these plans. The visualization tool not only improves the execution efficiency of the prevention and control plan, but also enhances the operability of the green plant disease management. Through the real-time updated visualization platform, farmers can obtain the latest status of disease management at any time, quickly respond and adjust the prevention and control strategy to ensure the timeliness and accuracy of disease prevention and control.

[0058] In this specification, a green plant disease identification and early warning management system is provided for implementing the above-mentioned green plant disease identification and early warning management method. The green plant disease identification and early warning management system includes:

[0059] A part segmentation module, configured to obtain a high-definition image of a green plant; perform image preprocessing on the high-definition image of the green plant to generate a standard high-definition image of the green plant; perform green plant part image segmentation and integration on the standard high-definition image of the green plant to obtain a standard green plant part image;

[0060] A disease type identification module, configured to extract multi-dimensional disease features of the green plant part from the standard green plant part image to obtain multi-dimensional disease features of the green plant part; identify the lesion area of the standard green plant part image based on the multi-dimensional disease features of the green plant part to generate green plant part lesion area identification data; discriminate the green plant disease type for the standard green plant part image according to the green plant part lesion area identification data to generate green plant disease type discrimination data, where the green plant disease type discrimination data includes minor disease types and similar disease types;

[0061] A disease transmission analysis module, which is used to analyze the green plant disease transmission trend of green plants with minor lesion types based on minor disease types and similar disease types, generate green plant disease transmission risk data, and convert the green plant disease transmission risk data into a green plant disease abnormal risk map through graph theory methods;

[0062] A risk warning module, which is used to use the green plant disease abnormal risk map to perform green plant lesion risk warning on the image of the lesion area of the green plant part, generate green plant lesion risk warning data; construct a prevention and control plan according to the green plant lesion risk warning data to execute the green plant disease prevention and control management operation.

[0063] Through high-definition image acquisition and preprocessing, the present invention ensures the quality and consistency of the images, providing an accurate basis for subsequent processing. Image segmentation and integration isolate each part of the green plant, making subsequent disease diagnosis more accurate, being able to clearly identify and analyze the disease manifestations of each part, and avoiding confusion and misjudgment. Through the extraction of multi-dimensional disease characteristics of the green plant parts, the various manifestations of the disease can be comprehensively and accurately analyzed, improving the accuracy of disease diagnosis. The identification of the lesion area can accurately locate the disease area and classify it according to different disease types, reducing misjudgments of minor diseases and similar diseases, ensuring the accurate determination of the disease type, and providing a scientific basis for subsequent disease prevention and control. Through the analysis of the transmission trends of minor disease types and similar disease types, the risk of disease spread can be predicted in advance, potential high-risk areas can be identified, and green plant disease transmission risk data can be generated. This method enables disease prevention and control to shift from static diagnosis to dynamic warning, thus effectively reducing the spread of diseases. Using the disease abnormal risk map to warn of the green plant lesion risk can timely detect potential risk areas before the disease spreads, provide real-time feedback and early intervention. The prevention and control plan constructed based on these warning data has a high degree of accuracy, making disease prevention and control management more efficient and timely, reducing resource waste and improving the effect of disease control. Therefore, the present invention improves the accuracy and reaction timeliness of green plant disease identification and warning by introducing high-definition image processing, multi-dimensional feature extraction, transmission trend analysis and dynamic warning. Brief Description of the Drawings

[0064] Figure 1 It is a schematic diagram of the step flow of a method for identifying and warning the management of green plant diseases;

[0065] Figure 2 For Figure 1 a detailed implementation step flow diagram of step S2 in

[0066] Figure 3 For Figure 1 a detailed implementation step flow diagram of step S3 in

[0067] The realization, functional features and advantages of the objectives of the present invention will be further described in conjunction with the embodiments with reference to the accompanying drawings. Specific Embodiments

[0068] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0069] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus the repeated description thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0070] It should be understood that although the terms "first", "second", etc. may be used here to describe each unit, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be called the second unit, and similarly the second unit may be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed related items.

[0071] To achieve the above object, please refer to Figures 1 to 3 , a method for identifying and warning management of green plant diseases, the method comprising the following steps:

[0072] Step S1: Obtain a high-definition image of a green plant; perform image preprocessing on the high-definition image of the green plant to generate a standard high-definition image of the green plant; perform image segmentation and integration on the standard high-definition image of the green plant to obtain a standard image of the green plant part;

[0073] Step S2: Extract multi-dimensional disease features of the green plant part from the standard image of the green plant part to obtain multi-dimensional disease features of the green plant part; identify the diseased area of the standard image of the green plant part based on the multi-dimensional disease features of the green plant part to generate identification data of the diseased area of the green plant part; determine the type of green plant disease based on the identification data of the diseased area of the green plant part to generate discrimination data of the type of green plant disease, wherein the discrimination data of the type of green plant disease includes minor disease types and similar disease types;

[0074] Step S3: Based on the minor disease types and similar disease types, analyze the transmission trend of the minor lesion types of green plants, generate the green plant disease transmission risk data, and convert the green plant disease transmission risk data into a green plant disease abnormal risk map through graph theory methods;

[0075] Step S4: Use the green plant disease abnormal risk map to conduct a green plant lesion risk warning on the image of the lesion area of the green plant part, and generate the green plant lesion risk warning data; construct a prevention and control plan based on the green plant lesion risk warning data to execute the green plant disease prevention and control management operation.

[0076] The present invention ensures the standardization of the image quality by obtaining high-definition images and performing preprocessing, avoiding the influence of low-quality images on the recognition effect. Image segmentation and integration further improve the accuracy of disease recognition, making subsequent processing more targeted and accurate. By extracting multi-dimensional disease features of the green plant parts, the disease manifestations can be comprehensively analyzed, and different types of lesion areas can be identified. The discrimination of disease types not only improves the accuracy of disease diagnosis but also distinguishes minor diseases and similar disease types, reducing the possibility of misdiagnosis and missed diagnosis. By analyzing the transmission trends of minor disease types and similar disease types, the spread of diseases can be predicted, and potential high-risk areas can be identified in advance. The graph theory method converts the transmission risk data into a green plant disease abnormal risk map, which helps to intuitively display the spread trend of diseases and provides a scientific basis for formulating prevention and control strategies. By using the disease abnormal risk map to warn of the green plant lesion risk, problems can be identified in time before the spread of diseases, avoiding further spread of diseases. Constructing a prevention and control plan based on the warning data can achieve precise prevention and control management, improve the efficiency and effect of disease control, and reduce resource waste. Therefore, the present invention improves the accuracy and response timeliness of green plant disease recognition and warning by introducing high-definition image processing, multi-dimensional feature extraction, transmission trend analysis, and dynamic warning.

[0077] In the embodiment of the present invention, refer to Figure 1 As shown, it is a schematic flow chart of the steps of a green plant disease recognition and warning management method of the present invention. In this example, the green plant disease recognition and warning management method includes the following steps:

[0078] Step S1: Obtain a high-definition image of a green plant; perform image preprocessing on the high-definition image of the green plant to generate a standard high-definition image of the green plant; perform image segmentation and integration on the standard high-definition image of the green plant to obtain a standard image of the green plant part;

[0079] In the embodiments of the present invention, an image of a green plant is captured by using a high-resolution imaging device (such as a digital camera, a mobile phone or a professional image acquisition tool) to ensure that the image is clear and distortion-free, and the influence of light changes or shadows during shooting is minimized. To obtain high-quality images, shooting can be performed under natural light to avoid overexposure or loss of details in dark areas. The image is denoised by using a filter (such as Gaussian filter) to reduce the noise generated due to interference during the shooting process. By adjusting the brightness and contrast of the image, the details of the green plant are made more prominent, the sundries in the background are made more blurred, and the saliency of the green plant area is enhanced. If the color deviation of the image is large, color correction can be performed to ensure that the natural color tone of the green plant is retained. The preprocessed image is normalized, including unifying the size, resolution and color mode of the image (such as RGB mode or HSV mode), to generate a unified standard high-definition image of the green plant. The image size can be adjusted to a consistent pixel size (for example, 1024x1024), and the pixel density of the image is ensured to meet the requirements of subsequent image segmentation and analysis. An image segmentation algorithm (such as U-Net, threshold segmentation, region growing or a deep learning model) is used to segment the green plant area in the green plant image. The goal of segmentation is to extract parts such as the leaves and branches of the green plant from the image. If a deep learning method is used, a semantic segmentation model can be trained to identify different types of green plant parts (such as leaves, stems, roots, etc.). Post-processing is performed on the segmentation result, such as connected region analysis, noise removal, etc., to further clean up errors or small irrelevant regions in the image. Each part of the segmented green plant (such as leaves, stems, etc.) is integrated according to its spatial position and semantic information to form a final standard image of the green plant part. During integration, it is ensured that the boundaries between the parts are clear and the sizes of the part images are consistent.

[0080] Step S2: Extract multi-dimensional disease characteristics of the parts from the standard image of the green plant parts to obtain multi-dimensional disease characteristics of the green plant parts; identify the diseased area of the standard image of the green plant parts based on the multi-dimensional disease characteristics of the green plant parts to generate identification data of the diseased area of the green plant parts; determine the type of green plant disease for the standard image of the green plant parts according to the identification data of the diseased area of the green plant parts to generate discrimination data of the type of green plant disease, where the discrimination data of the type of green plant disease includes minor disease types and similar disease types;

[0081] In the embodiments of the present invention, by analyzing the color information of the images of the green plant parts, especially the color changes in the diseased area (such as yellow, brown, black spots, etc.). Color features can be extracted through color space conversion (such as from RGB to HSV or Lab). There are usually obvious color changes in the diseased area. By calculating features such as the mean, variance, and chromaticity of the color, it can provide support for subsequent diseased spot recognition. Use texture analysis algorithms (such as gray-level co-occurrence matrix (GLCM), local binary pattern (LBP), etc.) to extract the texture features of the images of the green plant parts. The texture of the diseased area is usually different from that of the healthy part. The texture features can help identify different types of diseases, such as spots, rots, or vein diseases, etc. The diseased area often has irregular or specific shape features. Extract the shape features of the diseased area in the image through edge detection (such as Canny edge detection) or contour extraction algorithms (such as boundary-based contour analysis), and analyze parameters such as the size, shape, and edge smoothness of the diseased spot. Multidimensional integration of the extracted color, texture, and shape features can be performed using dimensionality reduction methods such as principal component analysis (PCA), linear discriminant analysis (LDA), etc., to obtain the multi-dimensional disease features of the green plant parts. These features will provide a basis for subsequent diseased area recognition and disease type discrimination. According to the extracted multi-dimensional disease features (such as color, texture, shape, etc.), apply image segmentation techniques (such as threshold segmentation, region growing, edge detection, deep learning, etc.) to identify the diseased areas in the images of the green plant parts. Image learning-based methods (such as convolutional neural network (CNN) or U-Net network) can be used to automatically detect the diseased areas. By training the model, the diseased areas existing on different green plant parts are identified. During the diseased area recognition process, use post-processing methods (such as morphological processing, region connection analysis) to remove small noise areas and refine the diseased spot boundaries. This step can improve the recognition accuracy of the diseased area and ensure the accuracy of the recognition results. Store the identified diseased areas in data form to form the recognition data of the diseased areas of the green plant parts. This data can include information such as the location, size, shape, and color of the diseased spots, which is convenient for subsequent analysis. Store the identified diseased areas in data form to form the recognition data of the diseased areas of the green plant parts. This data can include information such as the location, size, shape, and color of the diseased spots, which is convenient for subsequent analysis. Perform type discrimination on each diseased area based on the trained model. The results output by the model should include the type of the disease and the severity of the disease (such as mild, severe). The green plant disease type discrimination data includes: The model judges that the diseased spot belongs to the type of mild disease (such as mild spot disease, initial downy mildew, etc.) according to the features such as the size, color, and distribution of the diseased spot. According to the similarity between different disease types (such as the similarity in morphology, color, texture, etc.), judge that some diseased spots belong to similar disease types, which is convenient for further refined diagnosis.The plant disease classification model comprehensively discriminates the disease types of the images of green plant parts and generates disease type discrimination data, which include the types of each lesion (such as leaf spot disease, downy mildew, etc.) and its severity (such as mild or similar types). Probability values can be used to represent the confidence levels of different types of diseases, facilitating further analysis and intervention. By comparing the manually labeled lesion area data with the results output by the model, the accuracy of lesion area recognition and disease type discrimination is verified. Methods such as cross-validation and confusion matrices can be used to evaluate the performance of the model. According to the verification results, the classification model is optimized and adjusted, such as increasing training data, improving feature extraction methods, adjusting model parameters, etc., to improve classification accuracy.

[0082] Step S3: Based on the mild disease types and similar disease types, analyze the propagation trend of green plant diseases for green plants with mild lesion types, generate green plant disease propagation risk data, and convert the green plant disease propagation risk data into a green plant disease abnormal risk map through graph theory methods;

[0083] In the embodiments of the present invention, a disease transmission model is established, which is based on the transmission characteristics of minor disease types and similar disease types. Common transmission models include: the SIR model (Susceptible-Infected-Recovered model) used to describe the process of disease transmission from susceptible green plants to infected green plants and then to recovery or disappearance. This model can adjust parameters such as the infection rate and recovery rate according to the transmission characteristics of different disease types. If the green plants are distributed in a certain spatial area, the green plants can be regarded as the nodes of a graph, and the disease transmission as the edges in the graph. By defining the transmission probability, disease transmission simulation is carried out based on the spatial positions and neighboring relationships of the green plants. Based on the distance between green plants and contact situations (such as transmission methods like wind and insects), the transmission probability between adjacent green plants is calculated to predict the disease transmission trend. Through historical data or experimental data, the transmission speed, direction, and intensity of minor diseases and similar diseases are evaluated. The future spread trend of diseases can be predicted through time series analysis (such as the ARIMA model), or high-risk areas of disease transmission can be determined through spatial analysis (such as hotspot analysis). Based on the above transmission model and trend analysis, the disease transmission risks of different green plants are calculated. The risk data should include the following: the rate of disease transmission from one green plant to another, the preferred direction of disease transmission, affected by factors such as wind speed and green plant density, and the degree of disease transmission per unit time, such as the number of newly infected green plants per unit area. Regarding the green plants as the nodes of a graph and the contact or transmission possibility between green plants as the edges in the graph. According to information such as the distance between green plants, contact frequency, and physical environment factors (such as air flow, temperature, and humidity), weights are established for each pair of green plant nodes to represent the probability or intensity of transmission between them. For example, if two green plants are very close and the wind is strong, the transmission probability between them can be set with a higher weight. Taking the disease transmission risk data of each green plant as the attribute of the node to describe the disease transmission risks (such as infection risk, disease development speed, etc.) faced by the node (green plant). Using graph theory methods (such as the adjacency matrix of the graph, edge weights of the graph), the transmission risk data between green plants is mapped to the edge weights of the graph, and a map of the green plant disease transmission network is generated. "Risk hotspot" areas can be generated in the map according to the disease transmission trend, and the green plants in these areas have a higher infection risk or transmission potential. For these high-risk areas, more precise prevention and control measures can be taken. Using centrality measures in graph theory (such as degree centrality, closeness centrality, betweenness centrality, etc.) to analyze the most critical green plant nodes in the disease transmission network. High centrality nodes usually represent the "transmission sources" or "transmission centers" of disease transmission, and these green plants have an important impact on the overall disease transmission. Degree centrality can reflect the number of neighboring green plants directly connected to each green plant. Nodes with high degree centrality are the main nodes of disease transmission. Betweenness centrality indicates the importance of a node in the shortest path between two nodes, and green plants with high betweenness centrality are the "bridges" of disease transmission.Closeness centrality measures how close a node is to other nodes. Green plants with high closeness centrality are prone to quickly spreading diseases. Graph theory algorithms (such as the shortest path algorithm, minimum spanning tree, etc.) are used to analyze the disease transmission paths. Potential disease transmission chains can be identified, thus providing effective early warning and intervention strategies for disease control. For example, if a disease transmission path extends from a high-risk area to a low-risk area, taking isolation measures in the early stage can effectively reduce the spread of the disease. Using connectivity analysis, determine whether the disease transmission network is fully connected and whether there are isolated disease transmission areas. According to the results of the connectivity analysis, local isolation measures can be taken to prevent the disease from spreading to other areas. Combine the transmission risk map with graph theory indicators such as the centrality of nodes, transmission paths, and connectivity to form the final abnormal risk map of green plant diseases. Through the identification of different colors, sizes, and shapes in the map, visually display the disease transmission risk level of each green plant. For example, high-risk green plant nodes can be represented by red, and low-risk green plants can be represented by green. This map can be used as a decision support tool to provide effective strategy references for green plant disease prevention and control work.

[0084] Step S4: Use the abnormal risk map of green plant diseases to conduct early warning of the lesion risk of the diseased spot area of the green plant part, and generate early warning data for the green plant lesion risk; construct a prevention and control plan according to the early warning data for the green plant lesion risk to execute the green plant disease prevention and control management operation.

[0085] In the embodiments of the present invention, by combining the abnormal risk map of green plant diseases with the image data of the lesion area as the model input. The feature data extracted from the image of the lesion area (such as the area, shape, color change of the lesion, etc.) is matched with the risk level in the risk map. A model based on deep learning (such as convolutional neural network) or machine learning (such as random forest, support vector machine, etc.) is used to predict the lesion risk. The model learns the relationship between the image features of the green plant lesion area and the disease transmission risk through training historical data, and generates lesion risk warning data. The CNN model can effectively extract the spatial features of the lesion area image, and comprehensively analyze by combining the risk data in the disease transmission risk map to predict the potential risk of disease development. Using a regression model, combining the disease transmission trend data of green plants with the attributes of the lesion image, to predict the future lesion risk of each green plant part. Classify the predicted lesion risks by green plant part to generate green plant lesion risk warning data. The warning data includes: classified according to the intensity of the warning, into low-risk, medium-risk and high-risk areas, predicting whether the disease will spread rapidly or slow down in the future, and giving the warning time window (for example: 1 day, 3 days, 7 days, etc.), and accurately positioning the location where the disease occurs according to the image of the lesion area of the green plant part. Use a graphical interface or map to display the lesion risk warning data. High-risk areas can be marked in red, and low-risk areas are marked in green or yellow, which is convenient for managers to quickly identify key prevention and control areas. It is possible to combine the green plant distribution map or growth model to form a real-time risk map of green plants to dynamically monitor risk changes. According to the warning data, generate a detailed green plant lesion risk warning report, and the report content includes: the location of the warned green plant, the lesion risk level, the disease type, the expected transmission speed, etc., to provide decision-making support for subsequent prevention and control work. For green plants in high-risk areas, prioritize the design of enhanced prevention and control plans. Include physical control (such as isolation, pruning, disinfection), chemical control (such as spraying pesticides, fungicides), biological control (such as introducing beneficial insects, microbial control). For medium-risk areas, formulate relatively conservative prevention and control measures. For example, use low-toxicity, broad-spectrum plant protectants for prevention to reduce losses. For low-risk areas, adopt the method of observation and tracking, regularly check the changes of lesions, and avoid waste of resources caused by over-prevention. According to the discriminant data of green plant disease types (such as minor disease types, similar disease types, etc.), combined with the risk warning data, design special prevention and control plans for different types of diseases. For example, for fungal diseases, specific fungicides can be selected for spraying, while for insect pests, insecticides or physical insect-catching equipment are required. Considering the transmission speed and diffusion trend of the disease (analysis based on the transmission model), dynamically adjust the prevention and control plan. If the transmission speed of a certain disease is too fast, it is necessary to increase the frequency of monitoring and prevention and control, or change the prevention and control strategy. Establish a green plant disease monitoring system to obtain the health status of green plants and the prevention and control effect in real time. Through technical means such as drones, sensors or remote sensing images, monitor the prevention and control effect in real time.If it is found that the prevention and control measures fail to achieve the expected effect, adjust the prevention and control strategy in a timely manner. Continuously optimize the prevention and control plan according to the actual changes in the lesion risk. For example, if the spread rate of the disease in a certain area is faster than expected, the prevention and control measures can be strengthened, or more powerful agents can be used. According to the green plant lesion risk warning data and the prevention and control plan, formulate a detailed task allocation table for the prevention and control of green plant diseases, clarifying information such as the responsible person, operation time, and operation location. The tasks can be assigned to relevant personnel or teams through a management system or application program, and the implementation status can be updated in real time through mobile devices.

[0086] Preferably, step S1 includes the following steps:

[0087] Step S11: Obtain high-definition images of green plants using a high-definition digital camera;

[0088] Step S12: Perform image filtering on the high-definition images of green plants to generate high-definition filtered images of green plants;

[0089] Step S13: Perform image brightness enhancement on the high-definition filtered images of green plants to generate high-definition brightness-enhanced images of green plants;

[0090] Step S14: Perform image geometric transformation on the high-definition brightness-enhanced images of green plants to generate standard high-definition images of green plants; perform image segmentation and integration of the green plant parts on the standard high-definition images of green plants to obtain standard green plant part images.

[0091] In the embodiments of the present invention, a high-resolution digital camera (such as a DSLR or a professional camera) is used to photograph green plants. Ensure sufficient lighting and clear and unobstructed green plants during image acquisition. Appropriate camera settings (such as exposure time, white balance, and aperture) can be selected to ensure the acquisition of details and colors of green plants. Apply image filtering techniques, such as Gaussian filtering or median filtering, to remove noise in the image. This step can be achieved by passing the image through a convolutional filter, which helps to smooth the image, remove unnecessary details, and make the main features of green plants more prominent. Use brightness enhancement algorithms (such as histogram equalization, contrast stretching, or gamma correction) to improve the brightness and contrast of the image, which helps to make the colors of green plants more vivid and clear. Especially in darker areas of the image, the details of green plants can be enhanced. According to actual needs, perform geometric transformations (such as rotation, scaling, cropping, etc.) to make the composition of the image more in line with standardized requirements. For example, the direction of green plants can be kept consistent by rotating the image, or the green plants can be centered in the image by cropping the image. Use image segmentation algorithms (such as threshold segmentation, k-means clustering, deep learning segmentation methods) to segment the green plant parts in the image and extract the specific areas of green plants. Then, the segmented green plant parts are integrated into a standard green plant part image to highlight the main parts of green plants and remove the background or irrelevant areas.

[0092] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0093] Step S21: Extract multi-dimensional disease characteristics of the green plant part from the standard green plant part image to obtain the multi-dimensional disease characteristics of the green plant part;

[0094] Step S22: Identify the lesion area of the standard green plant part image based on the multi-dimensional disease characteristics of the green plant part to generate the identification data of the lesion area of the green plant part;

[0095] Step S23: Separate the lesion background of the standard green plant part image according to the identification data of the lesion area of the green plant part to obtain the lesion area image of the green plant part, and extract the morphological characteristics of the lesion area image of the green plant part to obtain the morphological data of the lesion area;

[0096] Step S24: Determine the type of green plant disease for the lesion area image of the green plant part through the morphological data of the lesion area to generate the discrimination data of the type of green plant disease, where the discrimination data of the type of green plant disease includes minor disease types and similar disease types.

[0097] In an embodiment of the present invention, by collecting images of standard parts of green plants, these images need to have a high resolution to ensure that the disease features can be accurately captured. Use image processing algorithms (such as color histograms, texture features, morphological filtering, etc.) to extract multiple dimensional features in the image. A variety of methods can be used, such as feature extraction based on regional growth or feature learning methods based on convolutional neural networks (CNNs). The extracted disease features are calibrated as multidimensional disease features to form a data set containing information such as color, texture, and shape. Using the disease features extracted in S21, an image classification model or a semantic segmentation model is trained. The model can use a convolutional neural network (CNN) or other deep learning methods. The standard green plant part image is input into the trained model to automatically identify the diseased area. The model will separate the diseased area from the healthy part by analyzing the features in the image, and generate green plant part diseased area identification data, in which the specific location and morphology of the diseased spot are identified. According to the diseased area identified in S22, segmentation techniques (such as image thresholding, edge detection, etc.) are used to separate the diseased spot from the background. Morphological operations (such as opening and closing operations) in image processing can be used to remove background noise. For the separated lesion area image, the geometric morphological features of the lesions, including area, perimeter, shape factor, etc., are extracted using morphological analysis methods. By analyzing the morphological features of the lesion area, morphological data of the lesion area are obtained, including the size, shape, edge smoothness and other features of the lesions. Based on the lesion morphological data extracted in S23, a classification model is trained. This model can be a traditional machine learning model (such as support vector machine, decision tree) or a deep learning model (such as CNN). The lesion area morphological data is input into the trained disease type discrimination model. The model will classify the lesions, determine the disease type to which they belong, and generate green plant disease type discrimination data, including different types of diseases (such as minor disease types) and similar disease types, for further disease management and prevention measures.

[0098] Preferably, step S21 includes the following steps:

[0099] Step S211: constructing a grayscale symbiosis matrix for the standard green plant part image, and using the constructed grayscale symbiosis matrix to perform green plant leaf texture feature analysis on the standard green plant part image to generate green plant leaf texture feature data;

[0100] Step S212: reconstructing the image of the standard green plant part according to the color histogram to generate a green plant part color histogram; performing RGB channel analysis on the green plant part color histogram to generate RGB channel data; performing HSV color mapping on the standard green plant part image according to the RGB channel data to generate green plant color feature data;

[0101] Step S213: Perform edge detection on the standard green plant part image based on the Sobel operator algorithm to generate green plant part edge detection data; calculate the serration degree of the green plant part edge detection data to obtain the green plant part edge serration degree; extract the structural features of the standard green plant part image according to the green plant part edge serration degree to obtain the green plant part structural feature data;

[0102] Step S214: Integrate the green plant leaf texture feature data, the green plant color feature data, and the green plant part structural feature data into the green plant part multi-dimensional disease features.

[0103] In the embodiments of the present invention, by performing grayscale processing on the standard green plant part image, the color image is converted into a grayscale image to reduce the interference of color information and focus on texture features. Using the Gray-Level Co-Occurrence Matrix (GLCM) technique, by calculating the gray-level correlation between image pixels, co-occurrence matrices in multiple directions (such as 0°, 45°, 90°, 135°) and different distances (such as 1, 2 pixels, etc.) are obtained. According to the co-occurrence matrix, multiple texture features are extracted, such as contrast, homogeneity, energy, entropy, etc. Through these texture features, the subtle changes on the leaf surface can be captured. The extracted texture feature data is converted into green plant leaf texture feature data, which serves as the basis for subsequent analysis. Construct a color histogram of the standard green plant part image. This step reconstructs the color characteristics of the green plant by calculating the frequency distribution of each color channel (red, green, blue) in the image. Calculate the color histogram of the RGB channels to generate the color histogram of the green plant part, which describes the color distribution of the image. Analyze the red, green, and blue (RGB) channels separately, and extract statistical features such as the mean and variance of each channel, which helps to describe the color performance of the green plant on different channels. Convert the RGB image into the HSV color space because the HSV color space is more consistent with the way humans perceive colors and can effectively reflect the hue, saturation, and brightness of the image. Further analyze the three components of H (hue), S (saturation), and V (brightness) in the HSV space, and extract their statistical features, such as the mean and variance. According to the HSV color mapping, generate the green plant color feature data, and combine the analysis results of the RGB channels to provide a more comprehensive color description. Use the Sobel operator to perform edge detection on the standard green plant part image, calculate the gradient of the image, and obtain the edge information of the image. The Sobel operator will highlight the boundaries and lesion areas of the green plant. The result of edge detection is a binary edge image that identifies the boundaries of the green plant leaves or lesions. Analyze the result of edge detection and calculate the serration degree of the edge. The serration degree refers to the roughness of the edge and can be used to describe the edge characteristics of the lesion area. A high serration degree usually indicates obvious boundary irregularities of the leaf lesion. Based on the edge data extracted by the Sobel operator, extract the structural features of the green plant part, such as the contour and shape of the leaf and the edge characteristics of the lesion. Through this information, further understand the geometric characteristics of the green plant leaves and lesions. Integrate the green plant leaf texture features extracted in S211, the color features in S212, and the structural features in S213 into a multi-dimensional feature dataset. This dataset comprehensively reflects multiple aspects of the green plant, such as texture, color, and structure. By means of weighted average or feature splicing, each feature is combined into a multi-dimensional feature vector as the input for further disease identification. Finally, the generated multi-dimensional disease features of the green plant part will be used in subsequent steps such as lesion area identification and type discrimination.

[0104] Preferably, step S24 includes the following steps:

[0105] Step S241: Divide the image of the diseased area of the green plant part into image blocks to generate image blocks of the diseased area of the green plant part; perform wavelet transform on the image blocks of the diseased area of the green plant part through the morphological data of the diseased area to generate a transformed image of the diseased area of the green plant part;

[0106] Step S242: Extract the high-frequency details of the transformed image of the diseased area of the green plant part to obtain the detail data of the diseased area of the green plant part; use the detail data of the diseased area of the green plant part to identify minor diseases in the transformed image of the diseased area of the green plant part to generate identification data of minor disease types;

[0107] Step S243: Calculate the occurrence frequency of minor lesion types of the image blocks of the diseased area of the green plant part based on the images of minor disease types to obtain minor lesion frequency data;

[0108] Step S244: When the minor lesion frequency data is greater than 1, perform hierarchical analysis on the image blocks of the diseased area of the green plant part according to the minor lesion frequency data to generate a set of hierarchical images of the diseases of the green plant part; perform analysis of symptom similarity on the set of hierarchical images of the diseases of the green plant part to generate identification data of similar disease types;

[0109] Step S245: Use the identification data of minor disease types and the identification data of similar disease types to discriminate the types of green plant diseases in the image of the diseased area of the green plant part to generate discrimination data of green plant disease types, where the discrimination data of green plant disease types includes minor disease types and similar disease types.

[0110] In the embodiments of the present invention, the diseased area image of the green plant part is divided into regular or adaptive blocks (such as dividing into image blocks of a fixed size or using the region growing method). Each image block should contain a certain amount of diseased spot information for subsequent transformation and analysis. Each small block after division will be used as an independent processing unit. The image blocks can be marked according to the image content, such as marking the diseased spot or non-diseased spot areas. Apply wavelet transform to the diseased spot image blocks, using methods such as discrete wavelet transform (DWT) or two-dimensional wavelet transform to extract the multi-scale features of the image. Wavelet transform can decompose the image at different scales and extract the detailed information and general outline. Generate the diseased spot transform images through wavelet transform, and these transform images can help reveal the high-frequency details and low-frequency background information in the image. Specifically, different analyses can be performed on the low-frequency part (approximation image) and the high-frequency part (detail image). Extract the high-frequency part of the image from the wavelet transform, which usually reflects the edge and detail information in the image. Analyze the high-frequency part of the wavelet of the diseased area image of the green plant to reveal the subtle diseased spot morphology and texture features. Extract these high-frequency detail information into a feature data set for further analysis. According to the extracted detail data, identify minor diseases through a pre-trained classification model (such as SVM, CNN, etc.). This step aims to identify the early or minor disease types of the diseased spots through the detail image. Convert the results of minor disease identification into minor disease type identification data, and mark the diseased area as a minor disease or a normal area. According to the results of minor disease type identification, calculate the number of times of minor lesions in each diseased spot image block. Count the occurrence frequency of the area identified as a minor disease in the image block. Generate minor lesion frequency data by counting the number of minor lesions in each image block, which is used as the basis for the next hierarchical analysis. When the minor lesion frequency data is greater than 1, it indicates that the image block contains areas with multiple minor lesions. At this time, adopt the hierarchical analysis method to hierarchically divide the diseased spot image blocks according to the lesion frequency. Classify these image blocks with higher frequencies into different levels to generate multiple disease hierarchical image sets. Perform similarity analysis on the generated disease hierarchical image sets, using methods such as structural similarity index (SSIM), normalized mutual information (NMI), etc., to calculate the similarity of different diseased areas. Based on the similarity analysis, identify the areas with similar disease types and classify them into similar disease types. Utilize the minor disease type identification data identified in step S242 and the similar disease type identification data analyzed in step S244, and through comprehensive analysis, determine the disease type of the green plant part. Specifically, by fusing the discrimination results of the two and selecting a suitable model (such as decision tree, logistic regression, etc.) for the final discrimination. Finally, generate the green plant disease type discrimination data, marked as minor disease type and similar disease type.

[0111] Preferably, step S244 includes the following steps:

[0112] Step S2441: When the minor lesion frequency data is greater than 1, perform hierarchical analysis on the lesion image blocks of the green plant part according to the minor lesion frequency data to generate a hierarchical image set of diseases of the green plant part;

[0113] Step S2442: Extract similar features from the hierarchical image set of diseases of the green plant part to obtain similar feature data; calculate the texture and morphological similarity of the hierarchical image set of diseases of the green plant part using the Euclidean distance formula based on the similar feature data to obtain the first similarity data;

[0114] Step S2443: Calculate the hue and saturation similarity of the hierarchical image set of diseases of the green plant part using the cosine similarity formula based on the similar feature data to obtain the second similarity data;

[0115] Step S2444: Construct a disease symptom similarity matrix based on the first similarity data and the second similarity data, and group the lesion image blocks of the green plant part into similar disease types according to the constructed disease symptom similarity matrix to generate grouped image blocks of similar disease types;

[0116] Step S2445: Calibrate the disease types of the grouped image blocks of similar disease types to generate similar disease type recognition data.

[0117] In the embodiment of the present invention, the minor lesion frequency data in step S243 is checked. If the lesion frequency of some image blocks is greater than 1, it is considered that the image block contains multiple lesion areas. For these image blocks with a frequency greater than 1, hierarchical analysis is performed. Analyze the spatial distribution and frequency of the lesion areas to determine the level of each lesion image block, and generate a hierarchical image set of diseases of the green plant part. According to the frequency information of the minor lesions, the lesion image blocks are divided into multiple levels to distinguish the severity or other characteristics (such as texture, morphology, etc.) of different lesion areas. Each level of image blocks generates a separate hierarchical image set, and these image sets can be implemented by image segmentation methods, region growing algorithms, or threshold-based techniques. Extract texture and morphological features from the hierarchical image set of diseases. Texture features can be extracted by methods such as gray-level co-occurrence matrix (GLCM), local binary pattern (LBP), etc. Morphological features can be extracted by edge detection, shape descriptors, etc. The extracted texture and morphological features will be combined into similar feature data to prepare for subsequent similarity calculations. Use the Euclidean distance formula to calculate the texture and morphological similarity of the extracted similar feature data. The formula is as follows: where d is the similarity, x i and y i$v_{i}$ is the value of two image patches in the $i$-th feature dimension, and $n$ is the number of feature dimensions. The first similarity data is generated from the calculated similarity values, representing the similarity in texture and morphology of different disease regions. Hue and Saturation information is extracted from the images of the lesion areas, and these features are particularly prominent in the HSV color space. The RGB or HSV color model of the image is used for extraction. The Hue and Saturation features are extracted as the similarity feature data. Based on the extracted Hue and Saturation data, the cosine similarity formula is used to calculate the Hue and Saturation similarities between different image patches. The formula is as follows: where $A$ and $B$ represent the Hue and Saturation feature vectors of two image patches respectively, and $\|A\|$ and $\|B\|$ are their respective norms. The second similarity data is generated from the calculated cosine similarity values, representing the similarity in Hue and Saturation of different disease regions. Based on the first similarity data (texture and morphology) and the second similarity data (Hue and Saturation), a disease symptom similarity matrix is constructed. Each element in the matrix represents the similarity between two sets of image patches, and the matrix dimension is the number of image patches in the lesion area image. The first and second similarity data are combined and comprehensively processed using methods such as weighted average and principal component analysis (PCA). Based on the similarity matrix, clustering algorithms (such as K-means, hierarchical clustering, etc.) are used to group the lesion image patches into similar disease types. The clustering results will provide labels for the image patches grouped into similar disease types, forming disease type groups. The disease type of the image patches grouped into similar disease types is calibrated, which can be performed through manual calibration, rule inference, or model-based automatic calibration. The calibrated type for each group can be confirmed based on existing disease standards or expert knowledge. According to the calibration results, similar disease type recognition data is generated, which includes labels for different disease types, such as minor disease types or similar disease types.

[0118] As an example of the present invention, refer to Figure 3 shown. In this example, step S3 includes:

[0119] Step S31: Continuously analyze the images of the lesion areas of the green plant parts based on the minor disease types to generate a continuous image comparison set of the lesion areas;

[0120] Step S32: Perform short-term extended anomaly monitoring on the green plants with minor lesion types through the continuous image comparison set of the lesion areas to generate short-term disease anomaly monitoring data;

[0121] Step S33: Perform dynamic video monitoring of the green plant parts on the images of the lesion areas of the green plant parts based on the similar disease types to generate long-term anomaly monitoring data of the lesion areas;

[0122] Step S34: Analyze the short-term disease abnormal monitoring data and the long-term abnormal monitoring data of the diseased area based on the preset environmental sensors to generate the green plant disease transmission trend data, and convert the green plant disease transmission trend data into a green plant disease abnormal risk map through the graph theory method.

[0123] In the embodiments of the present invention, by continuously comparing the images of the diseased spot areas of the green plant parts, focusing on the change patterns of minor diseases, the evolution and expansion of the diseased areas can be identified. Select the images of the diseased spot areas of the green plants within the time span, and analyze the characteristics such as the morphology, color tone, and texture of the diseased areas through frame-by-frame comparison. Use image registration techniques (such as feature point matching, image transformation) to ensure that the images at different time points can be accurately aligned. Pixel-level comparison of the diseased spot areas is performed through differential image methods or optical flow analysis to highlight the changes in the diseased spot areas. After each image comparison, the difference areas are extracted and retained, and finally a continuous image comparison set of the diseased areas is generated. These image sets contain the diseased areas in each time period of the disease evolution. Based on the continuous image comparison set of the diseased areas, monitor the abnormal expansion of minor disease types in a short period of time, and identify potential rapid spread or acute diseases. Through time series data comparison, analyze the expansion trend of the diseased areas in the short term, and generate short-term disease anomaly monitoring data. Utilize the continuous image comparison set, combined with image change detection (such as differential images, edge detection), to monitor the anomalies such as the growth and morphological changes of the diseased areas. Define the criteria for abnormal expansion through parameters such as the size and change speed of the diseased areas, and set a threshold. If the threshold is exceeded, generate anomaly monitoring data. When there are sharp changes in the diseased areas, generate short-term disease anomaly monitoring data, indicating that the area is entering the disease outbreak period. Use the continuously captured video stream to perform real-time monitoring of the green plant parts, and identify the long-term changes in the diseased areas. Extract each frame in the dynamic video, and perform image analysis on the diseased spot areas to identify the characteristics such as the size, shape, and color of the diseased spots. Use image analysis and change detection methods (such as optical flow method) to monitor the long-term changes in the diseased spot areas, and evaluate the change rate and trend of the diseased spots. Through video monitoring, continuously track the expansion of the diseased spots, generate long-term anomaly monitoring data of the diseased areas, and help judge the long-term development trend of the disease. Use environmental data (such as temperature, humidity, etc.) combined with disease monitoring data, and analyze the spread dynamics of the disease through a mathematical model to predict the future expansion trend of the disease. Combine the data provided by environmental sensors (such as humidity, temperature, precipitation, etc.) with short-term and long-term anomaly monitoring data to form a multi-dimensional data set. Analyze the potential paths, spread speed, and influence range of disease spread through graph theory methods (such as shortest path, network propagation model, etc.). By constructing a propagation network, speculate on the future expansion areas of the disease. Based on the results of the spread trend analysis, generate disease spread risk data to identify high-risk areas and time periods. According to the spread risk data, use graph theory methods (such as Dijkstra algorithm, minimum spanning tree, etc.) to draw an abnormal risk map of green plant diseases, marking the spread hotspots and susceptible areas. As the disease dynamically changes, continuously update the risk map for real-time risk assessment.

[0124] Preferably, the analysis of the green plant disease transmission trend based on the short-term disease abnormal monitoring data and the long-term abnormal monitoring data of the diseased area by the preset environmental sensor includes:

[0125] Collect the temperature and humidity data of the green plant lesions according to the preset temperature and humidity sensor for the short-term disease abnormal monitoring data and the long-term abnormal monitoring data of the diseased area, and obtain the temperature and humidity data of the green plant diseased area;

[0126] Conduct time series analysis on the temperature and humidity data of the green plant diseased area to generate the temperature and humidity change data of the green plant diseased area; Based on the temperature and humidity change data of the green plant diseased area, confirm the temperature and humidity change area of the green plant lesion, and use a multispectral imager to perform regional spectral imaging to generate a regional spectral imaging map;

[0127] Map the temperature and humidity data of the green plant diseased area to the spectral band of the regional spectral imaging map to generate the spectral band data of the green plant diseased area; Conduct direction change analysis on the spectral band data of the green plant diseased area to generate the propagation wind direction data of the green plant diseased area;

[0128] Use the propagation wind direction data of the green plant diseased area to predict the propagation trend of the short-term disease abnormal monitoring data and the long-term abnormal monitoring data of the diseased area, and generate the green plant disease transmission risk data.

[0129] In the embodiments of the present invention, temperature and humidity sensors are installed in a preset area to obtain the temperature and humidity data of the green plant disease area. The sensors are configured to periodically collect short-term disease anomaly monitoring data and long-term lesion area data. Ensure the accuracy and precision of the sensors to ensure that the collected data can reflect the temperature and humidity changes in the green plant disease area. The collected temperature and humidity data are processed through time series analysis. Statistical analysis methods, such as moving average or autoregressive integrated moving average (ARIMA) models, are used to capture and analyze the fluctuations of temperature and humidity, generating the temperature and humidity change data of the green plant lesion area, providing accurate climate information for subsequent analysis. Based on the temperature and humidity change data, the areas with significant temperature and humidity changes are determined, and these areas are the areas where green plant diseases occur or spread. In these areas, a multispectral imager is used for spectral imaging to obtain image data in different bands, helping to identify the regional characteristics of the diseases. The temperature and humidity data of the green plant lesion area are mapped to each spectral band of the multispectral imaging map, generating the spectral band data of the green plant lesion area. Through the comparison and analysis of the band data, the growth status and disease characteristics of the green plants are identified, such as leaf lesions, withering or other visual characteristics. Direction change analysis is performed on the spectral band data. Through image processing algorithms (such as gradient change analysis or direction field estimation), the wind direction of disease spread is determined. Using the historical data of wind speed and wind direction, the spread trend of green plant diseases is calculated, predicting the spread path of diseases from one area to another. Based on the spread wind direction data, combined with short-term disease anomaly monitoring data and long-term anomaly monitoring data of the lesion area, prediction models (such as regression analysis, machine learning algorithms, etc.) are used to predict the spread trend of diseases, generating the green plant disease spread risk data, helping agricultural workers understand the future disease spread trend and providing guidance for taking prevention and control measures. The predicted disease spread trend, risk data and temperature and humidity change information are presented through a visualization tool to form a disease spread risk map, which is convenient for agricultural managers to refer to when making decisions. Provide data-based emergency response suggestions, such as strengthening the protection measures in certain areas or regularly monitoring the lesion areas.

[0130] Preferably, step S4 includes the following steps:

[0131] Step S41: Use the green plant disease anomaly risk map to carry out early warning of the green plant lesion risk for the image of the lesion area of the green plant part, generating green plant lesion risk early warning data;

[0132] Step S42: Construct a prevention and control plan according to the green plant lesion risk early warning data, generating a green plant lesion prevention and control plan; Visualize the green plant lesion prevention and control plan to execute the green plant disease prevention and control management operation.

[0133] In the embodiments of the present invention, an image of the diseased area of the green plant part is obtained by using a multispectral imager, and combined with data such as temperature, humidity, and propagation wind direction, a risk map of abnormal green plant diseases is generated. The risk map is calibrated according to the size, distribution, and severity of the green plant diseased spots, and classified and labeled through machine learning models (such as support vector machines, random forests, etc.) to indicate areas of different risk levels. Based on the generated risk map, the image of the diseased area of the green plant part is analyzed, and real-time risk assessment is carried out by combining historical data and current environmental factors. The risk assessment results are used to give early warnings of the risk of green plant lesions, automatically generate early warning data for the risk of green plant lesions, indicate which areas of green plants are at high risk, and make preparations for prevention and control in advance. The early warning data for the risk of green plant lesions is output to a monitoring platform or relevant management system to timely understand the spread situation and degree of the disease. The early warning data can be presented in the form of charts, maps, or digital reports for agricultural managers to intuitively view and analyze. According to the early warning data for the risk of green plant lesions, combined with the experience of agricultural experts, historical prevention and control data, and the disease transmission law, a targeted prevention and control plan for green plant lesions is constructed. The prevention and control plan should include information such as the identification of disease types, prevention and control measures (such as spraying pesticides, adjusting temperature and humidity, etc.), response strategies (such as regional blockade, rotation planting, etc.), and implementation time points. Visualization tools (such as GIS systems, 3D modeling software, etc.) are used to visualize the relevant data of the prevention and control plan (such as prevention and control areas, operation times, pesticide application amounts, etc.). The prevention and control plan for green plant lesions is converted into easy-to-understand graphics, maps, or dynamic charts to ensure that agricultural managers can quickly grasp the implementation steps of the prevention and control measures. The generated prevention and control plan for green plant lesions is integrated into the farmland management system to provide specific operation guidelines and resource allocation suggestions. Monitoring and data feedback during the implementation process will update the prevention and control plan in real time to dynamically adjust the response strategy. During the implementation of the prevention and control work, data is collected for feedback to timely adjust the prevention and control measures. According to the monitoring results and implementation effects, the prevention and control plan is optimized and updated to continuously improve the accuracy and efficiency of green plant disease prevention and control.

[0134] In this specification, a green plant disease identification and early warning management system is provided for implementing the above-mentioned green plant disease identification and early warning management method. The green plant disease identification and early warning management system includes:

[0135] A part segmentation module, configured to obtain a high-definition image of a green plant; perform image preprocessing on the high-definition image of the green plant to generate a standard high-definition image of the green plant; perform image segmentation and integration on the standard high-definition image of the green plant to obtain a standard image of the green plant part;

[0136] The disease type recognition module is used to extract multi-dimensional disease characteristics of the standard green plant part images to obtain multi-dimensional disease characteristics of the green plant parts; identify the diseased area of the standard green plant part images based on the multi-dimensional disease characteristics of the green plant parts, and generate the recognition data of the diseased area of the green plant parts; determine the type of green plant diseases for the standard green plant part images according to the recognition data of the diseased area of the green plant parts, and generate the discrimination data of the green plant disease types, where the discrimination data of the green plant disease types includes minor disease types and similar disease types.

[0137] The disease transmission analysis module is used to analyze the transmission trend of green plant diseases for the green plants with minor lesion types based on the minor disease types and similar disease types, generate the risk data of green plant disease transmission, and convert the risk data of green plant disease transmission into the abnormal risk map of green plant diseases through the graph theory method.

[0138] The risk warning module is used to give early warning of the risk of green plant lesions for the images of the diseased areas of the green plant parts by using the abnormal risk map of green plant diseases, and generate the early warning data of green plant lesion risks; construct a prevention and control plan according to the early warning data of green plant lesion risks to perform the prevention and control management operation of green plant diseases.

[0139] Through the acquisition and preprocessing of high-definition images, the present invention ensures the quality and consistency of the images, providing an accurate basis for subsequent processing. Image segmentation and integration isolate each part of the green plant, making the subsequent disease diagnosis more accurate, enabling clear identification and analysis of the disease manifestations of each part, and avoiding confusion and misjudgment. By extracting multi-dimensional disease characteristics of the green plant parts, various manifestations of diseases can be comprehensively and accurately analyzed, improving the accuracy of disease diagnosis. The identification of the diseased area can accurately locate the disease area and classify it according to different disease types, reducing misjudgments of minor diseases and similar diseases, ensuring the accurate determination of the disease type, and providing a scientific basis for subsequent disease prevention and control. By analyzing the transmission trend of minor disease types and similar disease types, the risk of disease spread can be predicted in advance, potential high-risk areas can be identified, and the risk data of green plant disease transmission can be generated. This method enables disease prevention and control to shift from static diagnosis to dynamic early warning, effectively reducing the spread of diseases. Using the abnormal risk map of green plant diseases to give early warning of the risk of green plant lesions can timely detect potential risk areas before the spread of diseases, provide real-time feedback and early intervention. The prevention and control plan constructed based on these early warning data has high accuracy, making the prevention and control management of diseases more efficient and timely, reducing resource waste and improving the effect of disease control. Therefore, by introducing high-definition image processing, multi-dimensional feature extraction, transmission trend analysis and dynamic early warning, the present invention improves the accuracy and response timeliness of green plant disease identification and early warning.

[0140] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Thus, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be encompassed within the present invention.

[0141] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A green plant disease identification and early warning management method, characterized in that: The following steps are involved: Step S1: Acquire a high-definition image of green plants; perform image preprocessing on the high-definition image of green plants to generate a standard high-definition image of green plants; perform image segmentation and integration on the standard high-definition image of green plants to obtain a standard image of green plant parts; Step S2: extracting multi-dimensional disease features of the parts of the standard green plant images to obtain multi-dimensional disease features of the parts of the green plant; identifying the diseased areas of the standard green plant images based on the multi-dimensional disease features of the parts of the green plant, and generating identification data of the diseased areas of the parts of the green plant; distinguishing the types of green plant diseases on the standard green plant images based on the identification data of the diseased areas of the parts of the green plant, and generating identification data of the types of green plant diseases, wherein the identification data of the types of green plant diseases include minor disease types and similar disease types; Step S3: analyzing the green plant disease propagation trend of green plants with slight disease types based on slight disease types and similar disease types, generating green plant disease propagation risk data, and converting the green plant disease propagation risk data into a green plant disease abnormality risk map through a graph theory method; Step S4: Use the abnormal risk map of green plant diseases to conduct green plant disease risk warning on the images of the diseased areas of green plant parts, and generate green plant disease risk warning data; construct a prevention and control plan based on the green plant disease risk warning data to perform green plant disease prevention and control management operations.

2. The green plant disease identification and early warning management method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: using a high-definition digital camera to obtain a high-definition image of green plants; Step S12: performing image filtering on the green plant high-definition image to generate a green plant high-definition filtered image; Step S13: performing image brightness enhancement on the green plant high-definition filtered image to generate a green plant high-definition brightness enhanced image; Step S14: performing image geometric transformation on the green plant high-definition brightness enhanced image to generate a standard green plant high-definition image; performing green plant part image segmentation and integration on the standard green plant high-definition image to obtain a standard green plant part image.

3. The green plant disease identification and early warning management method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: extracting multi-dimensional disease features of the parts of the standard green plant from the images to obtain multi-dimensional disease features of the parts of the green plant; Step S22: identifying the diseased area of ​​the standard green plant part image based on the multi-dimensional disease characteristics of the green plant part, and generating green plant part diseased area identification data; Step S23: separating the diseased spot from the background of the standard green plant part image according to the green plant part diseased spot area recognition data to obtain a green plant part diseased spot area image, and extracting morphological features of the green plant part diseased spot area image to obtain diseased spot area morphological data; Step S24: discriminating the green plant disease type of the green plant spot area image by using the spot area morphology data to generate green plant disease type discrimination data, wherein the green plant disease type discrimination data includes a minor disease type and a similar disease type.

4. The green plant disease identification and early warning management method according to claim 3, characterized in that: Step S21 includes the following steps: Step S211: constructing a grayscale symbiosis matrix for the standard green plant part image, and using the constructed grayscale symbiosis matrix to perform green plant leaf texture feature analysis on the standard green plant part image to generate green plant leaf texture feature data; Step S212: reconstructing the image of the standard green plant part according to the color histogram to generate a green plant part color histogram; performing RGB channel analysis on the green plant part color histogram to generate RGB channel data; performing HSV color mapping on the standard green plant part image according to the RGB channel data to generate green plant color feature data; Step S213: performing edge detection on the standard green plant part image based on the Sobel operator algorithm to generate green plant part edge detection data; performing serration calculation on the green plant part edge detection data to obtain the green plant part edge serration; performing structural feature extraction on the standard green plant part image according to the green plant part edge serration to obtain the green plant part structural feature data; Step S214: Integrate the green plant leaf texture feature data, the green plant color feature data and the green plant part structure feature data into a multi-dimensional disease feature of the green plant part.

5. The green plant disease identification and early warning management method according to claim 3, characterized in that: Step S24 includes the following steps: Step S241: dividing the image of the diseased spot area of ​​the green plant into image blocks to generate image blocks of the diseased spot of the green plant; performing wavelet transformation on the image blocks of the diseased spot of the green plant according to the morphological data of the diseased spot area to generate a transformed image of the diseased spot of the green plant; Step S242: extracting high-frequency details of the transformed image of the diseased spot at the green plant site to obtain detailed data of the diseased spot at the green plant site; using the detailed data of the diseased spot at the green plant site to identify minor diseases in the transformed image of the diseased spot at the green plant site to generate minor disease type identification data; Step S243: calculating the occurrence frequency of the minor lesion type of the diseased spot image block at the plant site based on the minor disease type image, and obtaining minor lesion frequency data; Step S244: when the frequency data of minor lesions is greater than 1, the lesion image blocks of the green plant parts are layered analyzed according to the frequency data of minor lesions to generate a layered image set of green plant parts diseases; the symptom similarity analysis is performed on the layered image set of green plant parts diseases to generate similar disease type identification data; Step S245: Use the minor disease type identification data and the similar disease type identification data to identify the plant disease type of the diseased spot area image of the green plant part, and generate green plant disease type identification data, wherein the green plant disease type identification data includes minor disease types and similar disease types.

6. The green plant disease identification and early warning management method according to claim 5, characterized in that: Step S244 includes the following steps: Step S2441: when the frequency data of slight lesions is greater than 1, a layered analysis is performed on the diseased spot image blocks of the green plant parts according to the frequency data of slight lesions to generate a layered image set of diseases of the green plant parts; Step S2442: extract similar features from the layered image set of green plant disease parts to obtain similar feature data; calculate texture and morphological similarity of the layered image set of green plant disease parts using the Euclidean distance formula based on the similar feature data to obtain first similarity data; Step S2443: using the cosine similarity formula to calculate the hue and saturation similarity of the disease layered image set of the green plant parts according to the similar feature data, to obtain second similarity data; Step S2444: constructing a disease symptom similarity matrix based on the first similarity data and the second similarity data, and grouping the diseased spot image blocks of the green plant parts into similar disease types according to the constructed disease symptom similarity matrix to generate image blocks grouped into similar disease types; Step S2445: perform disease type calibration on the image blocks grouped with similar disease types, thereby generating similar disease type identification data.

7. The green plant disease identification and early warning management method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing continuous image comparison analysis on the images of the diseased areas of the green plant parts based on the types of minor diseases, and generating a continuous image comparison set of the diseased areas; Step S32: Perform short-term extended abnormal monitoring on green plants with slight lesions through continuous image comparison of the lesion area to generate short-term disease abnormal monitoring data; Step S33: performing dynamic video monitoring of the green plant parts on the images of the diseased areas of the green plant parts based on similar disease types, and generating long-term abnormal monitoring data of the diseased areas; Step S34: Analyze the green plant disease propagation trend based on the short-term disease abnormality monitoring data and the long-term abnormality monitoring data of the lesion area according to the preset environmental sensors, generate green plant disease propagation risk data, and convert the green plant disease propagation risk data into a green plant disease abnormality risk map through graph theory methods.

8. The green plant disease identification and early warning management method according to claim 7, characterized in that: The green plant disease propagation trend analysis based on the short-term disease abnormality monitoring data and the long-term abnormality monitoring data of the diseased area according to the preset environmental sensors includes: According to the preset temperature and humidity sensors, the temperature and humidity data of green plant lesions are collected for the short-term abnormal disease monitoring data and the long-term abnormal disease area monitoring data, and the temperature and humidity data of the green plant lesion area are obtained; Perform time series analysis on the temperature and humidity data of the green plant disease area to generate temperature and humidity change data of the green plant disease area; confirm the temperature and humidity change area of ​​the green plant disease based on the temperature and humidity change data of the green plant disease area, and use a multi-spectral imager to perform regional spectral imaging to generate a regional spectral imaging map; The temperature and humidity data of the green plant disease area are mapped to the spectral bands of the regional spectral imaging map to generate the spectral band data of the green plant disease area; the spectral band data of the green plant disease area are analyzed for direction changes to generate the propagation wind direction data of the green plant disease area; The wind direction data of the green plant disease area is used to predict the propagation trend of the short-term disease abnormal monitoring data and the long-term abnormal monitoring data of the disease area, and the green plant disease propagation risk data is generated.

9. The green plant disease identification and early warning management method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: using the green plant disease abnormal risk map to perform green plant disease risk warning on the image of the diseased spot area of ​​the green plant part, and generating green plant disease risk warning data; Step S42: construct a prevention and control plan based on the green plant disease risk warning data to generate a green plant disease prevention and control plan; visualize the green plant disease prevention and control plan data to perform green plant disease prevention and control management operations.

10. A green plant disease identification and early warning management system, characterized in that: Used to execute the green plant disease identification and early warning management method as claimed in claim 1, the green plant disease identification and early warning management system comprises: The part segmentation module is used to obtain a high-definition image of green plants; perform image preprocessing on the high-definition image of green plants to generate a standard high-definition image of green plants; perform green plant part image segmentation and integration on the standard high-definition image of green plants to obtain a standard image of green plant parts; The disease type recognition module is used to extract multi-dimensional disease features of standard green plant parts images to obtain multi-dimensional disease features of green plant parts; identify the diseased area of ​​standard green plant parts images based on the multi-dimensional disease features of green plant parts to generate green plant part diseased area recognition data; discriminate the green plant disease type of standard green plant parts images based on the green plant part diseased area recognition data to generate green plant disease type discrimination data, wherein the green plant disease type discrimination data includes minor disease types and similar disease types; The disease propagation analysis module is used to analyze the disease propagation trend of green plants with slight lesions based on slight disease types and similar disease types, generate green plant disease propagation risk data, and convert the green plant disease propagation risk data into a green plant disease abnormality risk map through graph theory methods; The risk warning module is used to use the abnormal risk map of green plant diseases to conduct green plant disease risk warning on the images of diseased areas of green plant parts, and generate green plant disease risk warning data; and to construct prevention and control plans based on the green plant disease risk warning data to perform green plant disease prevention and control management operations.

Images