A method and system for monitoring and early warning of pests and diseases of white peony root and oil tea intercropping

Abstract

This invention discloses a method and system for monitoring and early warning of pests and diseases in intercropping of white peony and camellia oleifera, mainly relating to the field of data processing technology. It includes: acquiring site environmental data and basic intercropping data for the white peony-camellia intercropping plot; constructing a three-dimensional spatial structure module containing the canopy layer, white peony layer, and ground surface layer; obtaining pest and disease monitoring paths and a set of monitoring viewpoints; obtaining a set of periodically acquired image sequences from the monitoring viewpoints; performing implicit iterative semantic analysis of pests and diseases within the sequences; and using the pest and disease monitoring paths as indexes, performing anomaly analysis on the iterative feature set of pests and diseases at the monitoring viewpoints. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring and early warning information is obtained. The beneficial effects of this invention are: solving the technical problem in the prior art of lacking linkage analysis in pest and disease monitoring of white peony and camellia oleifera intercropping, leading to delayed monitoring and early warning, and achieving the technical effect of improving the timeliness and reliability of monitoring and early warning response.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for monitoring and early warning of pests and diseases in intercropping of white peony and camellia oleifera. Background Technology

[0002] Intercropping white peony with camellia oleifera has been widely adopted in agricultural production in recent years because it effectively improves land utilization, enhances ecological benefits, and the camellia oleifera trees provide good shade, which is beneficial to the growth of white peony. However, pest and disease control is a key issue during intercropping, especially in complex agricultural environments where pest and disease monitoring and control are often affected by environmental changes and vegetation species. Traditional pest and disease monitoring methods usually rely on manual inspections or monitoring using single sensor devices.

[0003] Currently, traditional manual inspection methods are costly and inefficient, failing to achieve comprehensive, 24 / 7 monitoring. Secondly, existing monitoring equipment typically cannot simultaneously acquire information on the spatial distribution of pests and diseases and the detailed characteristics of lesions, resulting in insufficient accuracy and comprehensiveness of the information. Therefore, current pest and disease monitoring often lacks a dynamic update mechanism; once changes occur, timely feedback cannot be obtained, leading to missed opportunities for optimal control.

[0004] The existing technology lacks a linkage analysis for monitoring pests and diseases in the intercropping of white peony and camellia, resulting in a technical problem of delayed monitoring and early warning. Summary of the Invention

[0005] This application provides a method and system for monitoring and early warning of pests and diseases in the intercropping of white peony and camellia oleifera, which is used to address the technical problem that the lack of linkage analysis in the monitoring of pests and diseases in the intercropping of white peony and camellia oleifera in the prior art leads to the lag in monitoring and early warning.

[0006] In view of the above problems, this application provides a method and system for monitoring and early warning of diseases and pests in the intercropping of white peony and camellia oleifera.

[0007] The first aspect of this application provides a method for monitoring and early warning of pests and diseases in intercropping of white peony and camellia oleifera, the method comprising:

[0008] Environmental data and basic data on the intercropping of white peony and camellia are obtained to construct a three-dimensional spatial structure module containing the canopy layer, white peony layer, and ground surface layer. Using the camera parameters of a monitoring drone as constraints, monitoring path and viewpoint identification are performed on the three-dimensional spatial structure module to obtain a set of pest and disease monitoring paths and viewpoints. The monitoring drone is used to periodically monitor pests and diseases based on the monitoring paths and viewpoints to obtain a set of periodically acquired image sequences from the monitoring viewpoints. Implicit iterative semantic analysis of pests and diseases within the sequence is performed on the set of periodically acquired image sequences from the monitoring viewpoints to determine the iterative feature set of pests and diseases at each monitoring viewpoint. Using the pest and disease monitoring paths as indexes, anomaly analysis of pest and disease linkage is performed on the iterative feature set of pests and diseases at each monitoring viewpoint. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring early warning information is obtained.

[0009] In one possible implementation, using the camera parameters of the monitoring drone as constraints, the three-dimensional spatial structure module is used to identify monitoring paths and monitoring viewpoints to obtain a pest and disease monitoring path and a set of monitoring viewpoints. This includes: analyzing the canopy height distribution of the three-dimensional spatial structure module to obtain canopy height distribution characteristics; based on the canopy height distribution characteristics and flight altitude constraints in the camera parameters, analyzing the monitoring path of the three-dimensional spatial structure module to obtain a pest and disease monitoring path; extracting the focal length and pitch angle bandwidth from the camera parameters, and configuring monitoring viewpoints for the pest and disease monitoring path in conjunction with the three-dimensional spatial structure module to obtain a set of monitoring viewpoints.

[0010] In one possible implementation, the canopy height distribution of the three-dimensional spatial structure module is analyzed to obtain canopy height distribution characteristics, including: dividing the three-dimensional spatial structure module into multiple grids according to a preset grid size; based on the canopy information of the three-dimensional spatial structure module, performing canopy height mean processing on the multiple grids to determine the canopy height mean of the multiple grids; and performing mean drift distribution analysis on the canopy height mean of the multiple grids to determine the canopy height distribution characteristics.

[0011] In one possible implementation, the focal length and pitch angle bandwidth of the camera parameters are extracted, and the monitoring viewpoints of the pest and disease monitoring path are configured in conjunction with the three-dimensional spatial structure module to obtain a set of monitoring viewpoints. This includes: determining the monitoring viewpoint interval bandwidth based on the focal length and pitch angle bandwidth; taking the starting point of the pest and disease monitoring path as the initial first monitoring viewpoint, and performing texture quality certification on the initial first monitoring viewpoint. If the texture quality certification passes, the initial first monitoring viewpoint is used as the first monitoring viewpoint. The texture quality certification includes quality certification in four dimensions: perspective quality, luminance quality, structural exclusivity, and visual integrity; using the monitoring viewpoint interval bandwidth as a constraint, and combining it with the first monitoring viewpoint, determining the initial second monitoring viewpoint, and performing texture quality certification on the initial second monitoring viewpoint. If the texture quality certification passes, the initial second monitoring viewpoint is used as the second monitoring viewpoint, and so on, to obtain a set of monitoring viewpoints.

[0012] In one possible implementation, the starting point of the pest and disease monitoring path is used as the initial first monitoring viewpoint. Texture quality authentication of the initial first monitoring viewpoint is performed, including: if the texture quality authentication fails, the initial first monitoring viewpoint is offset according to the flight path offset bandwidth of the monitoring drone to determine a set of candidate first monitoring viewpoints; the set of candidate first monitoring viewpoints is traversed to perform texture quality authentication, and if there is a candidate first monitoring viewpoint that passes the texture quality authentication, it is added to the set of optional first monitoring viewpoints; the optional first monitoring viewpoint with the highest texture quality in the set of optional first monitoring viewpoints is used as the first monitoring viewpoint.

[0013] In one possible implementation, an intra-sequence implicit iterative analysis of pest and disease semantics is performed on the periodically acquired image sequence set of the monitoring viewpoint to determine the monitoring viewpoint pest and disease iterative feature set. This includes: performing pest and disease semantic parsing on each periodically acquired image in the periodically acquired image sequence set of the monitoring viewpoint to determine a pest and disease semantic parsing information sequence set; extracting a first pest and disease semantic parsing information sequence from the pest and disease semantic parsing information sequence set, performing intra-sequence implicit iterative analysis of pest and disease semantics to obtain a first monitoring viewpoint pest and disease iterative feature; and adding the first monitoring viewpoint pest and disease iterative feature to the monitoring viewpoint pest and disease iterative feature set.

[0014] In one possible implementation, extracting a first pest semantic parsing information sequence from the pest semantic parsing information sequence set, performing implicit iterative analysis of pest semantics within the sequence, and obtaining iterative features of pests at a first monitoring viewpoint includes: performing pest feature analysis on the first pest semantic parsing information sequence to obtain a first pest feature sequence; extracting the first and second pest features from the first pest feature sequence and performing implicit iterative analysis to obtain a second iterative pest feature; using the second iterative pest feature to perform implicit iterative analysis on the third pest feature in the first pest feature sequence to obtain a third iterative pest feature; and so on, using the third iterative pest feature to perform implicit iterative analysis on the first pest feature sequence to obtain iterative features of pests at the first monitoring viewpoint.

[0015] In one possible implementation, the first and second pest features in the first pest feature sequence are extracted and implicitly iteratively analyzed to obtain the second iterative pest feature. This includes: calculating the Euclidean distance between features of the same type in the first and second pest features to obtain a set of Euclidean distances; normalizing the set of Euclidean distances to construct an implicit iterative analysis matrix; and iterating the second pest feature using the implicit iterative analysis matrix to obtain the second iterative pest feature.

[0016] In one possible implementation, using the pest and disease monitoring path as an index, a pest and disease linkage anomaly analysis is performed on the pest and disease iteration feature set of the monitoring viewpoint. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring early warning information is obtained. This includes: traversing the pest and disease iteration feature set of the monitoring viewpoint to identify pest and disease anomalies and obtain an abnormal monitoring viewpoint pest and disease iteration feature set; based on the pest and disease monitoring path, identifying the linkage cluster density of the abnormal monitoring viewpoint pest and disease iteration feature set and obtaining the linkage cluster density; if the linkage cluster density is greater than or equal to a preset threshold, then the abnormal monitoring viewpoint pest and disease iteration feature set and the linkage cluster density are used as the anomaly analysis result; if the linkage cluster density is less than the preset threshold, then the abnormal monitoring viewpoint pest and disease iteration feature set is used as the anomaly analysis result.

[0017] A second aspect of this application provides a pest and disease monitoring and early warning system for intercropping white peony and camellia oleifera, the system comprising:

[0018] The system includes a three-dimensional spatial structure module for acquiring site environmental data and intercropping basic data for peony-camellia intercropping plots, and constructing a three-dimensional spatial structure module containing a canopy layer, a peony layer, and a ground surface layer; a monitoring viewpoint identification module for identifying monitoring paths and monitoring viewpoints of the three-dimensional spatial structure module, constrained by the camera parameters of the monitoring drone, to obtain a set of pest and disease monitoring paths and monitoring viewpoints; a periodic monitoring module for using the monitoring drone to conduct periodic pest and disease monitoring based on the monitoring paths and monitoring viewpoints, obtaining a set of periodically acquired image sequences for the monitoring viewpoints; an implicit iterative analysis module for performing implicit iterative semantic analysis of pest and disease within the sequence of the periodically acquired image sequences for the monitoring viewpoints, determining a set of iterative pest and disease features for the monitoring viewpoints; and a monitoring and early warning information acquisition module for performing pest and disease linkage anomaly analysis on the iterative feature set of pests and diseases for the monitoring viewpoints, using the pest and disease monitoring paths as indexes, and obtaining pest and disease monitoring and early warning information if the anomaly analysis result indicates the presence of anomalies.

[0019] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0020] This application acquires site environmental data and intercropping basic data for peony-camellia intercropping plots, constructs a three-dimensional spatial structure module including the canopy layer, peony layer, and surface layer. Then, using the camera parameters of a monitoring drone as constraints, it identifies monitoring paths and viewpoints within the three-dimensional spatial structure module, obtaining a set of pest and disease monitoring paths and viewpoints. The monitoring drone then conducts periodic pest and disease monitoring based on these paths and viewpoints, obtaining a set of periodically acquired image sequences from each viewpoint. Next, it performs implicit iterative semantic analysis of the periodically acquired image sequences within the sequences to determine the iterative feature set of pests and diseases at each viewpoint. Using the pest and disease monitoring path as an index, it performs anomaly analysis on the iterative feature set of pests and diseases at each viewpoint. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring and early warning information is obtained. This achieves the technical effect of analyzing the actual situation of intercropped plots and improving the reliability of monitoring and early warning. Attached Figure Description

[0021] Appendix Figure 1 This is a schematic diagram of a pest and disease monitoring and early warning method for intercropping white peony and camellia oleifera according to an embodiment of the present invention.

[0022] Appendix Figure 2 This is a schematic diagram of a pest and disease monitoring and early warning system for intercropping white peony and camellia oleifera provided in an embodiment of the present invention.

[0023] The labels shown in the attached diagram:

[0024] The module includes a three-dimensional spatial structure construction module 11, a monitoring viewpoint identification module 12, a periodic monitoring module 13, an implicit iterative analysis module 14, and a monitoring and early warning information acquisition module 15. Detailed Implementation

[0025] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims. It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices.

[0026] Example 1, as shown in the appendix Figure 1 As shown, this application provides a method for monitoring and early warning of pests and diseases in intercropping of white peony and camellia, wherein the method includes:

[0027] S100: Obtain plot environmental data and intercropping basic data for the white peony-camellia intercropping plots, and construct a three-dimensional spatial structure module including the canopy layer, white peony layer and surface layer;

[0028] In one embodiment, site environmental data is used to describe the basic environmental conditions of the site, such as soil type, topography, and climate conditions. Intercropping basic data is used to describe agronomic data related to the intercropping of white peony and camellia.

[0029] Preferred intercropping data includes: Positioning planting holes 4-5 meters apart in the center of the ridge, digging holes 80cm in diameter and 80cm deep; filling the holes with 300-400 catties / mu of well-rotted organic fertilizer (equivalent to 8.5-11.5 catties / hole based on 35 trees / mu); and 8 catties / hole of compound fertilizer (equivalent to 280 catties / mu). Mix these with the soil at the bottom of the hole to form a 20cm thick fertile soil layer. Place the camellia seedlings in the holes, spreading out the roots, and backfill with soil to 2-3cm below the grafting point, gently compacting the soil. After planting, water each seedling with 5L of water to ensure close contact between the roots and the soil. After watering, cover the trunk with 5-8cm of straw or wheat straw to reduce water evaporation. Specifically, 35 camellia trees are planted per acre, with a spacing of 4-5 meters between trees and 4.5-5 meters between rows, to ensure sufficient growing space for the trees and a stable shading rate of 40% for the leaves in the later stages.

[0030] Simultaneously, plant white peony on the ridges between the rows of camellia trees, with the planting strip at least 1.5m away from the camellia tree trunks. Plant at a density of 6000-9000 plants per acre using row sowing. Dig planting trenches 15-18cm deep along the ridges, with a trench spacing of 20-25cm. Place peony seedlings in the trenches at a spacing of 8-12cm, backfill the soil to 2-3cm above the roots, and compact it. After planting, use drip irrigation. Lay drip irrigation tape along the planting trenches with drippers spaced 15cm apart, and water each peony with 0.5-1L of water, ensuring the soil is moist but not waterlogged.

[0031] Through on-site surveys, high-precision GPS was used to collect topographic features of the plot, such as slope and soil type. Based on intercropping data, the distribution, canopy height, and row spacing of camellia oleifera and peony plants were recorded. Based on the collected data, a three-dimensional spatial structure model of the plot was constructed using 3D modeling software. Specifically, based on the canopy height and distribution of the camellia oleifera trees, a canopy layer model was constructed using GPS data and on-site measurements, considering the shape, density, and shading characteristics of the canopy. The canopy layer model can express the height, crown width, and crown morphology of each tree using 3D modeling software. The planting density, leaf distribution, and plant height of the peony were used to construct the peony layer model. Through measurement and data processing, the spatial distribution, plant height, and row spacing of each peony plant were determined, thus establishing the three-dimensional structure of the peony layer. Through on-site measurements and remote sensing data, the surface features of the plot were determined, including the growth areas of herbaceous plants, waterlogged areas, and soil type. Thus, a three-dimensional spatial structure module containing the canopy layer, peony layer, and surface layer was constructed.

[0032] S200: Using the camera parameters of the monitoring drone as constraints, the three-dimensional spatial structure module is used to identify the monitoring path and the monitoring viewpoint to obtain the pest and disease monitoring path and the set of monitoring viewpoints;

[0033] Furthermore, using the camera parameters of the monitoring drone as constraints, the three-dimensional spatial structure module is subjected to monitoring path identification and monitoring viewpoint identification to obtain the pest and disease monitoring path and monitoring viewpoint set. Step S200 in this embodiment further includes:

[0034] The canopy height distribution of the three-dimensional spatial structure module is analyzed to obtain the canopy height distribution characteristics;

[0035] Based on the height distribution characteristics of the canopy layer and the flight height constraint in the camera parameters, the monitoring path of the three-dimensional spatial structure module is analyzed to obtain the pest and disease monitoring path;

[0036] Extract the focal length and pitch angle bandwidth from the camera parameters, and combine them with the three-dimensional spatial structure module to configure the monitoring viewpoints for the pest and disease monitoring path, thereby obtaining a set of monitoring viewpoints.

[0037] In one embodiment, the camera parameters are various parameters set by the UAV's camera equipment, including flight altitude, focal length, pitch angle, and shooting angle, used to determine the image range captured by the UAV. A detailed analysis of the canopy height distribution is performed using data from the camellia tree canopy in a three-dimensional spatial model. Canopy height distribution analysis can reveal dense and sparse areas of the canopy, helping to avoid missing pest and disease monitoring areas in the white peony layer due to canopy shading. Furthermore, by combining the obtained canopy height distribution characteristics with the flight altitude constraints in the camera parameters, the UAV's flight path is determined.

[0038] Preferably, multiple samples of canopy height distribution features, multiple samples of flight height constraints, multiple samples of 3D spatial structure modules, and multiple samples of pest and disease monitoring paths are obtained as training data to construct a training framework based on a feedforward neural network, comprising an input layer, hidden layers, and an output layer. The input layer receives the input training data, including canopy height distribution features, flight height constraints, and 3D spatial structure modules, which are encoded and then input into the neural network. The hidden layer processes the input data. It uses activation functions such as ReLU or Sigmoid for nonlinear transformation to capture complex relationships in the data. Through the connection of multiple layers of neurons, the network can learn the implicit relationship between the input features and the pest and disease monitoring paths. The output layer outputs the pest and disease monitoring paths.

[0039] By optimizing the loss function, the neural network can gradually learn the optimal mapping relationship between input features and output. Training data from the input layer is passed through each layer of the neural network, processed by activation functions, and finally generated as predicted values ​​at the output layer. These predicted values ​​represent the network's estimation of pest and disease monitoring paths. Mean squared error is used as the loss function to evaluate the error between the predicted values ​​and the actual targets. The gradient of the loss function with respect to each weight in the neural network is calculated using the backpropagation algorithm. Then, gradient descent algorithms, such as Adam or SGD, are used to update the weights based on the gradients, allowing the network's output to gradually approach the true values. The backpropagation process is performed in each training iteration until the error between the network's predicted and actual values ​​reaches an acceptable range, resulting in a trained monitoring path analyzer. The trained monitoring path analyzer is then used to analyze the monitoring paths of the three-dimensional spatial structure module to obtain pest and disease monitoring paths. The input data includes canopy height distribution characteristics, flight height constraints in camera parameters, and the three-dimensional spatial structure module; the output data is the pest and disease monitoring path.

[0040] Furthermore, based on the focal length and pitch angle bandwidth in the camera parameters, the area that each monitoring viewpoint can cover is determined. Then, combined with the actual situation of the intercropping plot reflected by the three-dimensional spatial structure module, the monitoring viewpoints of the pest and disease monitoring path are configured to obtain a set of monitoring viewpoints. If the focal length is short and the pitch angle is large, the interval between viewpoints is small, and more monitoring points may be needed to cover the entire area. Conversely, if the focal length is long and the pitch angle is small, the interval between viewpoints may be large, resulting in a wider coverage area. By optimizing these parameters, it can be ensured that the monitoring viewpoints can maximize coverage of all key areas of the intercropping plot, especially the white peony layer under the canopy.

[0041] Furthermore, the canopy height distribution of the three-dimensional spatial structure module is analyzed to obtain the canopy height distribution characteristics. In this embodiment, step S200 further includes:

[0042] The three-dimensional spatial structure module is divided according to a preset grid size to obtain multiple grids;

[0043] Based on the canopy layer information of the three-dimensional spatial structure module, the average height of the canopy layer within the grid is processed to determine the average height of the canopy layer of the multiple grids;

[0044] The mean shift distribution of the average height of the canopy layer in the multiple grid cells is analyzed to determine the distribution characteristics of the canopy layer height.

[0045] In one possible embodiment, the preset grid size is the size of the area to be divided, pre-defined by those skilled in the art based on the actual area of ​​the intercropping plot. For example, a 1m x 1m grid size is selected, with each grid representing a 1m² area. The canopy height within each grid is averaged. Canopy height varies within the plot, possibly due to differences in tree density, species, or growth status. By calculating the average canopy height within each grid cell, the overall canopy height information for that area can be obtained. Assuming the canopy heights within a grid are 5m, 6m, and 7m, the average canopy height for that grid is 6m. For larger plots, this processing needs to be performed on each grid to obtain the canopy height distribution across the entire area. After averaging the canopy height, a mean-shift algorithm is used to analyze the average canopy heights of multiple grids. This identifies the distribution patterns in the data, thereby extracting the distribution characteristics of the canopy height. Through mean-shift analysis, the canopy height distribution can be reliably identified. For example, the average canopy height is high in some areas, indicating that the trees in these areas are relatively tall and may cause significant shading, requiring a lower monitoring height; while the average canopy height is low in some areas, indicating that the shading of the lower peony layer is not severe, allowing for higher monitoring, thereby expanding the scope of a single monitoring session. This provides data support for subsequent monitoring path identification, achieving the technical effect of improving the speed of pest and disease monitoring and early warning response.

[0046] Preferably, one average grid canopy height is randomly selected from the plurality of average grid canopy heights as the first distribution resolution starting point. The average grid canopy heights of those grid canopy heights whose distance to the first distribution resolution starting point is within a preset distribution resolution distance range pre-defined by those skilled in the art are added to the neighborhood to obtain the neighborhood of the first distribution resolution starting point.

[0047] The number of times the difference between the average grid canopy height of the first distribution resolution starting point neighborhood and the first distribution resolution starting point falls within a pre-defined range by those skilled in the art is counted to obtain the number of neighborhoods of the first distribution resolution starting point. According to the preset distribution resolution distance range, edge diffusion is performed on the neighborhood of the first distribution resolution starting point. The average grid canopy heights of multiple grid canopy heights, each within the preset distribution resolution distance range, are added to the neighborhood of the first distribution resolution starting point to obtain the first diffusion distribution resolution starting point neighborhood. Similarly, the number of times the difference between the average grid canopy height of the first diffusion distribution resolution starting point neighborhood and the first distribution resolution starting point falls within a pre-defined range by those skilled in the art is counted to obtain the number of neighborhoods of the first diffusion distribution resolution starting point. The number of neighborhoods of the first distribution resolution starting point and the number of neighborhoods of the first diffusion distribution resolution starting point reflect the degree of height approximation between the corresponding neighborhood and the first distribution resolution starting point. A higher degree of height approximation indicates a more uniform distribution of the neighborhood, making it suitable for path identification during monitoring. The lower the degree of similarity, the more uneven the distribution of the neighborhood, requiring further subdivision to improve the accuracy of monitoring path identification. This approach achieves the goal of determining neighborhood edges from two dimensions: location and mean difference.

[0048] When the number of neighborhoods at the starting point of the first distribution resolution is less than the number of neighborhoods at the starting point of the first diffusion distribution resolution, diffusion continues in the neighborhoods at the starting point of the first diffusion distribution resolution, following the same principle as the diffusion described above, until the number of neighborhoods obtained in this diffusion is greater than or equal to the number of neighborhoods obtained in the previous diffusion. Diffusion then stops, and the neighborhoods at the starting point of the first target distribution resolution are obtained. The mean of the average heights of multiple raster canopy layers within the neighborhoods at the starting point of the first target distribution resolution is calculated, and the regional location and the calculated mean of the neighborhoods at the starting point of the first target distribution resolution are used as the first distribution feature.

[0049] Similarly, another grid canopy height average is randomly selected from the multiple grid canopy height averages as the second distribution analysis starting point. After mean drift analysis, the second distribution feature is obtained. This second distribution feature includes the regional location and calculated mean of the neighborhood of the second target distribution analysis starting point. This process is repeated multiple times until all grid canopy height averages are assigned to neighborhoods, resulting in multiple distribution analysis starting points and multiple distribution features. These multiple distribution features are then summarized to obtain the canopy height distribution feature. Furthermore, feeding the canopy height distribution feature back into the pest and disease monitoring path planning allows for more precise selection of monitoring viewpoints and optimization of the monitoring drone's flight path.

[0050] Furthermore, the focal length and pitch angle bandwidth are extracted from the camera parameters, and the monitoring viewpoints are configured for the pest and disease monitoring path in conjunction with the three-dimensional spatial structure module to obtain a set of monitoring viewpoints. Step S200 in this embodiment of the application also includes:

[0051] Determine the monitoring viewpoint interval bandwidth based on focal length and pitch angle bandwidth;

[0052] The starting point of the pest and disease monitoring path is taken as the initial first monitoring viewpoint. The texture quality of the initial first monitoring viewpoint is certified. If the texture quality certification is passed, the initial first monitoring viewpoint is taken as the first monitoring viewpoint. The texture quality certification includes quality certification in four dimensions: perspective quality, luminance quality, structural exclusivity, and visual integrity.

[0053] Using the monitoring viewpoint interval bandwidth as a constraint, an initial second monitoring viewpoint is determined in combination with the first monitoring viewpoint. Texture quality authentication is performed on the initial second monitoring viewpoint. If the texture quality authentication is successful, the initial second monitoring viewpoint is used as the second monitoring viewpoint, and so on, to obtain a set of monitoring viewpoints.

[0054] Furthermore, taking the starting point of the pest and disease monitoring path as the initial first monitoring viewpoint, and performing texture quality authentication on the initial first monitoring viewpoint, step S200 of this embodiment further includes:

[0055] If the texture quality certification fails, the initial first monitoring viewpoint is offset according to the flight path offset bandwidth of the monitoring drone to determine the candidate first monitoring viewpoint set;

[0056] The candidate first monitoring viewpoint set is traversed for texture quality authentication. If there is a candidate first monitoring viewpoint that has passed texture quality authentication, it is added to the selectable first monitoring viewpoint set.

[0057] The first monitoring viewpoint is selected as the one with the highest texture quality from the set of selectable first monitoring viewpoints.

[0058] It's important to note that pitch angle bandwidth refers to the range of angles the drone camera can capture vertically, affecting the image coverage. A larger pitch angle bandwidth allows the camera to capture a wider field of view. Monitoring viewpoint interval bandwidth refers to the distance between two viewpoints selected along the monitoring path. Perspective quality ensures the alignment of the image's shooting angle with the target area, avoiding image distortion due to angular deviations. Luminosity quality ensures that the image's brightness, contrast, and sharpness meet monitoring requirements, preventing images that are too bright or too dark from affecting analysis. Structure exclusivity assesses the clarity of camellia trees and white peony in the image, ensuring structural details are captured. Visual integrity ensures the image captures a sufficiently wide area, avoiding missing crucial monitoring areas. Offset bandwidth refers to adjusting the viewpoint according to a preset offset range when the initial monitoring viewpoint's texture quality certification fails. This offset operation helps find the most suitable monitoring viewpoint, ensuring a final image of acceptable quality.

[0059] The interval bandwidth between monitoring viewpoints is calculated based on the focal length and pitch angle bandwidth. The longer the focal length and the larger the pitch angle bandwidth, the greater the interval between monitoring viewpoints can be, in order to cover a larger area and reduce redundant data acquisition. For example, if the focal length is long and the pitch angle is large, the viewpoint interval might be set to 100 meters, while for a shorter focal length, the viewpoint interval could be set to 50 meters.

[0060] Starting from the beginning of the pest and disease monitoring path, select the first monitoring viewpoint and perform texture quality certification on it. Certification includes four aspects: perspective quality, photometric quality, structural exclusivity, and visual integrity. If the certification passes, the viewpoint will be considered a valid monitoring viewpoint. If the initially selected viewpoint is located under the canopy of a camellia tree, there may be obstruction, in which case the texture quality certification may fail. The reason for failure may be perspective quality issues due to angle problems or photometric problems due to uneven lighting.

[0061] In one possible embodiment, perspective quality is analyzed by detecting whether the shooting angle is accurately aligned with the peony layer and whether lesions and insect infestations are clearly presented. If the angle of the monitoring viewpoint deviates from the lesion area, the perspective quality of the image may be substandard. Light quality is analyzed by evaluating the lighting conditions of the image; if the lighting is too strong or too weak during shooting, it may affect the identification of pests and diseases. For example, strong light may cause overexposure of the lesion area, while dim light may make the details of pests and diseases unclear. The sharpness of the peony and camellia tree areas in the image is checked; if the sharpness is low, making it impossible to clearly identify pest and disease features, the structural exclusivity will be low, and the image quality certification will fail. The completeness of the peony and camellia tree areas in the image is determined; if the image does not cover the entire peony and camellia tree area due to viewing angle issues or is severely obscured, the visual integrity may be substandard.

[0062] If the texture quality certification of the initial viewpoint fails, its position can be adjusted based on the flight path offset bandwidth of the monitoring drone. The adjusted viewpoint set becomes a candidate viewpoint set, which will be traversed and certified subsequently. Texture quality certification is performed on each viewpoint in the candidate viewpoint set; if a viewpoint passes certification, it is added to the optional viewpoint set. By comparing all certified viewpoints in the optional viewpoint set, the one with the highest texture quality is selected as the final monitoring viewpoint, iteratively obtaining the complete monitoring viewpoint set. By progressively selecting and certifying all monitoring viewpoints, the image quality of each viewpoint is ensured to be acceptable. A reasonably spaced set of monitoring viewpoints is formed along the entire pest and disease monitoring path, covering all possible pest and disease occurrence areas.

[0063] S300: Using a monitoring drone to conduct periodic monitoring of pests and diseases based on the pest and disease monitoring path and the set of monitoring viewpoints, a set of periodically collected image sequences from the monitoring viewpoints is obtained;

[0064] It should be noted that the monitoring drone initiates its pest and disease monitoring mission based on a pre-planned monitoring path and a selected set of monitoring viewpoints. During periodic monitoring, the drone will pass through each monitoring viewpoint along the predetermined path and take pictures. For example, it may be set to collect pest and disease images every 3 days, or the collection frequency may be adjusted according to weather or seasonal changes to ensure continuous tracking of pest and disease development. During each monitoring session, the monitoring drone will collect images from different viewpoints. Each monitoring viewpoint can capture pest and disease information from different angles and in different areas, depending on its flight path and altitude.

[0065] Images captured by the drone each time it passes the monitoring viewpoint are saved and organized in a time-series format to obtain a set of periodically acquired image sequences from the monitoring viewpoint. Each image corresponds to a specific time point. During each monitoring cycle, the monitoring drone automatically updates the images according to the task settings.

[0066] S400: Perform implicit iterative semantic analysis of pests and diseases within the sequence of periodically acquired image sequences from the monitoring viewpoint to determine the iterative feature set of pests and diseases at the monitoring viewpoint;

[0067] Furthermore, the method involves performing implicit iterative semantic analysis of pest and disease characteristics within the periodically acquired image sequence set from the monitoring viewpoint to determine the iterative feature set of pest and disease characteristics at the monitoring viewpoint. In this embodiment, step S400 further includes:

[0068] Each periodically acquired image from a monitoring viewpoint in the set of periodically acquired image sequences is subjected to pest and disease semantic analysis to determine the set of pest and disease semantic analysis information sequences.

[0069] Extract the first pest semantic parsing information sequence from the pest semantic parsing information sequence set, perform implicit iterative analysis of pest semantics within the sequence, and obtain the iterative features of pests at the first monitoring viewpoint;

[0070] Add the pest and disease iterative features of the first monitoring viewpoint to the pest and disease iterative feature set of the monitoring viewpoint.

[0071] Furthermore, the first pest semantic parsing information sequence is extracted from the pest semantic parsing information sequence set, and implicit iterative analysis of pest semantics within the sequence is performed to obtain the iterative features of pests at the first monitoring viewpoint. In this embodiment, step S400 further includes:

[0072] Perform pest and disease feature analysis on the first pest and disease semantic parsing information sequence to obtain the first pest and disease feature sequence;

[0073] The first and second pest features in the first pest feature sequence are extracted and implicitly iteratively analyzed to obtain the second iterative pest feature.

[0074] The third-position iterative pest feature in the first pest feature sequence is obtained by implicitly iteratively analyzing the second-position iterative pest feature.

[0075] Similarly, the first pest and disease feature sequence is implicitly iteratively analyzed using the third iterative pest and disease feature to obtain the first monitoring viewpoint pest and disease iterative features.

[0076] Furthermore, implicit iterative analysis is performed on the first and second pest features extracted from the first pest feature sequence to obtain the second iterative pest feature. In this embodiment, step S400 further includes:

[0077] Euclidean distance is calculated for features of the same type in the first and second pest and disease characteristics to obtain a set of Euclidean distances.

[0078] The Euclidean distance set is normalized to construct an implicit iterative analysis matrix;

[0079] The implicit iterative analysis matrix is ​​used to iterate the second-order pest and disease feature to obtain the second-order iterative pest and disease feature.

[0080] In one embodiment, the pest and disease semantic analysis information sequence set refers to the information set processed by pest and disease semantic analysis from image sequences periodically collected from monitoring viewpoints. This set includes pest and disease information from all images periodically collected from each monitoring viewpoint, providing complete data on pest and disease evolution. By performing multiple iterative analyses on each pest and disease semantic analysis information sequence, the changes and spread trends of pest and disease characteristics at each monitoring viewpoint are revealed.

[0081] Image sequences collected by drones are first subjected to semantic analysis of pests and diseases. Using a deep learning model, the system can identify lesions, pests, affected plant parts, and the severity of pests and diseases in the images. This process converts the images into specific semantic information, such as lesion area: 10 cm², insect density: 20 insects / leaf, etc. Images from each monitoring viewpoint generate a set of semantic analysis information on pests and diseases, forming a sequence, namely the sequence of semantic analysis information on pests and diseases, representing the pest and disease situation at different time points.

[0082] For the periodically collected pest and disease semantic parsing information sequences from each monitoring viewpoint, the pest and disease features are iterated multiple times through implicit iterative analysis. The first pest and disease semantic parsing information sequence in the set is analyzed; here, "first" refers to any sequence in the set, not necessarily in order, indicating that the processing principle for each pest and disease semantic parsing information sequence in the set is consistent. Pest and disease features are extracted from the first pest and disease semantic parsing information sequence to form a pest and disease feature sequence.

[0083] Assuming the first set of data includes features such as lesion area, insect population density, and leaf damage degree, implicit iterative analysis updates these features iteratively, overlaying the pest and disease characteristics of this moment into the pest and disease development of the next moment or stage. This not only indicates the current pest and disease situation but also reflects the pest and disease trend. This process continues until a complete iterative sequence of pest and disease characteristics is predicted, showing the evolution of pests and diseases from mild to severe.

[0084] During the analysis, Euclidean distance is used to calculate the similarity between pest and disease characteristics. For example, if the first and second pest and disease characteristics are insect population density and lesion area, the difference between them can be calculated using Euclidean distance. By calculating the Euclidean distances between multiple characteristics, it is possible to identify which characteristics have significant correlations and which are relatively independent. The set of Euclidean distances is normalized, and then the normalization result is added to an initially empty matrix to construct an implicit iterative analysis matrix. This matrix helps analyze the changes in pest and disease characteristics over time.

[0085] Specifically, firstly, multiple sample datasets need to be collected, including an implicit iterative analysis matrix, second-order pest and disease features, and corresponding second-order iterative pest and disease features. The label for each data sample is the second-order iterative pest and disease feature obtained through implicit iterative analysis, which will serve as the network's output value. The input layer receives two data sets: the implicit iterative analysis matrix, which can be used for feature extraction through the convolutional layers of a convolutional neural network; and the second-order pest and disease features, which are input as pixel values ​​of the image and processed together with the implicit iterative analysis matrix.

[0086] The implicit iterative analysis matrix is ​​processed through multiple convolutional layers. Convolutional operations extract spatial features from the matrix, such as the relationships between pests and diseases. The purpose of convolutional layers is to extract useful local features from the input data, helping the model capture patterns in the implicit analysis matrix. Pooling layers are used to reduce the spatial dimensionality of the feature map, thereby reducing computational complexity and the risk of overfitting. Max pooling or average pooling is typically used to compress information. After convolutional and pooling layers, the feature map is flattened and processed through fully connected layers. The fully connected layers integrate the extracted local features and generate the final second-order iterative pest and disease features.

[0087] The iterative feature set of pests and diseases will be used for further linked analysis. For example, by analyzing the changing trend of a certain pest or disease feature, it can be determined whether the pest or disease is likely to spread in a certain area, or whether emergency control measures are needed. Through detailed analysis of iterative features, early warning of pests and diseases can be achieved, avoiding large-scale outbreaks.

[0088] S500: Using the pest and disease monitoring path as an index, perform pest and disease linkage anomaly analysis on the pest and disease iterative feature set of the monitoring viewpoint. If the anomaly analysis result indicates the presence of anomalies, obtain pest and disease monitoring early warning information.

[0089] Furthermore, using the pest and disease monitoring path as an index, a pest and disease linkage anomaly analysis is performed on the pest and disease iterative feature set of the monitoring viewpoint. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring early warning information is obtained. Step S500 in this embodiment of the application further includes:

[0090] The abnormal pest and disease feature set of the monitoring viewpoint is identified by traversing the iterative feature set of the abnormal monitoring viewpoint.

[0091] Based on the pest and disease monitoring path, the linkage cluster density is identified by performing linkage cluster density identification on the iterative feature set of pests and diseases at the abnormal monitoring viewpoints;

[0092] If the linkage cluster density is greater than or equal to the preset threshold, the abnormal monitoring viewpoint pest and disease iterative feature set and linkage cluster density will be used as the abnormal analysis results.

[0093] If the cluster density is less than the preset threshold, the set of iterative features of pests and diseases at the abnormal monitoring viewpoint will be used as the result of the abnormal analysis.

[0094] In one embodiment, cluster density refers to the degree of spatial aggregation of abnormal pest and disease characteristics. Cluster analysis identifies spatial clusters of pests and diseases and determines their density. High density typically indicates that pests and diseases may be spreading rapidly in certain areas.

[0095] The iterative feature set of pests and diseases at monitoring points is traversed to analyze whether these features are abnormal. By comparing the trends of pests and diseases with historical data or under normal conditions, monitoring points exhibiting abrupt changes, rapid changes, or patterns that deviate from the norm are identified. For example, a sudden increase in insect population density or a rapid expansion of lesion area in certain areas may be a precursor to the spread of pests and diseases.

[0096] Once the abnormal pest and disease feature sets at monitoring viewpoints are identified, the next step is to analyze the cluster density. Spatial analysis methods, such as clustering algorithms, are used to group the abnormal pest and disease features spatially. Cluster density reflects the concentration of these abnormal features in certain specific areas. For example, if pest and disease features from multiple abnormal monitoring viewpoints are highly clustered in a certain area, it may indicate that the pest and disease are spreading rapidly in that area, requiring close monitoring. The cluster density is then assessed based on a preset threshold. If the cluster density is greater than or equal to the preset threshold, it indicates that the pest and disease may be in a state of severe spread, requiring immediate emergency measures, such as local isolation or pesticide spraying. If the cluster density is below the preset threshold, although the anomaly exists, the pest and disease may not have spread widely or may still be in a localized stage, requiring continued monitoring and appropriate observation measures.

[0097] Ultimately, pest and disease monitoring and early warning information is output as an early warning result and pushed to relevant personnel through the agricultural management system or farmer terminals. This information can help farmers understand potential pest and disease risks in a timely manner and make corresponding management decisions.

[0098] Example 2, based on the same inventive concept as the method for monitoring and early warning of pests and diseases in the intercropping of white peony and camellia in the previous examples, as shown in the appendix. Figure 2 As shown, this application provides a pest and disease monitoring and early warning system for intercropping white peony and camellia oleifera. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0099] The three-dimensional spatial structure module 11 is used to obtain the plot environmental data and intercropping basic data of the white peony-camellia intercropping plot, and construct a three-dimensional spatial structure module including the canopy layer, white peony layer and surface layer.

[0100] The monitoring viewpoint recognition module 12 is used to identify the monitoring path and the monitoring viewpoint of the three-dimensional spatial structure module by using the camera parameters of the monitoring drone as constraints, so as to obtain the pest and disease monitoring path and the monitoring viewpoint set.

[0101] The periodic monitoring module 13 is used to use a monitoring drone to conduct periodic monitoring of pests and diseases according to the pest and disease monitoring path and the set of monitoring viewpoints, and to obtain a set of periodic image sequences collected from the monitoring viewpoints.

[0102] Implicit iterative analysis module 14 is used to perform implicit iterative semantic analysis of pests and diseases within the sequence of periodically acquired image sequences from the monitoring viewpoint, and to determine the iterative feature set of pests and diseases at the monitoring viewpoint.

[0103] The monitoring and early warning information acquisition module 15 is used to perform pest linkage anomaly analysis on the pest iterative feature set of the monitoring viewpoint using the pest monitoring path as an index. If the anomaly analysis result indicates the existence of anomalies, pest monitoring and early warning information is obtained.

[0104] Furthermore, the monitoring viewpoint recognition module 12 is used to perform the following steps:

[0105] The canopy height distribution of the three-dimensional spatial structure module is analyzed to obtain the canopy height distribution characteristics;

[0106] Based on the height distribution characteristics of the canopy layer and the flight height constraint in the camera parameters, the monitoring path of the three-dimensional spatial structure module is analyzed to obtain the pest and disease monitoring path;

[0107] Extract the focal length and pitch angle bandwidth from the camera parameters, and combine them with the three-dimensional spatial structure module to configure the monitoring viewpoints for the pest and disease monitoring path, thereby obtaining a set of monitoring viewpoints.

[0108] Furthermore, the monitoring viewpoint recognition module 12 is used to perform the following steps:

[0109] The three-dimensional spatial structure module is divided according to a preset grid size to obtain multiple grids;

[0110] Based on the canopy layer information of the three-dimensional spatial structure module, the average height of the canopy layer within the grid is processed to determine the average height of the canopy layer of the multiple grids;

[0111] The mean shift distribution of the average height of the canopy layer in the multiple grid cells is analyzed to determine the distribution characteristics of the canopy layer height.

[0112] Furthermore, the monitoring viewpoint recognition module 12 is used to perform the following steps:

[0113] Determine the monitoring viewpoint interval bandwidth based on focal length and pitch angle bandwidth;

[0114] The starting point of the pest and disease monitoring path is taken as the initial first monitoring viewpoint. The texture quality of the initial first monitoring viewpoint is certified. If the texture quality certification is passed, the initial first monitoring viewpoint is taken as the first monitoring viewpoint. The texture quality certification includes quality certification in four dimensions: perspective quality, luminance quality, structural exclusivity, and visual integrity.

[0115] Using the monitoring viewpoint interval bandwidth as a constraint, an initial second monitoring viewpoint is determined in combination with the first monitoring viewpoint. Texture quality authentication is performed on the initial second monitoring viewpoint. If the texture quality authentication is successful, the initial second monitoring viewpoint is used as the second monitoring viewpoint, and so on, to obtain a set of monitoring viewpoints.

[0116] Furthermore, the monitoring viewpoint recognition module 12 is used to perform the following steps:

[0117] If the texture quality certification fails, the initial first monitoring viewpoint is offset according to the flight path offset bandwidth of the monitoring drone to determine the candidate first monitoring viewpoint set;

[0118] The candidate first monitoring viewpoint set is traversed for texture quality authentication. If there is a candidate first monitoring viewpoint that has passed texture quality authentication, it is added to the selectable first monitoring viewpoint set.

[0119] The first monitoring viewpoint is selected as the one with the highest texture quality from the set of selectable first monitoring viewpoints.

[0120] Furthermore, the implicit iterative analysis module 14 is used to perform the following steps:

[0121] Each periodically acquired image from a monitoring viewpoint in the set of periodically acquired image sequences is subjected to pest and disease semantic analysis to determine the set of pest and disease semantic analysis information sequences.

[0122] Extract the first pest semantic parsing information sequence from the pest semantic parsing information sequence set, perform implicit iterative analysis of pest semantics within the sequence, and obtain the iterative features of pests at the first monitoring viewpoint;

[0123] Add the pest and disease iterative features of the first monitoring viewpoint to the pest and disease iterative feature set of the monitoring viewpoint.

[0124] Furthermore, the implicit iterative analysis module 14 is used to perform the following steps:

[0125] Perform pest and disease feature analysis on the first pest and disease semantic parsing information sequence to obtain the first pest and disease feature sequence;

[0126] The first and second pest features in the first pest feature sequence are extracted and implicitly iteratively analyzed to obtain the second iterative pest feature.

[0127] The third-position iterative pest feature in the first pest feature sequence is obtained by implicitly iteratively analyzing the second-position iterative pest feature.

[0128] Similarly, the first pest and disease feature sequence is implicitly iteratively analyzed using the third iterative pest and disease feature to obtain the first monitoring viewpoint pest and disease iterative features.

[0129] Furthermore, the implicit iterative analysis module 14 is used to perform the following steps:

[0130] Euclidean distance is calculated for features of the same type in the first and second pest and disease characteristics to obtain a set of Euclidean distances.

[0131] The Euclidean distance set is normalized to construct an implicit iterative analysis matrix;

[0132] The implicit iterative analysis matrix is ​​used to iterate the second-order pest and disease feature to obtain the second-order iterative pest and disease feature.

[0133] Furthermore, the monitoring and early warning information acquisition module 15 is used to perform the following steps:

[0134] The abnormal pest and disease feature set of the monitoring viewpoint is identified by traversing the iterative feature set of the abnormal monitoring viewpoint.

[0135] Based on the pest and disease monitoring path, the linkage cluster density is identified by performing linkage cluster density identification on the iterative feature set of pests and diseases at the abnormal monitoring viewpoints;

[0136] If the linkage cluster density is greater than or equal to the preset threshold, the abnormal monitoring viewpoint pest and disease iterative feature set and linkage cluster density will be used as the abnormal analysis results.

[0137] If the cluster density is less than the preset threshold, the set of iterative features of pests and diseases at the abnormal monitoring viewpoint will be used as the result of the abnormal analysis.

[0138] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0139] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0140] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for monitoring and early warning of pests and diseases in intercropping white peony and camellia oleifera, characterized in that, The method includes: Obtain plot environmental data and intercropping basic data for white peony-camellia intercropping plots, and construct a three-dimensional spatial structure module including the canopy layer, white peony layer and surface layer; Using the camera parameters of the monitoring drone as constraints, the monitoring path and monitoring viewpoint are identified in the three-dimensional spatial structure module to obtain the pest and disease monitoring path and monitoring viewpoint set; Using a monitoring drone, periodic monitoring of pests and diseases is conducted based on the pest and disease monitoring path and the set of monitoring viewpoints to obtain a set of periodically collected image sequences from the monitoring viewpoints; An implicit iterative semantic analysis of pests and diseases within the sequence is performed on the periodically acquired image sequence set of the monitoring viewpoint to determine the iterative feature set of pests and diseases at the monitoring viewpoint. Using the pest and disease monitoring path as an index, anomaly analysis of pest and disease linkage is performed on the iterative feature set of the monitoring viewpoint. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring and early warning information is obtained.

2. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 1, characterized in that, Using the camera parameters of the monitoring drone as constraints, the three-dimensional spatial structure module is subjected to monitoring path identification and monitoring viewpoint identification to obtain a set of pest and disease monitoring paths and monitoring viewpoints, including: The canopy height distribution of the three-dimensional spatial structure module is analyzed to obtain the canopy height distribution characteristics; Based on the height distribution characteristics of the canopy layer and the flight height constraint in the camera parameters, the monitoring path of the three-dimensional spatial structure module is analyzed to obtain the pest and disease monitoring path; Extract the focal length and pitch angle bandwidth from the camera parameters, and combine them with the three-dimensional spatial structure module to configure the monitoring viewpoints for the pest and disease monitoring path, thereby obtaining a set of monitoring viewpoints.

3. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 2, characterized in that, The canopy height distribution of the three-dimensional spatial structure module is analyzed to obtain canopy height distribution characteristics, including: The three-dimensional spatial structure module is divided according to a preset grid size to obtain multiple grids; Based on the canopy layer information of the three-dimensional spatial structure module, the average height of the canopy layer within the grid is processed to determine the average height of the canopy layer of the multiple grids; The mean shift distribution of the average height of the canopy layer in the multiple grid cells is analyzed to determine the distribution characteristics of the canopy layer height.

4. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 3, characterized in that, Extracting the focal length and pitch angle bandwidth from the camera parameters, and combining them with the three-dimensional spatial structure module, configures the monitoring viewpoints for the pest and disease monitoring path to obtain a set of monitoring viewpoints, including: Determine the monitoring viewpoint interval bandwidth based on focal length and pitch angle bandwidth; The starting point of the pest and disease monitoring path is taken as the initial first monitoring viewpoint. The texture quality of the initial first monitoring viewpoint is certified. If the texture quality certification is passed, the initial first monitoring viewpoint is taken as the first monitoring viewpoint. The texture quality certification includes quality certification in four dimensions: perspective quality, luminance quality, structural exclusivity, and visual integrity. Using the monitoring viewpoint interval bandwidth as a constraint, an initial second monitoring viewpoint is determined in combination with the first monitoring viewpoint. Texture quality authentication is performed on the initial second monitoring viewpoint. If the texture quality authentication is successful, the initial second monitoring viewpoint is used as the second monitoring viewpoint, and so on, to obtain a set of monitoring viewpoints.

5. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 3, characterized in that, The starting point of the pest and disease monitoring path is used as the initial first monitoring viewpoint. Texture quality verification is performed on the initial first monitoring viewpoint, including: If the texture quality certification fails, the initial first monitoring viewpoint is offset according to the flight path offset bandwidth of the monitoring drone to determine the candidate first monitoring viewpoint set; The candidate first monitoring viewpoint set is traversed for texture quality authentication. If there is a candidate first monitoring viewpoint that has passed texture quality authentication, it is added to the selectable first monitoring viewpoint set. The first monitoring viewpoint is selected as the one with the highest texture quality from the set of selectable first monitoring viewpoints.

6. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 1, characterized in that, An implicit iterative semantic analysis of pest and disease characteristics within the periodically acquired image sequence set from the monitoring viewpoint is performed to determine the iterative feature set of pest and disease characteristics at the monitoring viewpoint, including: Each periodically acquired image from a monitoring viewpoint in the set of periodically acquired image sequences is subjected to pest and disease semantic analysis to determine the set of pest and disease semantic analysis information sequences. Extract the first pest semantic parsing information sequence from the pest semantic parsing information sequence set, perform implicit iterative analysis of pest semantics within the sequence, and obtain the iterative features of pests at the first monitoring viewpoint; Add the pest and disease iterative features of the first monitoring viewpoint to the pest and disease iterative feature set of the monitoring viewpoint.

7. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 6, characterized in that, Extract the first pest and disease semantic parsing information sequence from the pest and disease semantic parsing information sequence set, perform implicit iterative analysis of pest and disease semantics within the sequence, and obtain the iterative features of pests and diseases from the first monitoring viewpoint, including: Perform pest and disease feature analysis on the first pest and disease semantic parsing information sequence to obtain the first pest and disease feature sequence; The first and second pest features in the first pest feature sequence are extracted and implicitly iteratively analyzed to obtain the second iterative pest feature. The third-position iterative pest feature in the first pest feature sequence is obtained by implicitly iteratively analyzing the second-position iterative pest feature. Similarly, the first pest and disease feature sequence is implicitly iteratively analyzed using the third iterative pest and disease feature to obtain the first monitoring viewpoint pest and disease iterative features.

8. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 7, characterized in that, Implicit iterative analysis is performed on the first and second pest and disease features extracted from the first pest and disease feature sequence to obtain the second iterative pest and disease feature, including: Euclidean distance is calculated for features of the same type in the first and second pest and disease characteristics to obtain a set of Euclidean distances. The Euclidean distance set is normalized to construct an implicit iterative analysis matrix; The implicit iterative analysis matrix is ​​used to iterate the second-order pest and disease feature to obtain the second-order iterative pest and disease feature.

9. The method for monitoring and early warning of pests and diseases in intercropping white peony and camellia as described in claim 1, characterized in that, Using the pest and disease monitoring path as an index, a pest and disease linkage anomaly analysis is performed on the pest and disease iterative feature set of the monitoring viewpoint. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring early warning information is obtained, including: The abnormal pest and disease feature set of the monitoring viewpoint is identified by traversing the iterative feature set of the abnormal monitoring viewpoint. Based on the pest and disease monitoring path, the linkage cluster density is identified by performing linkage cluster density identification on the iterative feature set of pests and diseases at the abnormal monitoring viewpoints; If the linkage cluster density is greater than or equal to the preset threshold, the abnormal monitoring viewpoint pest and disease iterative feature set and linkage cluster density will be used as the abnormal analysis results. If the cluster density is less than the preset threshold, the set of iterative features of pests and diseases at the abnormal monitoring viewpoint will be used as the result of the abnormal analysis.

10. A pest and disease monitoring and early warning system for intercropping white peony and camellia oleifera, characterized in that, The system is used to implement the pest and disease monitoring and early warning method for intercropping white peony and camellia as described in any one of claims 1-9, the system comprising: The three-dimensional spatial structure module is used to acquire the plot environmental data and intercropping basic data of the white peony-camellia intercropping plot, and construct a three-dimensional spatial structure module including the canopy layer, white peony layer and surface layer; The monitoring viewpoint recognition module is used to identify the monitoring path and monitoring viewpoint of the three-dimensional spatial structure module by using the camera parameters of the monitoring drone as constraints, so as to obtain the pest and disease monitoring path and monitoring viewpoint set; The periodic monitoring module is used to conduct periodic monitoring of pests and diseases using a monitoring drone based on the pest and disease monitoring path and the set of monitoring viewpoints, and to obtain a set of periodically acquired image sequences from the monitoring viewpoints. An implicit iterative analysis module is used to perform implicit iterative semantic analysis of pests and diseases within the sequence of periodically acquired image sequences from the monitoring viewpoint, and to determine the iterative feature set of pests and diseases at the monitoring viewpoint. The monitoring and early warning information acquisition module is used to perform pest and disease linkage anomaly analysis on the pest and disease iterative feature set of the monitoring viewpoint using the pest and disease monitoring path as an index. If the anomaly analysis result indicates the presence of anomalies, pest and disease monitoring and early warning information is obtained.

Images