Tea garden disease and pest monitoring method and system based on photovoltaic light supplement

Abstract

The invention relates to the technical field of tea garden disease and insect pest monitoring, in particular to a tea garden disease and insect pest monitoring method and system based on photovoltaic light supplement. The method comprises the following steps that photovoltaic light supplementing equipment is arranged, and tea garden environment data are obtained; carrying out photovoltaic light supplement operation according to the tea garden environment data, and carrying out tea tree image acquisition through camera equipment to obtain tea tree image data; tea tree trunk feature extraction and tea tree tea feature extraction are carried out according to the tea tree image data, and tea tree trunk feature data and tea tree tea feature data are obtained; performing pest assessment according to the tea tree trunk feature data and the tea tree tea feature data to obtain pest assessment data; and performing pest distribution generation according to the pest evaluation data, the tea tree trunk feature data and the tea tree tea feature data to obtain pest distribution data. According to the invention, an intelligent and automatic tea garden disease and insect pest monitoring and prevention and control system is formed by combining photovoltaic light supplement and disease and insect pest simulation.

Description

Technical Field

[0001] The present invention relates to the technical field of tea garden pest monitoring, and in particular to a tea garden pest monitoring method and system based on photovoltaic supplementary lighting. Background Art

[0002] In modern agriculture, the monitoring and prevention of tea tree pests and diseases is crucial to the production management of tea gardens. Early detection and timely treatment of pests and diseases can effectively reduce tea production losses and improve the quality of tea. However, traditional tea tree pest and disease monitoring methods rely on manual inspections and empirical judgments, and have problems such as low detection efficiency and high misjudgment rate, which are difficult to meet the actual needs of large-scale tea gardens. In addition, due to the complex growth environment of tea trees, different tea varieties, climatic conditions, and light changes have different manifestations of pests and diseases, which poses a challenge to the accurate identification of pests and diseases. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a tea garden pest and disease monitoring method and system based on photovoltaic supplementary lighting to solve at least one of the above technical problems.

[0004] The present application provides a method for monitoring tea garden pests and diseases based on photovoltaic supplementary lighting, the method comprising:

[0005] S1. Arrange photovoltaic lighting equipment and obtain tea garden environmental data;

[0006] S2, perform photovoltaic supplementary lighting operations according to the tea garden environmental data, and collect tea tree images through a camera device to obtain tea tree image data;

[0007] S3, extracting tea tree trunk features and tea tree leaf features according to the tea tree image data, and obtaining tea tree trunk feature data and tea tree leaf feature data;

[0008] S4. Perform pest and disease assessment based on the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease assessment data; generate pest and disease distribution based on the pest and disease assessment data, the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease distribution data.

[0009] The present invention realizes continuous monitoring of tea gardens in the case of insufficient light (such as at night or on cloudy days) by arranging photovoltaic lighting equipment, ensures that the pest monitoring process is not interrupted, and improves the coverage and real-time performance of tea garden pest monitoring. The method dynamically adjusts the operation mode of photovoltaic lighting equipment by acquiring and analyzing the environmental data of the tea garden, so that the lighting equipment can adapt to different environmental conditions (such as temperature, humidity, wind speed, etc.), provides optimal lighting conditions, and ensures that the collected tea tree image data is clear and reliable. Through the accurate extraction of tea tree trunk feature data and tea tree tea leaf feature data, the method can efficiently identify the pest and disease situation of tea trees, and the pest and disease assessment is more accurate, reducing the problems of missed detection or misdetection in traditional manual inspections. Using the tea tree trunk feature data and tea tree tea leaf feature data combined with the pest and disease assessment model, pests and diseases can be intelligently assessed. Its automated assessment method reduces the deviation caused by human judgment and improves the efficiency and accuracy of monitoring. By combining pest and disease assessment data, tea tree trunk characteristic data and tea tree tea leaf characteristic data to generate pest and disease distribution data, it is possible to clearly display the distribution range and extent of pests and diseases, and provide strong data support for pest and disease prevention and control work in tea gardens.

[0010] Optionally, S1 includes:

[0011] Get tea tree location data;

[0012] The photovoltaic lighting equipment placement position is generated according to the tea tree position data to obtain the photovoltaic lighting equipment placement position data;

[0013] The photovoltaic fill lighting equipment is arranged according to the placement data of the photovoltaic fill lighting equipment, and the tea garden environment data is obtained.

[0014] By acquiring the tea tree location data, the present invention can accurately determine the placement of the photovoltaic lighting equipment according to the specific distribution of the tea trees in the tea garden, thereby reducing the blind placement of the lighting equipment, ensuring that the lighting equipment can effectively cover the tea trees, and maximizing the lighting effect. The precise arrangement of the photovoltaic lighting equipment can provide sufficient light sources for the tea trees when the light is insufficient, thereby improving the photosynthesis efficiency of the tea trees, thereby promoting the healthy growth of the tea trees and improving the overall yield and quality of the tea garden.

[0015] Optionally, S2 includes:

[0016] Perform supplementary lighting equipment operation scheduling according to tea garden environmental data to obtain supplementary lighting equipment operation scheduling data;

[0017] Carry out photovoltaic lighting operations according to the lighting equipment operation scheduling data;

[0018] Collect tea tree images through camera equipment to obtain preliminary tea tree image data;

[0019] Perform tea tree region segmentation on the preliminary tea tree image data according to the supplementary lighting equipment operation scheduling data and the tea garden environment data to obtain tea tree region segmentation data;

[0020] The tea tree region segmentation data is subjected to illumination adaptation image enhancement and aggregation to obtain tea tree image data.

[0021] The present invention performs operation scheduling of the supplementary lighting equipment according to the tea garden environmental data, and can intelligently control the operation time, light intensity and angle of the photovoltaic supplementary lighting equipment, effectively adapt to the real-time environmental changes of the tea garden, improve the light energy utilization rate and operation efficiency, and reduce unnecessary energy waste. The optimization of the lighting environment enables the camera equipment to obtain more uniform lighting when collecting tea tree images, thereby improving the quality of the preliminary image data and reducing image distortion or noise caused by uneven lighting. The preliminary tea tree image is segmented according to the supplementary lighting equipment operation scheduling data and the tea garden environmental data, and the main area of ​​the tea tree can be accurately divided. By performing image enhancement for lighting adaptation on the tea tree area segmentation data, the clarity and detail retention effect of the image can be further improved. The lighting adaptation process can correct the lighting deviation in the image, so that the images collected under different lighting conditions remain consistent, and the reliability of the overall data is improved.

[0022] Optionally, the tea tree region segmentation comprises:

[0023] Perform illumination compensation calculation on the preliminary tea tree image data according to the operation scheduling data of the supplementary lighting equipment and the tea garden environment data to obtain the illumination compensation data of the tea tree image;

[0024] Perform preliminary boundary detection on the illumination compensation data of the tea tree image to obtain preliminary boundary data of the tea tree;

[0025] According to the preliminary boundary data of the tea tree, the main leaf area of ​​the tea tree image illumination compensation data is segmented to obtain the main leaf area data;

[0026] The data of the main leaf area is occluded and repaired to obtain the segmentation data of the tea tree area.

[0027] In the present invention, by combining the operation scheduling data of the supplementary light equipment and the tea garden environment data, the preliminary tea tree image is subjected to illumination compensation calculation, and the image deviation caused by uneven illumination is effectively corrected. This process ensures that each area of ​​the tea tree in the image has uniform brightness, thereby improving the accuracy and stability of subsequent regional segmentation. By performing preliminary boundary detection on the image after illumination compensation, the outlines of the trunk and leaves of the tea tree can be identified, the basic accuracy of regional segmentation is ensured, and clear contour lines are provided for the trunk and leaf segmentation, reducing the mis-segmentation caused by complex background. Based on the boundary data of the preliminary detection, the trunk and leaves of the tea tree image are segmented, and the different parts of the tea tree can be clearly distinguished, the recognition ability of the tea tree structure is improved, and it is helpful to more accurately analyze the pests and diseases of different parts, and improve the detection efficiency and effect. Through the occlusion repair technology, the problem of partial data missing or segmentation error caused by occlusion (such as mutual overlap, illumination shadow, etc.) in the trunk and leaf areas of the tea tree is solved, and the occluded partial area can be restored to ensure the integrity and consistency of the tea tree regional segmentation.

[0028] Optionally, the illumination compensation calculation includes:

[0029] Generating the illumination unevenness area according to the supplementary lighting equipment operation scheduling data and the tea garden environment data to obtain the illumination unevenness area data;

[0030] Perform multi-scale segmentation on the preliminary tea tree image data according to the illumination inhomogeneity region data to obtain multi-scale segmentation data of the tea tree image;

[0031] Performing region detection based on multi-scale segmentation data of tea tree images to obtain heterogeneous region data of tea tree images;

[0032] Performing brightness gradient calculation on the uneven area data of the tea tree image to obtain brightness gradient data of the tea tree image;

[0033] Performing local compensation processing on the brightness gradient data to obtain local compensation data of the tea tree image;

[0034] Reconstructing the illumination boundary of the tea tree image local compensation data to obtain the illumination boundary data of the tea tree image;

[0035] Boundary superposition processing is performed based on the local compensation data of the tea tree image and the illumination boundary data of the tea tree image to obtain the illumination compensation data of the tea tree image.

[0036] In the present invention, by using the operation scheduling data of the supplementary lighting equipment and the tea garden environmental data, the method can dynamically generate the illumination unevenness area data according to the changes in the illumination and environmental conditions, and can cope with complex illumination conditions (such as illumination intensity in different time periods, shadow influence, etc.), thereby improving the robustness and flexibility of image processing. The multi-scale segmentation technology is used to process the tea tree image, and the details and structures in the image can be captured at different scales. Whether it is large-scale illumination unevenness or local small-scale illumination anomaly, it can be effectively identified and compensated. The brightness gradient calculation of the illumination unevenness area data enables the system to identify the trend and amplitude of the brightness change in the image and accurately locate the area of ​​illumination abnormality. After the brightness gradient calculation is completed, local compensation processing is performed to ensure that each uneven illumination area can obtain appropriate brightness adjustment. The illumination boundary reconstruction process fixes the boundary distortion and blurring problems caused by uneven illumination, ensuring that the reconstructed boundary lines are smoother and continuous. By superimposing the local compensation data with the illumination boundary data, it is ensured that the boundary information of the image is consistent with the brightness after compensation.

[0037] Optionally, S3 includes:

[0038] Reconstructing the main structure of tea leaves according to the tea tree image data to obtain the main structure reconstruction data of tea leaves;

[0039] Extract the trunk morphological features based on the tea trunk structure reconstruction data to obtain the tea tree trunk feature data;

[0040] Perform tea tree leaf detection according to the tea tree image data to obtain tea tree leaf data;

[0041] Tea tree and tea leaves features are extracted based on the tea tree and tea leaves data to obtain tea tree and tea leaves feature data.

[0042] In the present invention, the trunk structure is reconstructed through tea tree image data, and the three-dimensional morphology and structure of the tea tree trunk can be accurately reproduced. The extraction of trunk morphological features helps to identify the key structural features of the tea tree trunk, such as the thickness, shape, surface texture, etc. of the trunk. Extracting the key features of tea leaves (such as leaf color, shape, texture, size, etc.) from tea tree tea data can help identify the health status of tea leaves and the existing pests and diseases.

[0043] Optionally, S4 includes:

[0044] Using a preset tea tree pest and disease identification model to evaluate the tea tree trunk characteristic data and tea leaf characteristic data, to obtain pest and disease evaluation data;

[0045] Locate the pest area based on the pest assessment data to obtain pest area data;

[0046] The severity of pests and diseases is calculated based on the pest and disease area data, the tea tree trunk characteristic data and the tea tree tea leaf characteristic data to obtain the pest and disease severity data;

[0047] The pest distribution is generated based on the pest severity data and the pest area data to obtain the pest distribution data.

[0048] The present invention accurately identifies the presence and type of pests and diseases. This assessment is based on structured feature data (such as trunk morphology, leaf color, texture, etc.), ensuring that the identification of pests and diseases is more precise and accurate, reducing the situation of missed reports and false reports. The pest and disease area is located according to the pest and disease assessment data, which helps to determine the specific distribution location of pests and diseases in tea trees. The severity calculation of pests and diseases is based on the pest and disease area data, the characteristic data of the tea tree trunk and tea leaves, and can accurately assess the impact and threat level of pests and diseases on tea trees. The pest and disease distribution map generated according to the pest and disease severity data and regional data can clearly show the overall distribution of pests and diseases.

[0049] Optionally, the steps of constructing the preset tea tree pest and disease identification model include:

[0050] Obtain historical tea tree image data and corresponding historical pest and disease data;

[0051] Extract tea tree trunk features and tea tree leaf features based on historical tea tree image data to obtain historical tea tree trunk feature data and historical tea tree leaf feature data;

[0052] Data is divided according to the historical tea tree trunk feature data and the historical tea tree tea leaf feature data to obtain historical tea tree feature training data and historical tea tree feature test data;

[0053] A deep model was constructed based on historical tea tree characteristic training data to obtain a preliminary tea tree pest and disease identification model;

[0054] The preliminary tea tree pest and disease identification model was iteratively trained based on the historical tea tree characteristic test data to obtain the tea tree pest and disease identification model.

[0055] The present invention obtains a large amount of historical tea tree image data and corresponding pest and disease data to ensure that the data foundation for model training and testing is solid. By extracting the characteristics of the trunk and tea leaves from the historical tea tree image data, this method effectively captures the different performances of tea trees in health and pest and disease states. By reasonably dividing the historical tea tree feature data and generating training data and test data respectively, it is ensured that the model will not overfit during training, and the generalization ability of the model can be effectively tested in the testing phase. The preliminary tea tree pest and disease recognition model constructed using deep learning technology can effectively identify complex tea tree pest and disease characteristics. By iteratively training the model, the accuracy and stability of the initially constructed recognition model are improved in continuous updating and optimization.

[0056] Optionally, the pest severity calculation includes:

[0057] Perform pest and disease evolution simulation based on pest and disease assessment data to obtain pest and disease evolution simulation data;

[0058] According to the pest and disease evolution simulation data, the pest and disease area data, tea tree trunk characteristic data and tea tree tea leaf characteristic data are simulated in time series development to obtain pest and disease time series development data;

[0059] The transmission path of the pests and diseases is simulated based on the tea garden environmental data and the pest and disease area data to obtain the pest and disease transmission path data;

[0060] Pest and disease density is calculated based on pest and disease area data, tea tree trunk characteristic data, and tea tree tea leaf characteristic data to obtain pest and disease density data;

[0061] Extract pest and disease area characteristics based on pest and disease area data, tea tree trunk characteristic data, and tea tree tea leaf characteristic data to obtain pest and disease area characteristic data;

[0062] The local severity data of the region is obtained by calculating the local severity based on the time series development data of pests and diseases, the transmission path data of pests and diseases, the density data of pests and diseases, and the regional characteristic data of pests and diseases;

[0063] The severity data of pests and diseases were obtained by spatial integration based on the regional local severity data.

[0064] In the present invention, pest and disease evolution simulation is performed through pest and disease assessment data, and the development trend of pest and disease can be predicted. According to the pest and disease evolution simulation data, the pest and disease area, the main trunk characteristics of tea trees and the tea leaves of tea trees are simulated in time series, and the evolution trend of pest and disease in the future can be accurately predicted. The propagation path simulation is performed in combination with the tea garden environmental data and the pest and disease area data, and the diffusion path and speed of pest and disease can be effectively predicted. The pest and disease density is calculated according to the pest and disease area, the main trunk of tea trees and the tea leaves characteristic data, and the severity of pests and diseases can be accurately quantified. The regional feature extraction of pests and diseases can identify the differences in pest and disease characteristics in different regions by analyzing the spatial distribution characteristics of pests and diseases. The local severity calculation is combined with the time series development, propagation path, density data and regional characteristics of pests and diseases, so that the system can identify the severely affected areas and slightly affected areas of pests and diseases. By spatially integrating the local severity data and generating pest and disease severity data, global and comprehensive pest and disease assessment information can be provided.

[0065] Optionally, the present application further provides a tea garden pest and disease monitoring system based on photovoltaic supplementary lighting, which is used to execute the tea garden pest and disease monitoring method based on photovoltaic supplementary lighting as described above, and the tea garden pest and disease monitoring system based on photovoltaic supplementary lighting includes:

[0066] Tea garden environmental data collection module, used to arrange photovoltaic lighting equipment and obtain tea garden environmental data;

[0067] The tea tree image data acquisition module is used to perform photovoltaic supplementary lighting operations according to the tea garden environmental data, and to acquire tea tree images through a camera device to obtain tea tree image data;

[0068] A tea tree feature extraction module is used to extract tea tree trunk features and tea tree leaf features according to tea tree image data, and obtain tea tree trunk feature data and tea tree leaf feature data;

[0069] The pest and disease distribution calculation module is used to evaluate pests and diseases based on the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease evaluation data; and to generate pest and disease distribution based on the pest and disease evaluation data, the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease distribution data.

[0070] The objects of the present invention are:

[0071] 1. By arranging photovoltaic lighting equipment and combining it with the real-time environmental data collected from the tea garden, dynamic adjustment and intelligent lighting are achieved. The tea garden can still maintain good lighting conditions under different environmental conditions (such as cloudy days, nights, etc.), ensuring that the camera equipment can collect high-quality tea tree image data. Photovoltaic lighting equipment not only improves energy efficiency, but also achieves environmental protection and energy saving, reducing dependence on external energy.

[0072] 2. The characteristic data of tea tree trunks and tea leaves are combined with pest and disease assessment data, and dynamic monitoring of pests and diseases is achieved through pest and disease evolution simulation and propagation path prediction. The system can not only detect the current pest and disease situation, but also predict the future trend of pests and diseases based on historical data, pest and disease time series development simulation, environmental data and propagation path.

[0073] 3. The system can dynamically adjust the distribution map of pests and diseases, and continuously update it according to new data, providing a real-time panoramic view of the pest and disease status, helping managers to accurately grasp the spread and distribution of pests and diseases. Through multi-level analysis of the density, regional characteristics, temporal development and transmission path of pests and diseases, a severity assessment of pests and diseases is formed. Based on the regional severity data, managers can give priority to high-risk areas and effectively optimize the allocation of prevention and control resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0075] Figure 1 A flowchart showing a method for monitoring pests and diseases in a tea garden based on photovoltaic supplementary lighting according to an embodiment of the present invention is shown;

[0076] Figure 2 A flowchart showing a method for collecting environmental data of a tea garden according to an embodiment of the present invention is shown;

[0077] Figure 3 A flowchart showing a method for collecting tea tree image data according to an embodiment is shown;

[0078] Figure 4 A flowchart showing a method for extracting tea tree features according to an embodiment is shown;

[0079] Figure 5 A flowchart showing a method for calculating the distribution of pests and diseases according to an embodiment of the present invention is shown;

[0080] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0081] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0082] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying 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 implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0083] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0084] See also Figures 1 to 5 The present application provides a method for monitoring tea garden pests and diseases based on photovoltaic supplementary lighting, the method comprising:

[0085] S1. Arrange photovoltaic lighting equipment and obtain tea garden environmental data;

[0086] Specifically, determine the terrain and lighting conditions of the tea garden, and select the installation location of the photovoltaic lighting equipment; configure the light intensity and operating time of the photovoltaic lighting equipment according to the amount of sunshine and the light requirements of the tea trees; determine the wavelength and angle of the lighting to ensure uniform coverage of the tea trees.

[0087] Use environmental sensors (such as temperature and humidity sensors, light intensity sensors, etc.) to collect real-time environmental data of the tea garden, including temperature, humidity, light intensity, wind speed, etc.; transmit environmental data to the data acquisition terminal through a wireless sensor network; pre-process the collected data (such as denoising, data completion, etc.) to generate tea garden environmental data.

[0088] S2, perform photovoltaic supplementary lighting operations according to the tea garden environmental data, and collect tea tree images through a camera device to obtain tea tree image data;

[0089] Specifically, the insufficient light areas are analyzed based on the tea garden environmental data, and the light intensity and angle are adjusted using photovoltaic fill lighting equipment to perform automatic fill lighting operations; the working time and light intensity of the photovoltaic panels are set according to the fill lighting needs, and the fill lighting system is dynamically adjusted; the operating status of the photovoltaic equipment is monitored in real time, and equipment fault information is fed back.

[0090] High-resolution video equipment is used to capture full coverage of tea trees, collecting images of tea trees from multiple angles and multiple time periods. The collected tea tree images are preprocessed using image preprocessing algorithms (such as grayscale, noise removal, etc.) to obtain tea tree image data.

[0091] S3, extracting tea tree trunk features and tea tree leaf features according to the tea tree image data, and obtaining tea tree trunk feature data and tea tree leaf feature data;

[0092] Specifically, an edge detection algorithm (such as Canny edge detection) is used to extract the edge of the tea tree trunk; the edge of the trunk is further optimized through morphological processing (such as expansion and corrosion operations); a skeleton extraction algorithm is used to generate the skeleton features of the tea tree trunk, and the length, diameter, curvature and other data of the tea tree trunk are extracted to generate the tea tree trunk feature data.

[0093] Use color segmentation algorithm to separate tea leaves from the background (such as using K-means clustering for color segmentation); extract the shape characteristics of tea leaves (such as area, perimeter, shape coefficient, etc.); extract the texture characteristics of tea leaves through texture analysis algorithm (such as gray-level co-occurrence matrix), identify the health status of leaves, and generate tea tree and tea leaf characteristic data.

[0094] S4. Perform pest and disease assessment based on the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease assessment data; generate pest and disease distribution based on the pest and disease assessment data, the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease distribution data.

[0095] Specifically, combining the characteristic data of the tea tree trunk and the tea leaves, a machine learning model (such as a support vector machine) or a mathematical model is used to identify and classify pests and diseases; the characteristics of pests and diseases in different parts of the tea tree are analyzed to identify the types of pests and diseases (such as downy mildew, powdery mildew, insect pests, etc.); the severity of pests and diseases is evaluated, and pest and disease assessment data is generated; cluster analysis is performed on the pest and disease assessment data to determine the concentrated areas and distribution patterns of pests and diseases.

[0096] Based on the pest and disease assessment data, tea tree trunk characteristic data and tea tree tea leaf characteristic data, a spatial interpolation algorithm (such as Kriging interpolation method) is used to generate a pest and disease distribution map; the pest and disease distribution data is visualized and high-incidence areas of pests and diseases are marked; based on the distribution data, reference suggestions for tea garden pest and disease prevention and control are provided to optimize tea garden management strategies.

[0097] Specifically, the severity of pests and diseases is scored based on a weighted combination of features, with weights set by expert experience or learned from historical data through statistical learning methods (such as logistic regression). Based on the weighted scoring results, the severity of pests and diseases is divided into different levels (such as mild, moderate, and severe), and each level corresponds to different prevention and control measures.

[0098] Specifically, select areas with insufficient light in the tea garden (according to the terrain of the tea garden and the distribution of tea trees), arrange photovoltaic lighting equipment at intervals of 20 meters, and set the equipment height to 2.5 meters. The rated power of the photovoltaic panel is 100W, equipped with a 50Ah energy storage battery, which can continuously provide supplementary light for the tea trees on cloudy days and at night. Install environmental sensors, including temperature and humidity sensors, light intensity sensors, and wind speed sensors. Environmental sensors are arranged within 5 meters around each photovoltaic device. Temperature, humidity, light intensity and wind speed data are recorded once an hour and stored in the monitoring system.

[0099] When the environmental sensor detects that the light intensity is lower than 3000 lux, the photovoltaic lighting equipment starts to provide uniform lighting. The lighting time is about 2 hours a day, especially in the morning and evening. A 360-degree rotating camera with a resolution of 1080p is installed on each photovoltaic device. After the lighting operation is completed, the camera equipment automatically captures the front and side images of the tea tree, and each device takes 50 pictures. The image acquisition interval is set to once every 3 days, and image data is continuously captured for 3 months.

[0100] Use edge detection algorithms (such as the Canny algorithm) to extract the outline of the tea tree trunk from the image, and combine morphological processing (dilation and corrosion operations) to extract the trunk features. The extracted trunk features include: trunk diameter (average diameter is about 3.2cm), height (average height is 1.8m), curvature and number of forks (average number of forks is 12). Use the K-means clustering algorithm for color segmentation and extract the boundaries of the tea leaves. The characteristics of each tea leaf include: leaf area (average area is 42cm 2 ), color distribution (described by color histogram, mainly concentrated in the 120° hue range of green), and texture features (extracted using gray-level co-occurrence matrix). The total number of tea leaves detected is about 120 per tea tree.

[0101] The support vector machine (SVM) model was used to classify and identify the pests and diseases of tea trees. The model was trained using the trunk feature data and tea feature data of tea trees, and mainly identified three types of pests and diseases: downy mildew, powdery mildew, and aphid infestation. The accuracy of the model on the training set was 93%. After classification and evaluation of the test data, it was identified that 12% of the 1,000 tea trees were infected with downy mildew, 8% of the tea trees were infected with powdery mildew, and 3% of the tea trees were infested with aphids. K-means clustering was used to analyze the pest and disease assessment data and generate a pest and disease distribution map. The clustering results showed that downy mildew was mainly concentrated in the low-lying areas in the south of the tea garden, while powdery mildew was concentrated in the middle of the tea garden, and aphid infestation was found more in the north of the tea garden. The average size of the pest and disease area was 20 square meters, and the area of ​​the largest pest and disease concentration area was 50 square meters.

[0102] Optionally, S1 includes:

[0103] S11, obtaining tea tree location data;

[0104] Specifically, high-precision positioning equipment such as GPS or Beidou is deployed in the tea garden to cover the planting area of ​​tea trees; a high-precision map or drone photography is used to generate a plan of the tea garden, and it is aligned and calibrated with the positioning system. The accurate coordinate information of each tea tree is obtained through the positioning equipment, and the longitude, latitude, altitude and other tea tree location information is recorded; for large-scale tea gardens, drones or ground robots are used to scan and shoot line by line to improve the efficiency of location data collection; the acquired tea tree location data is stored as the basic data for the subsequent deployment of fill light equipment.

[0105] S12, generating a photovoltaic lighting equipment placement location according to the tea tree location data to obtain photovoltaic lighting equipment placement location data;

[0106] Specifically, based on the location information of the tea trees and the surrounding terrain, the lighting conditions of the tea trees are calculated in combination with lighting simulation algorithms (such as ray tracing algorithms). The solar path model (such as the PVGIS model) is used to analyze the angle and intensity of sunlight in different time periods to determine the lighting blind spots and shadow distribution. The location range of the tea trees that need supplementary lighting is determined based on the areas with insufficient light.

[0107] Use spatial optimization algorithms (such as Voronoi diagrams or K-means clustering algorithms) to cluster the locations of tea trees and divide the areas that require supplementary lighting into several sub-areas. Select the optimal location in each sub-area to arrange photovoltaic supplementary lighting equipment to ensure that the equipment covers the tea trees to the maximum extent while reducing redundant equipment arrangements. Generate placement data for photovoltaic supplementary lighting equipment based on the sub-areas, and mark the specific coordinates and supplementary lighting intensity requirements for each device.

[0108] Specifically, the longitude, latitude and altitude information of each tea tree in the tea garden is obtained through GPS or other high-precision positioning systems to generate a set of location information points of the tea trees. Using LiDAR or drone mapping technology, the surrounding terrain data of the tea garden is generated, including terrain undulations, obstacle locations, etc. A ray tracing algorithm is used to simulate the propagation path of natural light in the tea garden. Each ray tracing path starts from the position of the sun, passes through the atmosphere and terrain obstacles, and calculates the light intensity of each tea tree. The ray tracing calculation includes direct light and reflected light, and generates the light coverage of each tea tree in different time periods. Based on the ray tracing results, the light blind spots and shadow distribution in the tea garden are marked, and the amount of light for each tea tree at different times of the day is calculated. Especially in the morning and evening, tea trees in shadow areas may face insufficient light.

[0109] Use the PVGIS model to analyze the sun path and simulate the sun's angle and intensity in different seasons and time periods. Combined with the location information of the tea trees, simulate the path of the sun from sunrise to sunset in the tea garden in one day, and analyze the impact of the sun's angle on the tea trees at each time point. Combined with the PVGIS model, calculate the light intensity of each tea tree at different time points and generate a light intensity data table. The light intensity data is in lx (lux), which represents the light intensity distribution of the tea tree during sunlight. Mark areas below a certain threshold (such as 3000lx) as light blind areas. By analyzing the light blind areas, determine the location range of tea trees that need to be supplemented with light. Pay special attention to tea trees that have been in the shadow for a long time to ensure that they receive appropriate light compensation.

[0110] The location of the tea tree is divided into several sub-areas using the Voronoi diagram algorithm, and each area corresponds to a photovoltaic lighting device. The boundary of the Voronoi diagram represents the influence range of each device, ensuring that the tea trees in each area receive uniform light compensation. Input the location coordinate data of the tea tree. Construct Voronoi polygons, with the center of each polygon being the location of the tea tree and the surrounding boundaries determined by the location of the neighboring tea trees. Each polygon represents the lighting area of ​​the tea tree, and determines the optimal photovoltaic device placement point.

[0111] As another method, the K-means clustering algorithm can be used to cluster tea trees according to their spatial locations and determine the sub-areas that need supplementary lighting. The algorithm uses the location of tea trees as data points and determines the best clustering result by minimizing the distance from each cluster center to the tea trees. Set the number of clusters K (representing the number of photovoltaic supplementary lighting devices that need to be arranged). Randomly initialize K cluster centers. Calculate the distance between each tea tree and all cluster centers and assign it to the cluster center closest to it. Update the cluster center to the mean of all tea tree locations in the current cluster. Repeat the iteration until the cluster center is stable and the sub-area is generated.

[0112] In each Voronoi polygon or K-means clustering area, the optimal photovoltaic lighting device placement position is selected according to the distribution density and light requirements of the tea trees. The optimal position is determined by minimizing the distance between the device and all tea trees in the area and maximizing the light coverage area of ​​the tea trees. Optimization function: i is the tea tree sequence item, n is the tea tree quantity data, D is the relative distance item of the equipment tea tree location, P is the equipment placement point item, T i is the location data of the i-th tea tree in the sub-region. Combined with the blind area data in the region, the fill light intensity of the photovoltaic fill light equipment is set. The fill light intensity should be able to compensate for the insufficient light of the tea tree during the shadow period. The specific fill light intensity is adjusted according to the previous ray tracing results and the insufficient light determined by the PVGIS model.

[0113] S13, arranging photovoltaic supplementary lighting equipment according to the placement data of the photovoltaic supplementary lighting equipment, and obtaining tea garden environmental data.

[0114] Specifically, based on the generated placement location data, plan the layout route of the photovoltaic lighting equipment and determine the placement point of each device; use drones or automated equipment to install the photovoltaic lighting equipment at the designated location to ensure that its lighting angle is consistent with the position of the tea trees; adjust the working hours, light intensity and other parameters of the photovoltaic equipment to ensure the effective operation of the photovoltaic lighting system.

[0115] Environmental sensors are installed near the supplementary lighting equipment to monitor the environmental parameters of the tea garden, such as temperature, humidity, light intensity, etc., in real time. The environmental data is transmitted to the data processing center through the wireless sensor network for pre-processing such as data cleaning and denoising to generate tea garden environmental data.

[0116] Optionally, S2 includes:

[0117] S21, performing operation scheduling of the supplementary lighting equipment according to the tea garden environment data to obtain operation scheduling data of the supplementary lighting equipment;

[0118] Specifically, the collected tea garden environmental data (such as temperature, humidity, light intensity, etc.) are preprocessed and then analyzed using data modeling methods. A scheduling algorithm based on rules or machine learning (such as a reinforcement learning algorithm) is used to formulate a fill lighting strategy and adjust the start, stop and working parameters of photovoltaic fill lighting equipment. The optimal start time, duration and light intensity requirements of the fill lighting equipment are analyzed to generate fill lighting equipment operation scheduling data.

[0119] Use dynamic scheduling optimization algorithms (such as genetic algorithms or particle swarm optimization algorithms) to schedule and optimize the fill light tasks of multiple devices to ensure optimal working time and resource utilization of each device; generate scheduling plans to determine the specific operation sequence and schedule of the equipment.

[0120] S22, performing photovoltaic supplementary lighting operations according to supplementary lighting equipment operation scheduling data;

[0121] Specifically, based on the generated job scheduling data, the photovoltaic fill lighting equipment is automatically turned on and the light intensity, angle and other parameters are adjusted; the operating status of the fill lighting equipment is monitored in real time, and the feedback control system is used to ensure that the equipment operates according to the predetermined plan; during the fill lighting process, the working status of the equipment is dynamically adjusted according to changes in environmental data to optimize the fill lighting effect.

[0122] S23, collecting tea tree images by using a camera device to obtain preliminary tea tree image data;

[0123] Specifically, use multi-angle, high-resolution camera equipment to take all-round photos of the tea tree area; dynamically adjust the exposure parameters, focal length and viewing angle of the camera equipment according to the effect of photovoltaic fill light to ensure image clarity. The camera equipment is triggered regularly or on demand to obtain preliminary tea tree image data.

[0124] S24, performing tea tree region segmentation on the preliminary tea tree image data according to the supplementary lighting equipment operation scheduling data and the tea garden environment data to obtain tea tree region segmentation data;

[0125] Specifically, the labeled tea tree image data is used to train a deep learning model (such as U-Net) to automatically segment the tea tree area; based on the morphology, color and lighting characteristics of the tea tree, segmentation is performed through multi-scale feature extraction to improve segmentation accuracy.

[0126] The operation scheduling data and environmental data of the fill lighting equipment are used, combined with the changes in light intensity in the image, to enhance the segmentation model's ability to recognize tea tree areas under different lighting conditions. The segmentation model is applied to the preliminary tea tree image data to extract the tea tree area and generate tea tree area segmentation data.

[0127] U-Net consists of an encoder and a decoder, and retains high-resolution features through skip connections, which is suitable for the segmentation task of tea tree images.

[0128] U-Net structure:

[0129] Input layer: The input image size is 256×256 or 512×512, depending on the resolution of the original image.

[0130] Encoder part: Contains a series of convolution blocks, each of which consists of two convolution operations (kernel size = 3x3, stride = 1, padding = 'same') and one maximum pooling operation (kernel size = 2x2, stride = 2). After each convolution, the ReLU activation function is applied, and the maximum pooling reduces the size of the feature map. The role of the encoder is to extract low-level to high-level multi-scale features.

[0131] Decoder: The decoder gradually restores the resolution of the original image through deconvolution (upsampling), and concatenates the high-resolution features of the encoder with the features of the corresponding layer of the decoder through skip connections. After deconvolution, the convolution operation is applied again to extract features and restore image details.

[0132] S25. Perform illumination adaptation image enhancement and aggregation on the tea tree region segmentation data to obtain tea tree image data.

[0133] Specifically, based on the regional segmentation data, the tea tree image is enhanced for light adaptability (such as Histogram Equalization or Retinex algorithm) to correct the image deviation caused by uneven lighting; a multi-scale brightness balance algorithm is used to unify the lighting in different areas to ensure the retention of image details.

[0134] The enhanced tea tree area images are subjected to multi-angle and multi-time point image fusion (such as gradient-based stitching or image pyramid method) to generate complete tea tree image data; during the image aggregation process, redundant information is eliminated through deduplication and optimization algorithms to ensure the integrity and accuracy of the tea tree image data.

[0135] Optionally, the tea tree region segmentation comprises:

[0136] Perform illumination compensation calculation on the preliminary tea tree image data according to the operation scheduling data of the supplementary lighting equipment and the tea garden environment data to obtain the illumination compensation data of the tea tree image;

[0137] Specifically, based on the operation scheduling data of the supplementary lighting equipment and the tea garden environment data, a lighting model is constructed to calculate the light intensity of each tea tree image at different time points; the irradiance distribution model is used to analyze the lighting unevenness of each area in the tea tree image.

[0138] An illumination compensation algorithm based on histogram equalization or adaptive histogram equalization is applied to perform local illumination equalization on the tea tree image. The shadow area in the image is enhanced using the Retinex algorithm to restore details lost due to insufficient illumination. The compensated illumination compensation data of the tea tree image is generated to provide a more uniform illumination effect.

[0139] Perform preliminary boundary detection on the illumination compensation data of the tea tree image to obtain preliminary boundary data of the tea tree;

[0140] Specifically, an edge detection algorithm (such as Canny edge detection) is used to perform preliminary boundary detection on the tea tree image after illumination compensation; the detection result is smoothed to remove noise edges while maintaining the boundaries of key features.

[0141] Morphological operations (such as dilation and erosion) are applied to further enhance the boundary contours of the tea tree trunk and leaves; boundary tracking algorithms (such as the Suzuki-ABE algorithm) are used to extract preliminary boundary data of the tea tree trunk and leaves.

[0142] According to the preliminary boundary data of the tea tree, the main leaf area of ​​the tea tree image illumination compensation data is segmented to obtain the main leaf area data;

[0143] Specifically, based on the preliminary boundary data of the tea tree, the watershed algorithm is used to segment the trunk and leaves of the tea tree in the image. The GrabCut algorithm is used to further accurately segment the trunk and leaf areas through user annotations or preliminary boundary data to ensure the complete separation of the leaves and the trunk.

[0144] The segmentation results are post-processed, such as small area filtering and boundary refinement, to ensure that the segmentation boundaries of the trunk and leaves are more accurate; trunk and leaf area data are generated, and the shape, position and size information of each area are marked.

[0145] The data of the main leaf area is occluded and repaired to obtain the segmentation data of the tea tree area.

[0146] Specifically, an occlusion detection algorithm (such as an occlusion detection model based on deep learning) is used to identify the occlusion area between the trunk and the leaves; and the overlapping area between the trunk and the leaves of the tea tree is estimated through multi-view image analysis. Using an occlusion detection model based on deep learning, such as U-Net, the model is trained to identify the occlusion area in the image. After being trained with annotated occlusion data, the model can automatically detect the occlusion between the trunk and leaves of the tea tree. The front and side images of the tea tree are input into the occlusion detection model. The model extracts features through the convolution layer, and determines whether the trunk of the tea tree is occluded by the leaves through the pooling layer and the fully connected layer. The model outputs a mask map of the occlusion area, and the occlusion area is marked as the part of the trunk of the tea tree that is occluded by the leaves. These occlusion areas are usually manifested as a sudden change in morphology, color masking, or texture superposition. Morphological processing (such as dilation and erosion operations) is used to optimize the occlusion mask map to ensure that the boundaries of the occlusion area are more accurate.

[0147] The occluded area is filled with image restoration algorithms (such as texture and structure-based restoration algorithms) to restore the details of the occluded leaves or trunks; combined with multi-angle tea tree images, pixel interpolation and texture reconstruction technology are used to further improve the restoration effect. The trunk and leaf area data after occlusion restoration are integrated to generate tea tree area segmentation data.

[0148] Optionally, the illumination compensation calculation includes:

[0149] Generating the illumination unevenness area according to the supplementary lighting equipment operation scheduling data and the tea garden environment data to obtain the illumination unevenness area data;

[0150] Specifically, according to the operation scheduling data of the supplementary lighting equipment, the changes in light intensity in each area of ​​the tea garden are calculated. The sun trajectory model is used to simulate the changes in natural light in different time periods, and the tea garden environmental data (such as tree shadows and terrain) is combined to generate areas of uneven light. Through light intensity threshold detection, the areas of uneven light are marked as highlight areas and shadow areas to obtain the data of uneven light areas.

[0151] Perform multi-scale segmentation on the preliminary tea tree image data according to the illumination inhomogeneity region data to obtain multi-scale segmentation data of the tea tree image;

[0152] Specifically, multi-scale segmentation technology (such as a segmentation algorithm based on an image pyramid) is used to perform regional segmentation on tea tree images at different resolutions; small leaves and shadow areas are segmented at a higher resolution, and large areas of uneven lighting are segmented at a lower resolution; the segmentation results at each scale are fused to obtain multi-scale segmentation data of the tea tree image.

[0153] Performing region detection based on multi-scale segmentation data of tea tree images to obtain heterogeneous region data of tea tree images;

[0154] Specifically, we use edge detection and texture analysis algorithms to identify areas with significant brightness changes in the image. We use region growing or clustering algorithms (such as K-means) to classify the uneven regions of the multi-scale segmented image (clustering algorithms compare the brightness of cluster centers for classification), and mark areas with too strong or too weak light. Based on the detection results, we generate tea tree image uneven region data and mark the light differences in each area.

[0155] Performing brightness gradient calculation on the uneven area data of the tea tree image to obtain brightness gradient data of the tea tree image;

[0156] Specifically, a gradient calculation algorithm (such as the Sobel operator or the Laplace operator) is used to calculate the brightness gradient of the uneven area, and the gradient direction and intensity change of the illumination are marked; the gradient data is smoothed to remove noise interference and ensure the accuracy of the gradient direction and magnitude. The gradient data is classified, and the areas with gradients greater than a certain threshold are marked as areas that need to be compensated, and the brightness gradient data of the tea tree image is obtained.

[0157] Performing local compensation processing on the brightness gradient data to obtain local compensation data of the tea tree image;

[0158] Specifically, the adaptive histogram equalization (CLAHE) algorithm is used to locally compensate the brightness gradient data, enhance the areas with weaker light, and compress the brightness of the overexposed areas; the brightness in each uneven area is finely adjusted to ensure the brightness consistency of the local area. After local compensation processing, the local compensation data of the tea tree image is generated to improve the image quality problem caused by uneven lighting.

[0159] Reconstructing the illumination boundary of the tea tree image local compensation data to obtain the illumination boundary data of the tea tree image;

[0160] Specifically, based on the brightness distribution after local compensation, the boundary detection algorithm is used to regenerate the illumination boundary; the illumination boundary is morphologically processed (such as closing operation) to connect the broken boundary lines to generate illumination boundary data. The illumination boundary data is optimized, and the irregular illumination boundary is smoothed through multi-level edge detection and curve fitting.

[0161] Boundary superposition processing is performed based on the local compensation data of the tea tree image and the illumination boundary data of the tea tree image to obtain the illumination compensation data of the tea tree image.

[0162] Specifically, the illumination boundary data is superimposed with the local compensation data, and the illumination boundary is adjusted in combination with the local compensation brightness information; an image fusion algorithm (such as a fusion algorithm based on the Laplace pyramid) is used to fuse the local compensation brightness information with the illumination boundary information to ensure a natural transition of the overall illumination effect. Through boundary superposition and brightness fusion processing, the illumination compensation data of the tea tree image is generated.

[0163] Optionally, S3 includes:

[0164] S31, reconstructing the main structure of tea leaves according to the tea tree image data to obtain tea leaves main structure reconstruction data;

[0165] Specifically, a deep learning segmentation network (U-Net) is used to detect and segment the trunk area in the tea tree image to identify the trunk outline of the tea tree. Multi-view geometric reconstruction technology is used to generate three-dimensional point cloud data of the trunk through tea tree images taken from different angles; structured light or multi-view stereo vision algorithms are used to register and align multi-angle images to generate a complete three-dimensional model of the tea tree trunk; the reconstructed tea tree trunk structure data contains the height, thickness and shape information of the trunk, forming the tea trunk structure reconstruction data.

[0166] S32, extracting trunk morphological features according to the tea trunk structure reconstruction data to obtain tea tree trunk feature data;

[0167] Specifically, morphological features are extracted from the reconstructed three-dimensional trunk structure, including the trunk diameter, height, inclination and branch density; morphological operations (such as expansion and erosion) are used to refine the trunk morphological structure to ensure the accuracy of the features. The central axis of the trunk is extracted through the skeleton extraction algorithm, and the curvature and bifurcation angle of the trunk are calculated; gradient analysis and curvature calculation are applied to extract the key feature points of the trunk, such as the bending point and the top point, to generate the characteristic data of the tea tree trunk.

[0168] S33, performing tea tree and tea leaf detection according to the tea tree image data to obtain tea tree and tea leaf data;

[0169] Specifically, we use a target detection model based on a convolutional neural network (such as YOLO) to detect and locate tea leaves in tea tree images; we separate tea leaves from the background based on features such as color, shape, and texture, and filter out noise areas. We generate tea tree leaf data, annotate the location information of each tea leaf (such as center point coordinates, bounding box, etc.) and detection confidence, and obtain tea tree leaf data.

[0170] YOLO (You Only Look Once): is a real-time target detection algorithm that directly predicts the bounding box and category of the target by dividing the entire image into grids. The YOLO model is fast and suitable for real-time applications. YOLO model structure Input layer: The size of the input image is 416×416 or 608×608. In order to adapt to tea tree images of different sizes, the image is scaled and normalized. Backbone network: YOLOv3 uses Darknet-53 as the backbone network. It consists of multiple layers of convolutional layers and residual connections to extract features from tea images. The backbone network extracts image features such as color, shape, and edges from low to high levels, retaining the significant feature information of tea leaves. Multi-scale feature extraction: YOLOv3 uses a multi-scale feature extraction mechanism to detect at three different scales (large, medium, and small), and can identify tea leaves of different sizes. At each scale, the feature map is convolved and upsampled to better capture small targets (such as tiny tea leaves). Output layer (Detection Head): For each feature map grid cell, the following three outputs are predicted: the four coordinates of the bounding box (center coordinates, width, height), the confidence of the object's existence, and the category probability (here, the tea category). The model outputs multiple bounding boxes for each grid cell, each of which has the center point coordinates, width and height, and category confidence.

[0171] S34, extracting tea tree and tea leaves features based on the tea tree and tea leaves data to obtain tea tree and tea leaves feature data.

[0172] Specifically, for each detected tea leaf, shape features are extracted, such as the area, perimeter, shape coefficient, etc. of the leaf; edge detection algorithms (such as Canny edge detection) are used to accurately extract the outline of the tea leaf, and the smoothness and edge complexity of the leaf are calculated. The color features of the tea leaf are extracted through color space conversion (such as RGB to HSV or Lab), and the color distribution and health status of the leaf are analyzed; texture analysis algorithms (such as gray-level co-occurrence matrix or Laplacian of Gaussian) are used to extract the texture features of the tea leaf. All the extracted shape, color, and texture features are summarized to generate tea tree tea leaf feature data.

[0173] Optionally, S4 includes:

[0174] S41, using a preset tea tree pest and disease identification model to perform pest and disease assessment on the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease assessment data;

[0175] Specifically, a pre-trained pest and disease identification model (such as a convolutional neural network (CNN) based on deep learning or a classifier based on a support vector machine) is used to input the tea tree trunk feature data and tea leaf feature data; the model is used to classify the input features to identify common types of pests and diseases, such as downy mildew and powdery mildew.

[0176] The recognition results of the model are post-processed and combined with the identified pest and disease type, area, location and other information to generate pest and disease assessment data; the impact range and type confidence of pests and diseases are assessed, and the severity of pests and diseases in each affected area is marked.

[0177] S42, locating the pest area according to the pest assessment data to obtain pest area data;

[0178] Specifically, cluster-based regional positioning algorithms (such as K-means or DBSCAN) are used to cluster the affected areas according to the location, area and type in the pest assessment data to determine the areas where pests and diseases are concentrated; the initially located pest and disease areas are expanded through the regional growing algorithm to accurately calibrate the boundaries of the pest and disease areas. Independent labels are assigned to each located pest and disease area, and the center point, boundary range and area size of the area are marked to generate pest and disease area data.

[0179] S43, calculating the severity of pests and diseases according to the pest and disease area data, the tea tree trunk characteristic data, and the tea leaf characteristic data to obtain pest and disease severity data;

[0180] Specifically, the severity calculation model based on rules or machine learning is used to calculate the severity of pests and diseases in each area, combined with pest and disease area data, trunk feature data and tea feature data; severity calculation indicators include the proportion of affected area, the degree of lesions (such as leaf wilting and discoloration), and the hazard level of pest and disease types. The comprehensive severity score is calculated for each pest and disease area, and severity data is generated. The higher the score, the more serious the impact of pests and diseases in the area.

[0181] S44. Generate pest and disease distribution data based on the pest and disease severity data and the pest and disease area data to obtain pest and disease distribution data.

[0182] Specifically, interpolation algorithms (such as Kriging interpolation) are used to spatially interpolate the severity data of pests and diseases to generate a pest and disease severity distribution map for the entire tea garden; combined with pest and disease regional data, the spatial distribution of pests and diseases is visualized, and the density of pests and diseases in different areas is marked. Pest and disease distribution data is generated, and the severity, area and species distribution of pests and diseases in different areas are output; distribution data can be used for further tea garden management decisions, such as pest and disease control measures for specific areas.

[0183] Optionally, the steps of constructing the preset tea tree pest and disease identification model include:

[0184] Obtain historical tea tree image data and corresponding historical pest and disease data;

[0185] Specifically, a large amount of tea tree image data was collected from historical tea tree pest and disease monitoring, including tea tree images under different seasons, lighting, and environmental conditions; at the same time, the corresponding pest and disease data was collected, and the pest and disease type, severity, and location in each image were annotated. The pest and disease areas in the historical images were manually annotated to ensure the accuracy of the annotated data.

[0186] Extract tea tree trunk features and tea tree leaf features based on historical tea tree image data to obtain historical tea tree trunk feature data and historical tea tree leaf feature data;

[0187] Specifically, traditional image processing techniques (such as edge detection and morphological operations) are used to extract the shape and texture features of the tea tree trunk; color segmentation and region detection algorithms are applied to detect and extract features of the tea leaves, such as color, texture, shape, etc.

[0188] A semantic-based image segmentation algorithm is used to segment the historical tea tree image data, and deep feature extraction is performed on the segmented images. Then, multi-level feature extraction is performed on the tea tree trunk and tea leaves, such as texture and edge information extracted in the convolution layer, and shape and structure information in the pooling layer.

[0189] Data is divided according to the historical tea tree trunk feature data and the historical tea tree tea leaf feature data to obtain historical tea tree feature training data and historical tea tree feature test data;

[0190] Specifically, the historical tea tree trunk characteristic data and tea leaf characteristic data are randomly divided into training sets and test sets, usually 80% of the data is used for training and 20% of the data is used for testing; ensuring that the types and severity of pests and diseases are evenly distributed during the data division process to prevent model overfitting.

[0191] A deep model was constructed based on historical tea tree characteristic training data to obtain a preliminary tea tree pest and disease identification model;

[0192] Specifically, a deep learning framework (such as TensorFlow or PyTorch) is used to design a convolutional neural network model, which includes multiple layers of convolutional layers, pooling layers, and fully connected layers; the network architecture includes modules such as ResNet or Inception to improve the depth of feature extraction and the recognition accuracy of the model.

[0193] The historical tea tree feature training data was input for model training, and the model was optimized using the cross entropy loss function and Adam optimizer. During the training process, hyperparameters such as learning rate, batch size, and number of iterations were adjusted to ensure the convergence and accuracy of the model.

[0194] The preliminary tea tree pest and disease identification model was iteratively trained based on the historical tea tree characteristic test data to obtain the tea tree pest and disease identification model.

[0195] Specifically, the preliminary model was evaluated using historical tea tree characteristic test data, and the model's accuracy, recall, F1 score and other indicators were calculated; the model's performance under different types and severity of pests and diseases was analyzed to identify the model's weak links.

[0196] The model is adjusted using feedback from model evaluation, the network structure is optimized, regularization is increased, hyperparameters are adjusted, and iterative training is continued; through multiple rounds of training and evaluation, the recognition accuracy and generalization ability of the model are continuously improved. After completing multiple rounds of iterations, the optimal parameters of the model are determined to generate a tea tree pest and disease recognition model.

[0197] Optionally, the pest severity calculation includes:

[0198] Perform pest and disease evolution simulation based on pest and disease assessment data to obtain pest and disease evolution simulation data;

[0199] Specifically, the pest evolution model based on discrete time steps (such as the SEIR model) is used to combine historical pest data and current assessment data to simulate the dynamic evolution of pests and diseases; the future evolution trend of pests and diseases is predicted by statistically analyzing the spread speed, infection rate, cure rate and other parameters of pests and diseases. The tea tree characteristic data and environmental data are used to calibrate parameters, predict the distribution and severity of pests and diseases in different time periods, and generate pest and disease evolution simulation data.

[0200] The SEIR model based on discrete time steps consists of the following four state variables: S (Susceptible): tea trees that are susceptible to infection. E (Exposed): tea trees that have been exposed but not infected (incubation period). I (Infected): tea trees infected with pests and diseases. R (Recovered): tea trees that have recovered from pests and diseases. The model simulates the spread, exposure, infection and recovery of pests and diseases by describing the transitions between state variables. The dynamic changes in the spread of pests and diseases are simulated by discrete time steps.

[0201] State variables: S(t): the number of tea trees in the susceptible state at time t. E(t): the number of tea trees in the exposed state at time t. I(t): the number of tea trees in the infected state at time t. R(t): the number of tea trees in the recovered state at time t.

[0202] The SEIR model uses a set of differential equations to describe the dynamic changes of tea trees in different states, from susceptible to exposed states: β is the infection rate term, N is the number of tea trees; the transition from exposure to infection state: σ is the latent period conversion rate term; the transition from infection to recovery state: γ is the recovery rate term, which represents the speed at which tea trees recover from pests and diseases. Changes in recovery status: I(t). The infection rate β is calibrated by environmental data (such as temperature and humidity) and tea tree characteristics (such as density and health status), indicating the spread rate of pests and diseases in the tea garden, or set by preset parameters in the system. The incubation period conversion rate σ calculates the average time from exposure to disease of tea trees from historical data, calibrates the length of the incubation period, or is set by preset parameters in the system. The recovery rate γ combines the natural recovery characteristics of pests and diseases to estimate the time from infection to recovery of tea trees, or is set by preset parameters in the system.

[0203] The model uses discrete time steps for iterative calculations, updating the number of tea trees in different states at each time step. The time step is set to daily or weekly, depending on the speed of development of tea pests and diseases. The initial state is set to S(0), E(0), I(0), and R(0), and each state corresponds to the number of tea trees in the tea garden. The spread of pests and diseases in the tea garden is simulated by iteratively updating the state variables. At each time step, S(t+1), E(t+1), I(t+1), and R(t+1) at the next moment are calculated. Statistics of propagation speed and infection rate: The model calculates the propagation speed of pests and diseases (i.e., β parameters) and estimates the spread trend of pests and diseases based on the growth rate of the number of infected tea trees in the tea garden. Through iterative calculations of discrete time steps, the model predicts the distribution and infection level of pests and diseases in the next few days or weeks. For example, the model predicts the infected area of ​​pests and diseases in the tea garden and the size of the recovery area after 10 days. According to the number of infected tea trees I(t) and the number of exposed tea trees E(t) of pests and diseases in the future, the severity of pests and diseases in different areas of the tea garden is evaluated. The data is used to formulate prevention and control measures to prevent the spread of pests and diseases in advance. Output the distribution data of pests and diseases at each time step, including the number of tea trees that are susceptible to infection, exposed, infected, and recovered. Mark the transmission path and speed of pests and diseases, and mark key areas of concern.

[0204] According to the pest and disease evolution simulation data, the pest and disease area data, tea tree trunk characteristic data and tea tree tea leaf characteristic data are simulated in time series development to obtain pest and disease time series development data;

[0205] Specifically, the simulation data of pest evolution is modeled using time series analysis methods (such as ARIMA model or LSTM neural network) to predict the development trend of pests and diseases in different time periods; according to the change trend of historical tea pest and disease data, the expansion of pests and diseases in the trunk and leaves of tea trees is simulated to generate the time series development data of pests and diseases. Time series prediction is performed on different areas of tea trees respectively, and the expansion speed and range changes of pests and diseases in the trunk and leaves are calculated to generate the time series development data of pests and diseases.

[0206] The transmission path of the pests and diseases is simulated based on the tea garden environmental data and the pest and disease area data to obtain the pest and disease transmission path data;

[0207] Specifically, we use a propagation path simulation model based on random walk or fluid dynamics, combined with tea garden environmental data (such as wind speed, humidity, temperature, etc.), to simulate the propagation path of pests and diseases in the tea garden; calculate the possibility of pests and diseases spreading from an infected area to adjacent areas, and draw the propagation path. Based on the regional data and environmental conditions of pests and diseases, we generate the propagation path data of pests and diseases, and mark the propagation direction and speed of pests and diseases from the source to the target area.

[0208] Simulate the expansion path of pests and diseases in the tea garden. Divide the tea garden into grids, each grid represents a certain number of tea trees. Update the spread of pests and diseases at each time step, and combine factors such as tea tree density, leaf density, light, humidity, etc. to simulate the dynamic evolution of pests and diseases.

[0209] Pest and disease density is calculated based on pest and disease area data, tea tree trunk characteristic data, and tea tree tea leaf characteristic data to obtain pest and disease density data;

[0210] Specifically, we use a density calculation method based on rasterization to divide the pest area into multiple small areas, and calculate the pest density in each small area. Combined with the characteristics of the tea tree trunk and leaves, we estimate the changes in pest density in each area, especially the transition between high-density pest areas and low-density pest areas. We generate pest density data, and each pest density data point corresponds to the pest density at different locations in the tea garden, reflecting the distribution of pests in different areas.

[0211] Extract pest and disease area characteristics based on pest and disease area data, tea tree trunk characteristic data, and tea tree tea leaf characteristic data to obtain pest and disease area characteristic data;

[0212] Specifically, image processing and machine learning methods are used to extract features from pest and disease area data, such as the area, shape, boundary complexity, and lesion type of the pest and disease area; the pest and disease features on the trunk and leaves of the tea tree are extracted independently to distinguish the lesion features of the trunk and leaves. The extracted pest and disease area features are summarized to generate pest and disease area feature data, and the feature information in the area, such as size, shape, and location, is annotated.

[0213] The local severity data of the region is obtained by calculating the local severity based on the time series development data of pests and diseases, the transmission path data of pests and diseases, the density data of pests and diseases, and the regional characteristic data of pests and diseases;

[0214] Specifically, the weighted average method or logistic regression model is used to calculate the local severity of different pest and disease areas in the tea garden based on multiple factors such as time series development, transmission path, density and regional characteristics; the weight of each factor is adjusted according to its impact on the severity of pests and diseases to ensure the accuracy of the calculation results. Generate local severity data and mark the severity of pests and diseases in each area of ​​the tea garden. The higher the score, the more serious the pests and diseases in the area.

[0215] The severity data of pests and diseases were obtained by spatial integration based on the regional local severity data.

[0216] Specifically, a spatial weighted interpolation algorithm (such as Kriging interpolation) is used to spatially integrate local severity data to generate pest and disease severity data for the entire tea garden; during the interpolation process, weighting is performed according to the size and location of the region to ensure that the severity of different regions has a balanced impact on the overall data. Generate pest and disease severity data, mark the overall severity of pests and diseases in the tea garden, and provide a reference for pest and disease prevention and management in the tea garden.

[0217] Optionally, the present application further provides a tea garden pest and disease monitoring system based on photovoltaic supplementary lighting, which is used to execute the tea garden pest and disease monitoring method based on photovoltaic supplementary lighting as described above, and the tea garden pest and disease monitoring system based on photovoltaic supplementary lighting includes:

[0218] Tea garden environmental data collection module, used to arrange photovoltaic lighting equipment and obtain tea garden environmental data;

[0219] The tea tree image data acquisition module is used to perform photovoltaic supplementary lighting operations according to the tea garden environmental data, and to acquire tea tree images through a camera device to obtain tea tree image data;

[0220] A tea tree feature extraction module is used to extract tea tree trunk features and tea tree leaf features according to tea tree image data, and obtain tea tree trunk feature data and tea tree leaf feature data;

[0221] The pest and disease distribution calculation module is used to evaluate pests and diseases based on the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease evaluation data; and to generate pest and disease distribution based on the pest and disease evaluation data, the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease distribution data.

[0222] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0223] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may 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 the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A tea garden pest and disease monitoring method based on photovoltaic supplementary lighting, characterized in that: The method comprises: S1. Arrange photovoltaic lighting equipment and obtain tea garden environmental data; S2, perform photovoltaic supplementary lighting operations according to the tea garden environmental data, and collect tea tree images through a camera device to obtain tea tree image data; S3, extracting tea tree trunk features and tea tree leaf features according to the tea tree image data, and obtaining tea tree trunk feature data and tea tree leaf feature data; S4. Perform pest and disease assessment based on the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease assessment data; generate pest and disease distribution based on the pest and disease assessment data, the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease distribution data.

2. The method according to claim 1, characterized in that S1 includes: Get tea tree location data; The photovoltaic lighting equipment placement position is generated according to the tea tree position data to obtain the photovoltaic lighting equipment placement position data; The photovoltaic lighting equipment is arranged according to the location data of the photovoltaic lighting equipment, and the tea garden environment data is obtained.

3. The method according to claim 1, characterized in that S2 include: Perform supplementary lighting equipment operation scheduling according to tea garden environmental data to obtain supplementary lighting equipment operation scheduling data; Carry out photovoltaic lighting operations according to the lighting equipment operation scheduling data; Collect tea tree images through camera equipment to obtain preliminary tea tree image data; Perform tea tree region segmentation on the preliminary tea tree image data according to the supplementary lighting equipment operation scheduling data and the tea garden environment data to obtain tea tree region segmentation data; The tea tree area segmentation data is subjected to illumination adaptation image enhancement and aggregation to obtain tea tree image data.

4. The method according to claim 3, characterized in that The tea tree area segmentation includes: Perform illumination compensation calculation on the preliminary tea tree image data according to the operation scheduling data of the supplementary lighting equipment and the tea garden environment data to obtain the illumination compensation data of the tea tree image; Perform preliminary boundary detection on the illumination compensation data of the tea tree image to obtain preliminary boundary data of the tea tree; According to the preliminary boundary data of the tea tree, the main leaf area of ​​the tea tree image illumination compensation data is segmented to obtain the main leaf area data; The data of the main leaf area is occluded and repaired to obtain the segmentation data of the tea tree area.

5. The method according to claim 4, characterized in that The illumination compensation calculation includes: Generating the illumination unevenness area according to the operation scheduling data of the supplementary lighting equipment and the tea garden environment data to obtain the illumination unevenness area data; Perform multi-scale segmentation on the preliminary tea tree image data according to the illumination inhomogeneity region data to obtain multi-scale segmentation data of the tea tree image; Performing region detection based on multi-scale segmentation data of tea tree images to obtain heterogeneous region data of tea tree images; Performing brightness gradient calculation on the uneven area data of the tea tree image to obtain brightness gradient data of the tea tree image; Performing local compensation processing on the brightness gradient data to obtain local compensation data of the tea tree image; Reconstructing the illumination boundary of the tea tree image local compensation data to obtain the illumination boundary data of the tea tree image; Boundary superposition processing is performed based on the local compensation data of the tea tree image and the illumination boundary data of the tea tree image to obtain the illumination compensation data of the tea tree image.

6. The method according to claim 1, characterized in that S3 include: Reconstructing the main structure of tea leaves according to the tea tree image data to obtain the main structure reconstruction data of tea leaves; Extract the trunk morphological features based on the tea trunk structure reconstruction data to obtain the tea tree trunk feature data; Perform tea tree and tea leaf detection according to the tea tree image data to obtain tea tree and tea leaf data; Tea tree and tea leaves features are extracted based on the tea tree and tea leaves data to obtain tea tree and tea leaves feature data.

7. The method according to claim 1, characterized in that S4 include: Using a preset tea tree pest and disease identification model to evaluate the tea tree trunk characteristic data and tea leaf characteristic data, to obtain pest and disease evaluation data; Locate the pest area based on the pest assessment data to obtain pest area data; The severity of pests and diseases is calculated based on the pest and disease area data, the tea tree trunk characteristic data and the tea tree tea leaf characteristic data to obtain the pest and disease severity data; The pest distribution is generated based on the pest severity data and the pest area data to obtain the pest distribution data.

8. The method according to claim 7, characterized in that The steps for constructing the preset tea tree pest and disease identification model include: Obtain historical tea tree image data and corresponding historical pest and disease data; Extract tea tree trunk features and tea tree leaf features based on historical tea tree image data to obtain historical tea tree trunk feature data and historical tea tree leaf feature data; Data is divided according to the historical tea tree trunk feature data and the historical tea tree tea leaf feature data to obtain historical tea tree feature training data and historical tea tree feature test data; A deep model was constructed based on historical tea tree characteristic training data to obtain a preliminary tea tree pest and disease identification model; The preliminary tea tree pest and disease identification model was iteratively trained based on the historical tea tree characteristic test data to obtain the tea tree pest and disease identification model.

9. The method according to claim 7, characterized in that: The calculation of pest severity includes: Perform pest and disease evolution simulation based on pest and disease assessment data to obtain pest and disease evolution simulation data; According to the pest and disease evolution simulation data, the pest and disease area data, tea tree trunk characteristic data and tea tree tea leaf characteristic data are simulated in time series development to obtain pest and disease time series development data; The transmission path of the pests and diseases is simulated based on the tea garden environmental data and the pest and disease area data to obtain the pest and disease transmission path data; Pest and disease density is calculated based on pest and disease area data, tea tree trunk characteristic data, and tea tree tea leaf characteristic data to obtain pest and disease density data; Extract pest and disease area characteristics based on pest and disease area data, tea tree trunk characteristic data, and tea tree tea leaf characteristic data to obtain pest and disease area characteristic data; The local severity data of the region is obtained by calculating the local severity based on the time series development data of pests and diseases, the transmission path data of pests and diseases, the density data of pests and diseases, and the regional characteristic data of pests and diseases; The severity data of pests and diseases were obtained by spatial integration based on the regional local severity data.

10. A tea garden pest and disease monitoring system based on photovoltaic supplementary lighting, characterized in that: Used to execute the tea garden pest and disease monitoring method based on photovoltaic supplementary lighting as claimed in claim 1, the tea garden pest and disease monitoring system based on photovoltaic supplementary lighting comprises: Tea garden environmental data collection module, used to arrange photovoltaic lighting equipment and obtain tea garden environmental data; The tea tree image data acquisition module is used to perform photovoltaic supplementary lighting operations according to the tea garden environmental data, and to acquire tea tree images through a camera device to obtain tea tree image data; A tea tree feature extraction module is used to extract tea tree trunk features and tea tree leaf features according to tea tree image data, and obtain tea tree trunk feature data and tea tree leaf feature data; The pest and disease distribution calculation module is used to evaluate pests and diseases based on the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease evaluation data; and to generate pest and disease distribution based on the pest and disease evaluation data, the tea tree trunk characteristic data and the tea leaf characteristic data to obtain pest and disease distribution data.

Images