Forest tree disease monitoring method and monitoring system

Abstract

The invention discloses a forest tree disease monitoring method and monitoring system, and relates to the technical field of forest tree monitoring, and the method comprises the steps: obtaining the regional topographic features of a forest tree monitoring region; local data calling is performed on the unmanned aerial vehicle, and a monitoring flight path is generated; in the flight process of the forest monitoring area, forest data acquisition is carried out, and multi-modal monitoring data is obtained; performing spatial alignment on the multi-modal monitoring data to obtain a plurality of area monitoring information graphs; carrying out disease area identification frame selection to obtain disease risk area distribution; carrying out disease diffusion prediction according to the disease type, and carrying out frame selection to obtain a hierarchical disease diffusion region. According to the method, the technical problems of single monitoring means, inaccurate disease area identification, difficulty in accurately predicting the disease diffusion trend and low monitoring efficiency in the existing forest disease monitoring technology are solved, and the technical effects of accurately monitoring the forest diseases, efficiently identifying the disease areas and accurately predicting the disease types and the diffusion trend are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of forest monitoring, and in particular to a forest disease monitoring method and a monitoring system. Background Art

[0002] In forestry production, forest diseases seriously affect the growth, development, yield and quality of trees, posing a major threat to the stability and sustainable development of forest ecosystems. Traditional forest disease monitoring mainly relies on manual inspections, which has many limitations. The scope of manual inspections is limited, and it is difficult to cover large areas of forest, resulting in many diseased areas not being discovered in time; and manual judgment of the type and degree of the disease is easily affected by subjective factors and has low accuracy. In addition, traditional monitoring methods are inefficient and cannot meet the needs of timely and rapid acquisition of disease information. Measures are often taken only after the disease has broken out on a large scale, at which time the difficulty and cost of prevention and control are greatly increased.

[0003] The existing forest disease monitoring technology has the following technical problems: single monitoring means, inaccurate identification of disease areas, difficulty in accurately predicting disease spread trends, and low monitoring efficiency. Summary of the invention

[0004] The present application provides a forest disease monitoring method and a monitoring system, which are used to solve the technical problems in existing forest disease monitoring technologies, such as single monitoring means, inaccurate identification of disease areas, difficulty in accurately predicting disease spread trends, and low monitoring efficiency.

[0005] In view of the above problems, the present application provides a forest disease monitoring method and monitoring system.

[0006] In a first aspect of the present application, a forest disease monitoring method is provided, the method comprising: The regional terrain characteristics of the forest monitoring area are obtained by calling the GIS system; local data of the UAV is called to obtain the multimodal parameters of the multimodal sensor, and then a monitoring flight trajectory is generated by fitting the multimodal parameters and regional terrain characteristics; the UAV is controlled by the monitoring flight trajectory to fly in the forest monitoring area, and the multimodal sensor is started to synchronously collect forest data to obtain multimodal monitoring data; the multimodal monitoring data are spatially aligned to obtain multiple regional monitoring information maps; disease area identification and selection are performed through collaborative analysis of the multiple regional monitoring information maps to obtain the distribution of disease risk areas; starting from the distribution of disease risk areas, disease spread prediction is performed according to the disease type, and hierarchical disease spread areas are obtained by selection.

[0007] The second aspect of the present application provides a forest disease monitoring system, the system comprising: A regional terrain feature acquisition module, which is used to obtain the regional terrain features of the forest monitoring area from the GIS system; a monitoring flight trajectory generation module, which is used to call local data of the UAV, obtain the multimodal parameters of the multimodal sensor, and then generate a monitoring flight trajectory according to the multimodal parameters and regional terrain features; a multimodal monitoring data acquisition module, which is used to control the UAV to fly in the forest monitoring area using the monitoring flight trajectory, and start the multimodal sensor to synchronously collect forest data. , obtain multimodal monitoring data; a regional monitoring information map acquisition module, the regional monitoring information map acquisition module is used to spatially align the multimodal monitoring data to obtain multiple regional monitoring information maps; a disease risk area distribution acquisition module, the disease risk area distribution acquisition module is used to identify and select disease areas through collaborative analysis of the multiple regional monitoring information maps, and obtain disease risk area distribution; a hierarchical disease spread area acquisition module, the hierarchical disease spread area acquisition module is used to use the disease risk area distribution as a starting point, predict disease spread according to disease type, and select hierarchical disease spread areas.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: The regional terrain features of the forest monitoring area are obtained from the GIS system call; local data is called on the UAV to obtain the multimodal parameters of the multimodal sensor, and then a monitoring flight trajectory is generated; the UAV is controlled by the monitoring flight trajectory to fly in the forest monitoring area, and the multimodal sensor is started to synchronously collect forest data to obtain multimodal monitoring data; the multimodal monitoring data is spatially aligned to obtain multiple regional monitoring information maps; disease area identification and selection are performed through collaborative analysis of the multiple regional monitoring information maps to obtain the distribution of disease risk areas; starting from the distribution of disease risk areas, disease diffusion prediction is performed according to the disease type, and hierarchical disease diffusion areas are obtained by selection. The technical effect of accurate monitoring of forest diseases, efficient identification of disease areas, and accurate prediction of disease types and diffusion trends is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a forest disease monitoring method provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a forest disease monitoring system provided in an embodiment of the present application.

[0011] Explanation of the reference numerals: regional terrain feature acquisition module 10, monitoring flight trajectory generation module 20, multimodal monitoring data acquisition module 30, regional monitoring information map acquisition module 40, disease risk area distribution acquisition module 50, hierarchical disease diffusion area acquisition module 60. DETAILED DESCRIPTION

[0012] The present application provides a forest disease monitoring method and a monitoring system, which are used to solve the technical problems in the existing forest disease monitoring technology, such as single monitoring means, inaccurate identification of disease areas, difficulty in accurately predicting the disease spread trend and low monitoring efficiency.

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0014] Embodiment 1, as Figure 1 As shown, the present application provides a forest disease monitoring method, the method comprising: Step 100: Obtain regional topographic features of the forest monitoring area from the GIS system.

[0015] Specifically, the monitoring work first establishes a close connection with the advanced GIS (Geographic Information System), and obtains detailed and accurate regional terrain characteristics of the forest monitoring area through precise data call instructions. These characteristics cover a wealth of geographic information, including altitude data in the monitoring area. Its ups and downs clearly depict the uneven terrain, which helps to determine the light, temperature and humidity conditions in different areas, and then analyze their potential impact on forest growth and disease occurrence. Topography and landform types are also an important component. For example, the distribution of different landforms such as mountains, hills, and plains is of key significance for the planning of drone flight paths and the formulation of sensor data collection strategies. For example, mountainous terrain may require the adjustment of drone flight altitude and angle to adapt to complex terrain to ensure the comprehensiveness of data collection. In addition, it also includes information such as slope and slope direction. The slope size affects the degree of soil erosion and water distribution, and the slope direction determines the duration and intensity of light exposure to the forest. These factors are closely related to the health of the forest, providing indispensable basic data support for subsequent monitoring work, so that the entire forest disease monitoring process can be carried out based on accurate terrain information, improving the scientificity and effectiveness of monitoring.

[0016] Step 200: After calling local data of the UAV and obtaining multimodal parameters of the multimodal sensor, a monitoring flight trajectory is generated based on the multimodal parameters and regional terrain features.

[0017] Specifically, the monitoring system establishes a local data connection with the drone and issues precise data call instructions to obtain the rich and diverse multimodal parameters of the multimodal sensor. These parameters include the unique performance attributes of each sensor, such as the resolution, field of view, and focal length range of the visible light camera, which determine its ability and range to capture the details of the appearance of the forest; the band range and spectral resolution of the multispectral sensor affect the detection accuracy of the spectral characteristics of vegetation; the temperature resolution and temperature measurement range of the thermal imaging sensor are related to the accurate detection of the temperature changes on the surface of the forest; the number of spectral channels and spectral coverage of the hyperspectral sensor provide more detailed spectral information. After obtaining these multimodal parameters, they are deeply integrated with the regional terrain features previously obtained from the GIS system. Using the fitting algorithm, the most suitable monitoring flight trajectory is carefully planned by fully considering the factors such as the undulation, slope, slope direction, and obstacle distribution of the terrain, combined with the monitoring characteristics of the sensor. For example, in mountainous areas with complex terrain, the flight altitude and angle will be adjusted according to the slope and slope direction to ensure that the sensor can effectively cover the target area; at the same time, obstacles such as towering peaks will be avoided to ensure flight safety and continuity of data collection. The monitoring flight trajectory generated in this way can not only give full play to the advantages of multimodal sensors, but also adapt to complex terrain conditions, laying a solid foundation for the accurate collection of subsequent forest data.

[0018] Step 300: Using the monitoring flight trajectory to control the UAV to fly in the forest monitoring area, start the multimodal sensor to synchronously collect forest data to obtain multimodal monitoring data.

[0019] Specifically, according to the carefully planned monitoring flight trajectory, the drone is guided to fly smoothly in the forest monitoring area through a precise flight control system. During the flight, the multimodal sensors are triggered synchronously to start working, so that they can work together to collect forest data in all directions. The visible light camera records the appearance of the forest with its high resolution, including intuitive visual information such as the color, shape, and density of the branches and leaves, providing a basic basis for judging the overall growth status of the forest; the multispectral sensor obtains the spectral reflectance characteristics of the forest from multiple specific bands, and through the analysis of spectral data in different bands, accurately understands the changes in the physiological structure and chemical composition of the vegetation, so as to effectively identify the early signs of potential forest diseases; the thermal imaging sensor detects the temperature distribution on the surface of the forest in real time, and can keenly detect the temperature differences caused by plant physiological abnormalities caused by diseases, such as local heating or low temperature areas; the hyperspectral sensor, with its fine spectral resolution, provides richer spectral detail information, further enhancing the ability to analyze the physiological characteristics and pathological characteristics of the forest. Through the synchronous data collection of multimodal sensors during flight, massive and multi-dimensional multimodal monitoring data is obtained. These data complement and verify each other, comprehensively reflecting the health status of forests from different angles, laying a solid data foundation for the subsequent precise analysis of forest disease conditions, ensuring that possible disease problems can be discovered accurately and timely, and providing strong support for in-depth research on the causes and spread of diseases and the formulation of effective prevention and control measures.

[0020] Step 400: spatially align the multimodal monitoring data to obtain multiple regional monitoring information maps.

[0021] Specifically, after obtaining the multimodal monitoring data, the spatial alignment program is immediately started. The program first extracts and analyzes the geospatial coordinate information in the multimodal monitoring data, which is closely related to the flight trajectory of the drone and the actual geographical location of the forest monitoring area. Then, using advanced spatial registration algorithms, the data from different modal sensors (such as visible light cameras, multispectral sensors, thermal imaging sensors, and hyperspectral sensors) are accurately matched and calibrated in three-dimensional space based on regional terrain features. For the image data collected by the visible light camera, its spatial position information is precisely adjusted to align it with other sensor data in space, ensuring that the position of the trees in the image is consistent with the physical position reflected by other sensors. The data of the multispectral sensor is fused with the overall spatial framework based on its spectral band information and spatial coordinates, so that the spectral characteristics of the vegetation can accurately correspond to the corresponding geographical area. The temperature data obtained by the thermal imaging sensor is also mapped according to the spatial coordinates, so that the temperature distribution information is accurately matched with the actual position of the trees. The rich spectral detail data of the hyperspectral sensor is also strictly calibrated in space to ensure that it is in the same spatial reference system as other modal data. After this complex and sophisticated spatial alignment process, multiple regional monitoring information maps were finally generated. These include RGB image distribution maps, which use intuitive color images to show the appearance and color changes of trees; vegetation health index distribution maps, which analyze and calculate multi-spectral data to present the health status of vegetation in an intuitive graphical way; temperature distribution maps, which clearly depict the temperature distribution on the surface of trees, which helps to find abnormal temperature areas; spectral reflectance distribution maps, which show in detail the spatial distribution of spectral reflectance characteristics of different bands, and provide important basis for in-depth analysis of forest physiological characteristics and disease characteristics. These regional monitoring information maps provide comprehensive and accurate data support for subsequent disease area identification and analysis from multiple dimensions.

[0022] Step 500: collaboratively analyzing the multiple regional monitoring information graphs to identify and select diseased areas, and obtain the distribution of disease risk areas.

[0023] Specifically, a comprehensive and in-depth collaborative analysis process is initiated for the acquired regional monitoring information maps. First, the RGB image distribution map, vegetation health index distribution map, temperature distribution map, and spectral reflectance distribution map are placed in the same analysis framework for multi-dimensional feature comparison. In the RGB image, carefully observe the color changes, texture abnormalities, and morphological distortions of the branches and leaves of the trees, which may indicate visual clues of disease; in the vegetation health index distribution map, focus on the areas with low indexes, which often indicate that the vegetation growth vitality is insufficient and there may be a threat of disease; in the temperature distribution map, carefully screen the areas with obvious temperature differences from the surrounding normal areas, and abnormal temperature changes may be the external manifestation of physiological lesions of the trees; in the spectral reflectance distribution map, accurately analyze the abnormal fluctuations of the spectral reflectance of specific bands, which may reflect the changes in the internal chemical composition of the trees, which is also a potential signal of the occurrence of diseases. Subsequently, the image recognition technology, machine learning algorithms, and professional forestry disease knowledge models are used to perform pattern recognition and intelligent judgment on the abnormal features in these information maps. Once an area that meets the characteristics of the disease is found, advanced image processing tools are used to accurately select the suspected area of ​​the disease. Then, through cross-validation and comprehensive evaluation of suspected disease areas in multiple information maps, misjudgments caused by environmental factors or data errors were eliminated, and the distribution of disease risk areas was finally accurately determined. This distribution result is presented in an intuitive graphical way, clearly identifying the specific locations and ranges with higher disease risks in the forest monitoring area, providing a key and reliable basis for subsequent disease prevention and control decisions, helping to take targeted measures in a timely manner, effectively curb the spread of diseases, and maintain the health and stability of forest resources.

[0024] Step 600: Taking the disease risk area distribution as the starting point, predict the disease spread according to the disease type, and select the hierarchical disease spread area.

[0025] Specifically, based on the distribution of identified disease risk areas, we deeply explore historical forest disease data and comprehensively analyze the occurrence patterns, propagation characteristics and interrelationships between past disease types and environmental factors. Based on the currently identified disease types, we comprehensively consider the species characteristics of trees, the growth environment (including topography, soil conditions, climate factors, etc.) and the current ecological conditions, and use advanced disease diffusion prediction models for simulation calculations. This model fully considers the spatial spread trend of diseases, such as using geographic information system (GIS) technology to analyze the hindering or promoting effect of terrain on disease spread, and the impact of different vegetation distribution patterns on disease spread paths; at the same time, combined with the time dimension, we predict the possible development speed and spread range of diseases over time. A series of possible disease spread scenarios are obtained through model calculations, and they are divided into levels according to the severity and spread range of diseases under different scenarios with scientific and reasonable classification standards. For each level, the corresponding boundaries and features are determined, and professional image processing and geospatial analysis tools are used to accurately select the hierarchical disease spread areas on the regional monitoring information map. These hierarchical disease spread areas clearly demonstrate the scope of the disease that may be affected at different stages and degrees starting from the risk area, providing highly targeted and forward-looking guidance for forestry management departments to formulate graded prevention and control strategies in advance, rationally allocate resources, and scientifically arrange prevention and control work, which will help minimize the losses caused by the disease and maintain the stability and health of the forest ecosystem.

[0026] In a possible implementation, step S200 further includes: Step S210: interactively obtaining disease monitoring requirements.

[0027] Step S220: Analyze the disease monitoring requirements to obtain multiple monitoring data types, and perform sensor requirements analysis based on the multiple monitoring data types to obtain multiple sensor types, wherein the multiple sensor types include visible light cameras, multispectral sensors, thermal imaging sensors, and hyperspectral sensors.

[0028] Step S230: interactively obtain the load range of the drone, and perform sensor configuration based on the load range and multiple sensor types as constraints to obtain multiple sensor entities.

[0029] Step S240: traversing the installation reference position table using the multiple sensor types to obtain multiple installation positions, and then installing the multiple sensor entities on the drone using the multiple installation positions as a guide to complete the installation configuration of the multimodal sensor.

[0030] Specifically, a close interaction process with relevant forestry professionals was initiated to accurately obtain disease monitoring needs. Through a specially designed interactive interface, a series of key information was obtained through interaction. First, a detailed understanding of the scope of the monitoring area, including geographical boundaries, area size, etc., will help determine the scale of the monitoring work and the spatial scope required to be covered. Next, in-depth discussions on the monitoring objectives, such as common disease monitoring for specific types of trees, a comprehensive assessment of the health of the entire forest ecosystem, or focusing on early warning of emerging and unknown diseases, etc., in order to configure monitoring resources and methods in a targeted manner. At the same time, attention is paid to users' expectations for monitoring accuracy, and the accuracy requirements such as spatial resolution and temporal resolution that need to be achieved are clearly defined, which is directly related to the selection of sensors and the design of monitoring solutions. In addition, the needs for monitoring frequency will be collected to determine whether a one-time comprehensive census or regular, continuous dynamic monitoring, as well as key monitoring periods in different seasons or growth stages, etc. Through this all-round, multi-level interactive inquiry, various types of information are comprehensively collected. After sorting and analysis, we can finally obtain clear, accurate and detailed disease monitoring needs, provide clear direction and goals for subsequent monitoring work, and ensure that the entire monitoring process can closely meet actual needs and achieve efficient and accurate forest disease monitoring.

[0031] The acquired disease monitoring requirements are deeply analyzed by using intelligent parsing algorithms. By extracting and understanding key information such as descriptions of tree disease characteristics, monitoring targets, and expected detection accuracy, the various monitoring data types required for comprehensive and accurate disease monitoring are determined. For example, if the requirements emphasize monitoring of surface symptoms of trees such as discoloration, spots, and withering, then it is clear that a type of high-resolution image data is needed to clearly present these visual features, thus pointing to visible light cameras as one of the sensor types that meet the requirements of this data type. When it comes to monitoring changes in the physiological state of trees, such as changes in photosynthesis efficiency and fluctuations in moisture content, these information related to the internal physiological mechanisms of vegetation need to be obtained by analyzing spectral reflectance characteristics. At this time, it is determined that the data type provided by multispectral sensors is indispensable. For abnormal tree temperatures caused by diseases, whether it is local heating or overall temperature changes, the type of temperature distribution data collected by thermal imaging sensors becomes a key requirement, and thermal imaging sensors are determined to be necessary monitoring equipment. If we want to deeply explore the subtle changes in the chemical composition of trees due to diseases, hyperspectral sensors can provide rich and detailed spectral information, which meets the data type requirements and is therefore included in the sensor type selection range. Through this layer-by-layer derivation and analysis based on the monitoring data type requirements, we finally accurately determined a variety of sensor types including visible light cameras, multispectral sensors, thermal imaging sensors and hyperspectral sensors, laying a solid hardware foundation for building a comprehensive and effective forest disease monitoring system.

[0032] Start the interaction process with the equipment manager, and obtain the load range information of the drone through specific interaction channels (such as data input interface, communication protocol, etc.). This load range clearly defines the weight range that the drone can safely carry, which is the key constraint condition for subsequent sensor configuration. After obtaining the load range, use it together with the multiple sensor types previously determined as constraints, enter the sensor configuration link, and call the internally stored sensor database, which contains detailed parameters of various sensors, such as weight, size, power consumption, performance indicators, etc. Filter out a list of candidate sensors that meet the requirements based on multiple sensor types, and then match and evaluate each candidate sensor based on its weight and load range. During the evaluation process, it is necessary not only to ensure that the total weight of the selected sensor does not exceed the upper limit of the drone's load to ensure the flight safety and stability of the drone, but also to comprehensively consider whether the performance and function of the sensor can meet the needs of disease monitoring. For example, without exceeding the load, give priority to sensors with higher resolution, better accuracy, and stronger stability. After rigorous calculations and trade-offs, several suitable sensor entities were finally determined from the candidate sensor list. These sensor entities achieved the optimal balance in terms of weight, performance, and functionality, and are fully prepared for subsequent installation on drones to perform forest disease monitoring tasks.

[0033] Based on the multiple sensor types that have been determined, we began to conduct a comprehensive traversal in the preset installation reference position table. The installation reference position table is carefully compiled after a large number of experimental tests and actual flight experience summaries. It records in detail the optimal installation position information for different types of sensors on the drone. For each type of sensor, carefully find the corresponding recommended installation position in the table. The determination of these positions fully considers many factors such as the working principle of the sensor, the field of view, the signal transmission characteristics, and the mutual interference relationship with other sensors. For example, the visible light camera needs to be installed in a position where it can obtain the widest field of view and is not blocked by other components to clearly capture the overall appearance of the forest; the multispectral sensor should be installed in a place where its spectral collection light path is not interfered with to ensure accurate acquisition of vegetation spectral reflectance data; the installation position of the thermal imaging sensor should be conducive to its accurate detection of the surface temperature of the forest, while avoiding the influence of the heating components of the drone itself; the installation of the hyperspectral sensor must ensure that its fine spectral collection function can be fully utilized and does not interfere with other sensors when working together. After the traversal is completed, multiple installation positions for different sensor entities are successfully obtained. Subsequently, strictly following these location guidelines, with professional installation tools and skills, multiple sensor entities were accurately installed at the corresponding positions of the drone. During the installation process, it was ensured that each sensor was firmly fixed, the connection line was correct and stable, and the direction and angle of the sensor met the preset requirements, thus completing the installation and configuration of the multi-modal sensor, enabling the drone to have the ability to comprehensively and efficiently collect forest disease monitoring data, laying a solid hardware foundation for subsequent flight monitoring work.

[0034] In a possible implementation, step S400 further includes: Step S410: Generate a regional plan map of the forest monitoring area based on the projection of the regional terrain features.

[0035] Step S420: calling the multispectral image data collected by the multispectral sensor from the multimodal monitoring data.

[0036] Step S430: After calculating the NDVI value of the multispectral image data, the multispectral image data is superimposed on the regional plan according to the monitoring flight trajectory to obtain a first regional monitoring information map, wherein the first regional monitoring information map is a vegetation health index distribution map.

[0037] Step S440: Similarly, after preprocessing the multimodal monitoring data according to the multiple data preprocessing requirements of the multiple sensor types, spatial alignment is performed based on the regional plan to obtain the multiple regional monitoring information maps, wherein the multiple regional monitoring information maps include an RGB image distribution map, a vegetation health index distribution map, a temperature distribution map, and a spectral reflectance distribution map.

[0038] Specifically, the regional terrain characteristics of the forest monitoring area are used as the basic data, and a professional projection algorithm is used to convert the three-dimensional terrain data into a two-dimensional plane image, thereby generating a regional plane map of the forest monitoring area. This plane map accurately presents the geographical outline, topographic distribution, and key landmark locations of the monitoring area, providing a unified geographic spatial reference framework for subsequent data integration and analysis.

[0039] The multispectral image data collected by the multispectral sensor are accurately called from the previously collected multimodal monitoring data. These multispectral image data contain the reflectance information of forests in multiple specific spectral bands and are an important data source for analyzing the physiological characteristics and health status of forests.

[0040] The NDVI value is calculated for the multispectral image data extracted from the multimodal monitoring data. The NDVI value is a vegetation health index. The calculation process strictly follows a specific formula to accurately analyze the reflectance information of the red band and the near-infrared band in the multispectral image data. The NDVI value corresponding to each pixel is obtained by calculating the ratio of the difference and the sum of the two. These NDVI values ​​can keenly reflect the growth status of vegetation. For example, areas with higher NDVI values ​​usually indicate vigorous vegetation growth, rich chlorophyll content, and good health; while lower values ​​may indicate that vegetation growth is inhibited, there is a potential risk of disease or other adverse factors. After the NDVI value calculation is completed, the monitoring flight trajectory that was carefully planned and obtained before is used as a spatial position reference. The monitoring flight trajectory contains key information such as the flight path, altitude, direction, and time series of the drone when collecting multispectral image data, ensuring the accurate correspondence between the multispectral image data and the actual geographic spatial location. With this information, the multispectral image data with NDVI values ​​are accurately and accurately superimposed on the generated regional plane map. During the overlay process, the geographic coordinates of each pixel point are accurately matched, so that the vegetation health information represented by the multispectral image data can be accurately located at the corresponding position of the regional plan. Finally, the first regional monitoring information map, that is, the vegetation health index distribution map, was successfully obtained. This map clearly presents the distribution of vegetation health status in the entire forest monitoring area in an intuitive and visual way. The changes in color depth or numerical value directly reflect the differences in vegetation health in different areas, providing extremely valuable basic data and intuitive spatial distribution information for the subsequent rapid and accurate identification of areas where diseases may exist, greatly contributing to the early detection and precise positioning of forest diseases.

[0041] According to similar logic and methods, other data in the multimodal monitoring data are preprocessed according to the data preprocessing requirements of various sensor types. For example, for the data collected by the visible light camera, image enhancement, denoising and other processing are performed to optimize the RGB image quality; temperature calibration, thermal radiation correction and other operations are performed on the thermal imaging sensor data; spectral smoothing, band selection and other preprocessing are performed on the hyperspectral sensor data. After completing various preprocessing, based on the regional plan map, the spatial registration and alignment algorithms are used to accurately align the processed visible light image data, thermal imaging temperature data and hyperspectral reflectance data with the regional plan map, and finally obtain multiple regional monitoring information maps including RGB image distribution map, temperature distribution map and spectral reflectance distribution map. These information maps comprehensively display the appearance morphology, surface temperature distribution and spectral reflectance characteristics of the forest from different angles, providing rich and comprehensive data support for the comprehensive and in-depth analysis of forest disease conditions.

[0042] In a possible implementation, step S500 further includes: Step S510: using 1 / K of the forest monitoring area as an image segmentation scale to rasterize the multiple regional monitoring information maps, and obtaining multiple groups of segmented monitoring information maps.

[0043] Step S520: Perform sample data networking scheduling based on the multiple sensor types as constraints to obtain multiple sample monitoring information atlases and multiple sample disease type identification sets.

[0044] Step S530: constructing multiple disease type recognition models using the multiple sample monitoring information atlases and the multiple sample disease type identification sets.

[0045] Step S540: connect the multiple disease type identification models in parallel to complete the construction of the forest disease identification model.

[0046] Step S550: After receiving the multiple groups of segmented monitoring information graphs, the forest disease identification model automatically identifies the types of forest diseases and outputs multiple groups of segmented disease identifiers.

[0047] Step S560: According to the mapping relationship between the multiple groups of segmented disease identifications and the multiple groups of segmented monitoring information maps, disease area restoration is performed on the multiple regional monitoring information maps to obtain multiple disease distribution maps.

[0048] Step S570: Overlay analysis is performed on the multiple disease distribution maps to output the disease risk area distribution.

[0049] Specifically, based on the forest monitoring area, 1 / K (K is a constant determined according to actual needs and monitoring accuracy) is selected as the image segmentation scale, and multiple regional monitoring information maps (including RGB image distribution maps, vegetation health index distribution maps, temperature distribution maps, and spectral reflectance distribution maps, etc.) are rasterized. This process converts continuous image data into discrete grid cells, each of which has a specific attribute value, which facilitates subsequent analysis and calculation. In this way, multiple sets of segmented monitoring information maps are obtained. These segmented images are easier to extract local features and recognize patterns while maintaining the original information.

[0050] Based on the characteristics and requirements of the information obtained by various sensor types, sample data is networked and scheduled in a wide range of network data sources or local databases with these information as constraints. For the samples corresponding to the visible light camera data, a sample set of RGB images of trees with different disease characteristics and health status, as well as the corresponding disease type identification, is collected; for multispectral sensor data, a sample monitoring information set containing changes in various vegetation health indexes and corresponding disease identifications is obtained; and by analogy, sample data and disease type identifications related to temperature distribution and spectral reflectance are collected for thermal imaging sensors and hyperspectral sensors, respectively. Finally, multiple sample monitoring information atlases and multiple sample disease type identification sets are obtained. These sample data cover a wide range of forest disease conditions and provide sufficient training data for subsequent model construction.

[0051] The decision tree algorithm is used to build multiple disease type recognition models. First, operations are carried out for each sample monitoring information atlas and the corresponding sample disease type identification set. Taking the data set corresponding to a certain sensor type (such as a visible light camera) as an example, its sample monitoring information atlas contains many forest appearance image samples, each sample has multiple attributes such as color distribution and texture characteristics, and the sample disease type identification set clearly marks the disease type to which each sample belongs. Starting from the root node, the information gain of each attribute is calculated, and the attribute with the largest information gain is selected as the split node. For example, for image samples, color distribution may have a higher information gain and be preferred. The sample set is divided into multiple subsets according to different values ​​of the attribute, and the above process is repeated on each subset, and branch nodes are continuously constructed until the stopping condition is met (such as all samples in the subset belong to the same disease type or reach a predetermined depth, etc.). In this way, each branch path represents a decision rule for a disease type, and finally a decision tree model is formed. For data from other sensor types (multispectral sensors, thermal imaging sensors, hyperspectral sensors), their respective decision tree models are constructed in the same way to obtain multiple disease type recognition models, which can accurately identify forest disease types based on the characteristics of different types of data.

[0052] The multiple disease type recognition models that have been built are integrated in parallel. Through reasonable architecture design and data interaction mechanism, these models can work together to analyze the input monitoring information. When receiving multiple sets of segmented monitoring information graphs, each model can process the image from different angles at the same time, give full play to their respective advantages, improve the accuracy and comprehensiveness of disease recognition, and thus complete the construction of the forest disease recognition model.

[0053] The constructed forest disease recognition model begins to work. When receiving multiple sets of segmented monitoring information maps, the model automatically starts the recognition program. Each parallel sub-model analyzes and judges each grid unit in the input segmented monitoring information map based on the feature patterns learned through training, identifies the possible disease features, and outputs multiple sets of segmented disease identifications corresponding to them. These identifications accurately mark the disease type or health status category of each grid unit, providing a key basis for the subsequent restoration of the diseased area.

[0054] According to the mapping relationship between multiple groups of segmented disease identification and multiple groups of segmented monitoring information maps, the disease area restoration operation is performed on the original multiple regional monitoring information maps. By reversely mapping the segmented disease identification back to the original image, the image area corresponding to each identification is determined, so as to accurately restore the location and range of the disease area on multiple regional monitoring information maps such as RGB image distribution map, vegetation health index distribution map, temperature distribution map and spectral reflectance distribution map. After this process, multiple disease distribution maps are obtained, which show the distribution of diseases in the monitoring area from different angles, providing more detailed information for further analysis of the disease.

[0055] Multiple disease distribution maps are superimposed and analyzed, and the diseased areas in the RGB image are comprehensively compared and statistically analyzed with the corresponding areas in the vegetation health index distribution map, temperature distribution map, and spectral reflectance distribution map. Areas with higher disease risks are determined by analyzing the overlap, feature consistency, and spatial distribution patterns of diseased areas in different distribution maps. Finally, the disease risk area distribution is output, which clearly shows the spatial distribution of disease risks in the forest monitoring area in an intuitive graphical way, providing an important decision-making basis for forestry management departments to formulate accurate disease prevention and control strategies, and helping to take effective prevention and control measures in a timely manner to protect the health of forest resources.

[0056] In a possible implementation, step S560 further includes: Step S561: according to the mapping relationship between the first group of segmented disease identifications and the first group of segmented monitoring information maps, disease intensity calculation is performed in the first area monitoring information map to obtain a first disease pixel gradient map; Step S562: presetting a disease intensity threshold, and using the disease intensity threshold to traverse the first disease pixel gradient map to obtain a plurality of initial seed grids; Step S563: In the first disease pixel gradient map, similar disease grids are merged starting from the multiple initial seed grids until they are extended to the boundary area to obtain multiple disease sub-regions, wherein the multiple disease sub-regions have multiple sub-region disease type identifiers; Step S564: merging the multiple disease sub-areas of the same type according to the disease types of the multiple sub-areas to obtain a first disease distribution map; Step S565: Similarly, according to the mapping relationship between the multiple groups of segmented disease identifications and the multiple groups of segmented monitoring information maps, disease area restoration is performed on the multiple regional monitoring information maps to obtain the multiple disease distribution maps.

[0057] Specifically, based on the mapping association between the first group of segmented disease identification and the first group of segmented monitoring information maps, the disease information in the first regional monitoring information map is deeply mined to carry out the calculation of disease intensity. By carefully analyzing the position, range and characteristic performance of the disease area indicated by the segmented disease identification in the first regional monitoring information map, a disease intensity value is assigned to each pixel. These intensity values ​​can accurately quantify the degree to which the area represented by each pixel is affected by the disease. For example, the more obvious and severe the disease characteristics are, the higher the disease intensity value corresponding to the pixel. On this basis, the first disease pixel gradient map is generated, which intuitively displays the spatial variation trend of disease intensity in the form of pixel gradients. Areas with larger gradients indicate drastic changes in disease intensity, which may be the core area or edge transition area of ​​the disease.

[0058] A disease intensity threshold is pre-set. This threshold is determined based on the analysis of a large amount of historical disease data and expert experience, and is used to distinguish diseased areas from healthy areas or areas with less severe diseases. Then, the disease intensity threshold is used to traverse and screen the first disease pixel gradient map. When the disease intensity value of a certain pixel is greater than or equal to the preset threshold, the pixel is marked as an initial seed grid. In this way, multiple initial seed grids are screened out. These initial seed grids serve as the starting points for subsequent expansion and merging of diseased areas. They are distributed in areas with high and relatively concentrated disease intensities and are highly representative.

[0059] In the first disease pixel gradient map, the similar disease grid merging process is started with the multiple initial seed grids obtained as the starting point. Starting from each initial seed grid, according to the pre-set similarity judgment rule, the rule comprehensively considers the pixel's disease intensity, color characteristics (for image data), spectral reflectance characteristics (for spectral data) and spatial adjacent relationship and other factors to judge whether its surrounding grids have similar disease characteristics with the initial seed grid. If the surrounding grid is judged to be similar, it will be merged into the current disease area, and the judgment and merging operation will continue to be performed on the outer grids. This process continues, just like the water flow with the initial seed grid as the source continues to spread and expand until it encounters an area with significantly different disease characteristics (i.e., the boundary area). In this way, grids with similar disease characteristics are gradually merged into a whole to form multiple disease sub-areas, and each disease sub-area is assigned a corresponding sub-area disease type identifier, which is determined according to the disease type to which the initial seed grid belongs and the comprehensive characteristics of the grids involved in the merging process.

[0060] According to the disease types of multiple sub-areas, the same type of merging operation is carried out on multiple disease sub-areas. The disease sub-areas with the same disease type identification are merged and integrated, and the gaps and discontinuous areas between the sub-areas are eliminated to form a complete and continuous disease distribution area, thereby obtaining the first disease distribution map. This map clearly and accurately shows the distribution range, shape, severity and other information of the disease in the first area monitoring information map, which provides an important basis for further analyzing the law of disease transmission and formulating prevention and control measures.

[0061] According to the same logic and method, the disease area restoration operation is performed on multiple regional monitoring information maps (such as RGB image distribution map, temperature distribution map, spectral reflectance distribution map, etc.) in turn for the mapping relationship between the segmented disease identification and segmented monitoring information map of other groups. Repeat the steps of disease intensity calculation, initial seed grid screening, similar disease grid merging and similar merging to generate a corresponding disease distribution map for each regional monitoring information map. These multiple disease distribution maps comprehensively present the distribution of diseases in the forest monitoring area from different dimensions and angles, providing rich and detailed information support for the comprehensive assessment of forest health status and the precise implementation of disease prevention and control strategies.

[0062] In a possible implementation, step S600 further includes: Step S610: Aggregate the multiple sample disease type identification sets to obtain multiple sample disease types and multiple groups of disease intensity sequences.

[0063] Step S620: calling the disease diffusion scales with the plurality of disease intensity sequences as constraints to obtain a plurality of diffusion scale sequences.

[0064] Step S630: According to the consistency of the scale parameters of the multiple groups of diffusion scale sequences, the multiple groups of disease intensity sequences are graded to obtain multiple levels of disease intensity intervals.

[0065] Step S640: associatively storing the multiple sample disease types, multiple levels of disease intensity intervals, and multiple groups of diffusion scale sequences to complete the construction of a diffusion information library.

[0066] Step S650: traverse the diffusion information database using the disease risk area distribution to obtain multiple diffusion scale information.

[0067] Step S660: using the plurality of diffusion scale information to distribute in the disease risk area, selecting a disease diffusion area, and obtaining the hierarchical disease diffusion area.

[0068] Specifically, multiple sample disease type identification sets are aggregated, and the different disease types contained in the numerous sample disease type identification sets are carefully sorted out and summarized, and the identifications of the same type are classified and integrated, so as to clearly obtain multiple sample disease types. For example, different categories such as fungal infection diseases, bacterial infection diseases, and diseases caused by insect pests are identified. At the same time, for each sample disease type, the disease intensity-related data presented in each sample is deeply mined. By extracting and sorting these data, multiple groups of disease intensity sequences are constructed according to different disease types. These sequences record in detail the intensity changes of each disease type in different sample scenarios, such as the severity of disease symptoms, the size of the lesion range, and the degree of impact on forest growth. The quantitative data changes with the samples, providing a rich and critical data basis for the subsequent in-depth analysis of disease characteristics and prediction of disease spread trends.

[0069] Make full use of the multiple sets of disease intensity sequences that have been obtained as key constraints to accurately call the disease diffusion scale. Multiple sets of disease intensity sequences contain rich information, such as the trend of disease severity changes, intensity characteristics at different stages, etc. Based on this information, query and match operations are performed in the pre-established disease diffusion scale model or database, which stores a large amount of diffusion scale data corresponding to different disease intensities. These data are obtained through in-depth analysis of historical disease data, field observations, and professional research summaries. For example, when a set of disease intensity sequences shows that the disease is in the early stage and the intensity is low, a smaller diffusion scale parameter corresponding to it will be called, because the disease diffusion range is relatively small at this time; and when the disease intensity sequence shows that the disease has developed to a more serious stage, a larger diffusion scale parameter will be matched to reflect the disease's possible wider and faster diffusion trend under high severity. In this way, the corresponding appropriate diffusion scale parameters are found for each set of disease intensity sequences, thereby obtaining multiple sets of diffusion scale sequences, which will provide important basic data support for the subsequent accurate prediction of the disease diffusion range and area.

[0070] In-depth analysis of the consistency characteristics of scale parameters of multiple groups of diffusion scale sequences. Through detailed comparison and statistical analysis of the scale parameters in these sequences, explore the laws and commonalities. For example, observe whether the scale parameters of different groups of diffusion scale sequences show similar numerical ranges or change trends at the same disease development stage or similar disease intensity. According to the consistency of this scale parameter, scientific and reasonable intensity classification is carried out for multiple groups of disease intensity sequences, and the disease intensity ranges corresponding to similar diffusion scale characteristics are divided into the same level, so as to determine the multi-level disease intensity interval. For example, if the corresponding diffusion scale parameters are in a small and similar range under certain disease intensities, it indicates that the disease diffusion ability is relatively weak in this intensity interval, and these disease intensities are classified into a lower intensity level interval; on the contrary, if the diffusion scale parameters are large and tend to be consistent at a specific disease intensity, it means that the disease diffusion trend is strong, and it is classified into a higher intensity level interval. In this way, through the consistency analysis of scale parameters of multiple groups of diffusion scale sequences, multi-level disease intensity intervals are accurately obtained, which provides a reliable grading standard for the subsequent construction of an accurate diffusion information database and disease spread prediction.

[0071] The previously obtained multiple sample disease types, multi-level disease intensity intervals, and multiple groups of diffusion scale sequences are organically integrated and associated to build a complete diffusion information library. For each sample disease type, the corresponding relationship between it and the corresponding multi-level disease intensity interval is accurately established, and the characteristic manifestations and change rules of different disease types in each intensity interval are recorded in detail. For example, for a common forest fungal disease, the appearance symptoms, physiological index changes, and possible propagation speed of the forest trees in different intensity intervals such as mild, moderate, and severe are clearly defined. Then, each level of disease intensity interval is closely associated with the corresponding multiple groups of diffusion scale sequences to ensure that the matching disease diffusion scale information can be accurately obtained at each intensity level. This information covers the key parameters such as the possible diffusion range, direction, and diffusion speed of the disease under different environmental conditions and forest growth conditions. For example, within a certain intensity interval, based on historical data and model analysis, the different diffusion scales of the disease in flat terrain and mountainous terrain, as well as the differences in diffusion speed in different forest density areas, are determined. Through this meticulous associative storage method, multiple sample disease types, multiple levels of disease intensity intervals, and multiple groups of diffusion scale sequences are integrated into an organic whole to complete the construction of the diffusion information library. This provides comprehensive, accurate, and targeted data support for subsequent disease spread predictions, enabling rapid query and call of relevant information when faced with actual forest disease situations, laying a solid foundation for accurately predicting disease spread trends and formulating effective prevention and control strategies.

[0072] The distribution of the determined disease risk area is used as an index to traverse the diffusion information database. According to the disease type and disease intensity in the disease risk area, matching diffusion scale information is searched in the diffusion information database to obtain multiple diffusion scale information, which accurately reflects the possible diffusion characteristics of the disease in the current disease risk area.

[0073] Using the obtained multiple diffusion scale information, the disease diffusion area is framed in the distribution of disease risk areas. According to the diffusion range and direction determined by the diffusion scale information, the area that may be affected by the disease spread is accurately framed around the disease risk area, so as to obtain the hierarchical disease diffusion area. These hierarchical disease diffusion areas clearly show the different levels and ranges of the disease spreading outward from the risk area, which provides important decision-making support for forestry management departments to formulate targeted disease prevention and control strategies, and helps to take effective prevention and control measures in a timely manner to prevent the further spread of the disease and protect the health of the forest ecosystem.

[0074] In a possible implementation, step S400 further includes: Step S450: Call local data of the UAV to obtain multiple effective monitoring ranges of the multiple sensor entities in the multimodal sensor, wherein the multiple effective monitoring ranges constitute the multimodal parameters; use geometric Boolean operations to solve the spatial intersection of the multiple effective monitoring ranges to obtain a standard monitoring range; obtain the vertical span of the trajectory by vertically projecting the standard monitoring range; obtain the horizontal span of the trajectory by horizontally projecting the standard monitoring range; fit a horizontal spiral trajectory in the forest monitoring area with the horizontal span of the trajectory as a constraint; extract a terrain change sequence from the terrain features of the area according to the horizontal spiral trajectory; dynamically update the trajectory height of the horizontal spiral trajectory according to the terrain change sequence and the horizontal span of the trajectory to obtain the monitoring flight trajectory.

[0075] Specifically, firstly, the local data of the UAV is called. This process aims to obtain multiple effective monitoring range information of each sensor entity in the multimodal sensor. Each sensor has a specific effective monitoring range due to its own characteristics and installation location. These range information together constitute the multimodal parameters. For example, the visible light camera may be able to clearly image within a certain angle range and distance range in the horizontal direction. The effective monitoring range of the multispectral sensor depends on the effective area of ​​its spectral acquisition, etc. The effective monitoring range data of these different sensors are integrated to provide a basic basis for subsequent trajectory planning. Then, the geometric Boolean operation is used to solve the spatial intersection of multiple effective monitoring ranges. The geometric Boolean operation can accurately handle the regional relationship of different shapes and spatial positions. Through this operation, the common coverage area of ​​the effective monitoring range of all sensors is found, that is, the standard monitoring range is obtained. This standard monitoring range ensures that during the flight, the multimodal sensors carried by the UAV can simultaneously collect effective data in the area, avoiding the situation where the monitoring blind area or some sensors cannot be covered. After that, by vertically projecting the standard monitoring range, its span in the vertical direction can be determined, thereby obtaining the vertical span of the trajectory. This data is of great significance for planning the flight altitude of the UAV, ensuring that the UAV can cover all target areas within the standard monitoring range in the vertical direction, while avoiding the flight altitude being too high or too low to affect the data collection quality. Similarly, by horizontally projecting the standard monitoring range, the horizontal span of the trajectory is obtained. This span clarifies the range that the UAV needs to cover in the horizontal direction, providing a key lateral constraint for the subsequent fitting of the horizontal spiral trajectory. With the obtained horizontal span of the trajectory as a constraint, a horizontal spiral trajectory is fitted in the forest monitoring area. The design of the horizontal spiral trajectory enables the UAV to fly in a relatively uniform and efficient manner while covering the monitoring area, ensuring that the data collection of the entire area has good spatial continuity and integrity, and avoiding the problem of data omission or repeated collection. Then, the terrain change sequence is extracted from the regional terrain features according to the fitted horizontal spiral trajectory. Using geographic information system (GIS) technology or other terrain analysis tools, the terrain undulations, slope changes and other information along the trajectory are analyzed to obtain the terrain change sequence. This sequence reflects the difference in terrain height at different locations in the monitoring area, which is crucial for the subsequent adjustment of the UAV flight altitude to adapt to terrain changes. Finally, the trajectory height of the horizontal spiral trajectory is dynamically updated according to the sequence of terrain changes and the horizontal span of the trajectory. Taking into account the terrain ups and downs and the horizontal coverage requirements, the flight altitude of the drone is adjusted in real time to ensure that the drone can maintain a suitable altitude to obtain high-quality monitoring data during the entire flight process, while avoiding collision risks caused by terrain changes. Through such dynamic updates, the optimized monitoring flight trajectory was finally obtained, laying a solid foundation for the subsequent drones equipped with multi-modal sensors to collect efficient and accurate forest data.

[0076] Embodiment 2, based on the same inventive concept as a forest disease monitoring method in the above embodiment, Figure 2 As shown, the present application provides a forest disease monitoring system, and the system and method embodiments in the present application are based on the same inventive concept. The system includes: The regional terrain feature acquisition module 10 is used to obtain the regional terrain features of the forest monitoring area from the GIS system.

[0077] The monitoring flight trajectory generation module 20 is used to call local data of the UAV, obtain multimodal parameters of the multimodal sensor, and then generate a monitoring flight trajectory according to the multimodal parameters and regional terrain features.

[0078] The multimodal monitoring data acquisition module 30 is used to control the UAV to fly in the forest monitoring area using the monitoring flight trajectory, start the multimodal sensor to synchronously collect forest data, and obtain multimodal monitoring data.

[0079] The regional monitoring information map acquisition module 40 is used to spatially align the multimodal monitoring data to obtain a plurality of regional monitoring information maps.

[0080] A disease risk area distribution acquisition module 50, the disease risk area distribution acquisition module 50 is used to identify and select disease areas by collaboratively analyzing the multiple regional monitoring information graphs to obtain disease risk area distribution; The hierarchical disease diffusion area acquisition module 60 is used to take the disease risk area distribution as the starting point, perform disease diffusion prediction according to the disease type, and select the hierarchical disease diffusion area.

[0081] Furthermore, the monitoring flight trajectory generation module 20 also includes: A disease monitoring requirement acquisition unit is used to interactively obtain disease monitoring requirements.

[0082] A multiple sensor type acquisition unit is used to parse the disease monitoring requirements to obtain multiple monitoring data types, and perform sensor demand analysis based on the multiple monitoring data types to obtain multiple sensor types, wherein the multiple sensor types include visible light cameras, multi-spectral sensors, thermal imaging sensors and hyperspectral sensors.

[0083] A sensor entity acquisition unit is used to interactively obtain the load range of the drone, and perform sensor configuration based on the load range and multiple sensor types as constraints to obtain multiple sensor entities.

[0084] An installation position acquisition unit is used to traverse the installation reference position table using the multiple sensor types, and after obtaining multiple installation positions, the multiple sensor entities are installed on the drone based on the multiple installation positions to complete the installation configuration of the multimodal sensor.

[0085] Furthermore, the regional monitoring information map acquisition module 40 also includes: A regional plane map generating unit is used to generate a regional plane map of the forest monitoring area according to the projection of the regional terrain features.

[0086] A multispectral image data calling unit is used to call the multispectral image data collected by the multispectral sensor from the multimodal monitoring data.

[0087] A regional monitoring information map acquisition unit, wherein the regional monitoring information map acquisition unit is used to calculate the NDVI value of the multispectral image data, and then superimpose the multispectral image data on the regional plan according to the monitoring flight trajectory to obtain a first regional monitoring information map, wherein the first regional monitoring information map is a vegetation health index distribution map.

[0088] A plurality of regional monitoring information map acquisition units, wherein the plurality of regional monitoring information map acquisition units are used for, by analogy, after preprocessing the multimodal monitoring data according to the plurality of data preprocessing requirements of the plurality of sensor types, performing spatial alignment based on the regional plan view to obtain the plurality of regional monitoring information maps, wherein the plurality of regional monitoring information maps include an RGB image distribution map, a vegetation health index distribution map, a temperature distribution map, and a spectral reflectance distribution map.

[0089] Furthermore, the disease risk area distribution acquisition module 50 also includes: The segmented monitoring information map acquisition unit is used to rasterize the multiple regional monitoring information maps by taking 1 / K of the forest monitoring area as the image segmentation scale to obtain multiple groups of segmented monitoring information maps.

[0090] A sample data networking scheduling unit, wherein the sample data networking scheduling is used to perform sample data networking scheduling based on the multiple sensor types as constraints to obtain multiple sample monitoring information atlases and multiple sample disease type identification sets.

[0091] A disease type identification model construction unit is used to construct multiple disease type identification models using the multiple sample monitoring information atlases and multiple sample disease type identification sets.

[0092] A forest disease identification model construction unit is used to connect the multiple disease type identification models in parallel to complete the construction of the forest disease identification model.

[0093] A multi-group segmented disease identification output unit is used for the forest disease identification model to automatically identify the forest disease type and output a multi-group segmented disease identification after receiving the multi-group segmented monitoring information graph.

[0094] A disease distribution map acquisition unit is used to restore the disease area in the multiple regional monitoring information maps according to the mapping relationship between the multiple groups of segmented disease identifications and the multiple groups of segmented monitoring information maps, so as to obtain multiple disease distribution maps.

[0095] A disease risk area distribution output unit is used to perform superposition analysis on the multiple disease distribution maps and output the disease risk area distribution.

[0096] Furthermore, the disease distribution map acquisition unit further includes: The first disease pixel gradient map acquisition unit is used to calculate the disease intensity in the first area monitoring information map according to the mapping relationship between the first group of segmented disease identifiers and the first group of segmented monitoring information maps to obtain the first disease pixel gradient map.

[0097] An initial seed grid acquisition unit is used to preset a disease intensity threshold, and use the disease intensity threshold to traverse the first disease pixel gradient map to obtain multiple initial seed grids.

[0098] A plurality of disease sub-region acquisition units, wherein the plurality of disease sub-regions are used to merge similar disease grids in the first disease pixel gradient map with the plurality of initial seed grids as the starting point until the grids are expanded to the boundary area, thereby obtaining a plurality of disease sub-regions, wherein the plurality of disease sub-regions have a plurality of sub-region disease type identifiers.

[0099] The first disease distribution map acquisition unit is used to merge the multiple disease sub-areas of the same type according to the disease types of the multiple sub-areas to obtain a first disease distribution map.

[0100] A disease area restoration unit is used to perform disease area restoration on the multiple regional monitoring information maps based on the mapping relationship between the multiple groups of segmented disease identifications and the multiple groups of segmented monitoring information maps, and obtain the multiple disease distribution maps.

[0101] Furthermore, the hierarchical disease diffusion area acquisition module 60 further includes: A disease intensity sequence acquisition unit is used to aggregate the multiple sample disease type identification sets to obtain multiple sample disease types and multiple groups of disease intensity sequences.

[0102] A diffusion scale sequence acquisition unit is used to call the disease diffusion scale based on the multiple groups of disease intensity sequences as constraints to obtain multiple groups of diffusion scale sequences.

[0103] A disease intensity interval acquisition unit is used to perform intensity grading on the multiple groups of disease intensity sequences according to the consistency of scale parameters of the multiple groups of diffusion scale sequences to obtain multi-level disease intensity intervals.

[0104] A diffusion information library construction unit is used to associate and store the multiple sample disease types, multiple disease intensity intervals and multiple groups of diffusion scale sequences to complete the construction of the diffusion information library.

[0105] A diffusion scale information acquisition unit is used to traverse the diffusion information database using the disease risk area distribution to obtain multiple diffusion scale information.

[0106] A hierarchical disease diffusion area acquisition unit is used to select the disease diffusion area by using the multiple diffusion scale information distributed in the disease risk area to obtain the hierarchical disease diffusion area.

[0107] Furthermore, the regional monitoring information map acquisition module 40 also includes: An effective monitoring range acquisition unit, wherein the effective monitoring range acquisition unit is used to perform local data call on the UAV to obtain multiple effective monitoring ranges of the multiple sensor entities in the multimodal sensor, wherein the multiple effective monitoring ranges constitute the multimodal parameters.

[0108] A standard monitoring range acquisition unit is used to use geometric Boolean operations to solve the spatial intersection of the multiple effective monitoring ranges to obtain a standard monitoring range.

[0109] A track vertical span acquisition unit is used to obtain the track vertical span by vertically projecting the standard monitoring range.

[0110] A track horizontal span acquisition unit is used to obtain the track horizontal span by horizontally projecting the standard monitoring range.

[0111] A horizontal spiral trajectory fitting unit is used to fit a horizontal spiral trajectory in the forest monitoring area with the horizontal span of the trajectory as a constraint.

[0112] A terrain change sequence acquisition unit is used to extract and obtain a terrain change sequence from the regional terrain features according to the horizontal spiral trajectory.

[0113] A monitoring flight trajectory acquisition unit is used to dynamically update the trajectory height of the horizontal spiral trajectory according to the terrain change sequence and the trajectory horizontal span to obtain the monitoring flight trajectory.

[0114] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0115] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0116] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A forest disease monitoring method, characterized in that: The method comprises: Obtain regional topographic features of the forest monitoring area from the GIS system; After calling local data of the UAV and obtaining multimodal parameters of the multimodal sensor, a monitoring flight trajectory is generated by fitting the multimodal parameters and regional terrain features; The monitoring flight trajectory is used to control the UAV to fly in the forest monitoring area, and the multimodal sensor is started to synchronously collect forest data to obtain multimodal monitoring data; Performing spatial alignment on the multimodal monitoring data to obtain multiple regional monitoring information maps; By collaboratively analyzing the multiple regional monitoring information maps, disease area identification and selection are performed to obtain the distribution of disease risk areas; Taking the distribution of the disease risk areas as the starting point, the disease spread is predicted according to the disease type, and the hierarchical disease spread area is obtained by box selection.

2. A forest disease monitoring method as claimed in claim 1, characterized in that: The method comprises: Interactively obtain disease monitoring needs; Parsing the disease monitoring requirements to obtain a plurality of monitoring data types, and performing sensor requirements analysis based on the plurality of monitoring data types to obtain a plurality of sensor types, wherein the plurality of sensor types include visible light cameras, multispectral sensors, thermal imaging sensors, and hyperspectral sensors; Interactively obtain the load range of the UAV, and perform sensor configuration based on the load range and multiple sensor types as constraints to obtain multiple sensor entities; After traversing the installation reference position table using the multiple sensor types to obtain multiple installation positions, the multiple sensor entities are installed on the drone using the multiple installation positions as a guide to complete the installation configuration of the multimodal sensor.

3. A forest disease monitoring method as claimed in claim 2, characterized in that: The multimodal monitoring data is spatially aligned to obtain a plurality of regional monitoring information maps, the method comprising: Generate a regional plan view of the forest monitoring area based on the projection of the regional terrain features; Calling the multispectral image data collected by the multispectral sensor from the multimodal monitoring data; After calculating the NDVI value of the multispectral image data, the multispectral image data is superimposed on the regional plane map according to the monitoring flight trajectory to obtain a first regional monitoring information map, wherein the first regional monitoring information map is a vegetation health index distribution map; By analogy, after preprocessing the multimodal monitoring data according to multiple data preprocessing requirements of the multiple sensor types, spatial alignment is performed based on the regional plan to obtain the multiple regional monitoring information maps, wherein the multiple regional monitoring information maps include RGB image distribution maps, vegetation health index distribution maps, temperature distribution maps, and spectral reflectance distribution maps.

4. A forest disease monitoring method as claimed in claim 3, characterized in that: By collaboratively analyzing the multiple regional monitoring information graphs to identify and select diseased areas, the disease risk area distribution is obtained, and the method includes: Using 1 / K of the forest monitoring area as an image segmentation scale to rasterize the multiple regional monitoring information maps, and obtaining multiple groups of segmented monitoring information maps; Performing sample data networking scheduling based on the multiple sensor types as constraints to obtain multiple sample monitoring information atlases and multiple sample disease type identification sets; A plurality of disease type recognition models are constructed by using the plurality of sample monitoring information atlases and the plurality of sample disease type identification sets; Connecting the plurality of disease type identification models in parallel to complete the construction of a forest disease identification model; After receiving the multiple groups of segmented monitoring information graphs, the forest disease recognition model automatically identifies the types of forest diseases and outputs multiple groups of segmented disease identifiers; According to the mapping relationship between the multiple groups of segmented disease identifications and the multiple groups of segmented monitoring information maps, disease area restoration is performed on the multiple regional monitoring information maps to obtain multiple disease distribution maps; The multiple disease distribution maps are overlaid and analyzed to output the disease risk area distribution.

5. A forest disease monitoring method as claimed in claim 4, characterized in that: According to the mapping relationship between the multiple groups of segmented disease identifications and the multiple groups of segmented monitoring information maps, disease area restoration is performed on the multiple regional monitoring information maps to obtain multiple disease distribution maps, the method comprising: According to the mapping relationship between the first group of segmented disease identifications and the first group of segmented monitoring information maps, disease intensity calculation is performed in the first area monitoring information map to obtain a first disease pixel gradient map; Presetting a disease intensity threshold, and using the disease intensity threshold to traverse the first disease pixel gradient map to obtain a plurality of initial seed grids; In the first disease pixel gradient map, similar disease grids are merged starting from the multiple initial seed grids until they are extended to the boundary area to obtain multiple disease sub-areas, wherein the multiple disease sub-areas have multiple sub-area disease type identifiers; Merge the multiple disease sub-areas of the same type according to the disease types of the multiple sub-areas to obtain a first disease distribution map; By analogy, based on the mapping relationship between the multiple groups of segmented disease identifications and the multiple groups of segmented monitoring information maps, disease area restoration is performed on the multiple regional monitoring information maps to obtain the multiple disease distribution maps.

6. A forest disease monitoring method as claimed in claim 4, characterized in that: Taking the disease risk area distribution as a starting point, predicting the disease spread according to the disease type, and selecting a hierarchical disease spread area, the method includes: Aggregating the multiple sample disease type identification sets to obtain multiple sample disease types and multiple groups of disease intensity sequences; Using the multiple groups of disease intensity sequences as constraints, disease diffusion scales are called to obtain multiple groups of diffusion scale sequences; According to the consistency of the scale parameters of the multiple groups of diffusion scale sequences, the multiple groups of disease intensity sequences are graded to obtain multiple levels of disease intensity intervals; The plurality of sample disease types, the multi-level disease intensity intervals and the plurality of groups of diffusion scale sequences are stored in association to complete the construction of the diffusion information library; The disease risk area distribution is used to traverse the diffusion information database to obtain a plurality of diffusion scale information; The plurality of diffusion scale information are used to distribute the disease risk area to select the disease diffusion area, and the hierarchical disease diffusion area is obtained.

7. A forest disease monitoring method as claimed in claim 2, characterized in that: The multimodal monitoring data is spatially aligned to obtain a plurality of regional monitoring information maps, the method comprising: Performing local data call on the drone to obtain multiple effective monitoring ranges of the multiple sensor entities in the multimodal sensor, wherein the multiple effective monitoring ranges constitute the multimodal parameters; Using geometric Boolean operations to solve the spatial intersection of the multiple effective monitoring ranges to obtain a standard monitoring range; By vertically projecting the standard monitoring range, the vertical span of the track is obtained; The horizontal span of the track is obtained by horizontally projecting the standard monitoring range; Fitting a horizontal spiral trajectory in the forest monitoring area with the horizontal span of the trajectory as a constraint; Extracting a topographic change sequence from the regional topographic features according to the horizontal spiral trajectory; The trajectory height of the horizontal spiral trajectory is dynamically updated according to the terrain change sequence and the horizontal span of the trajectory to obtain the monitored flight trajectory.

8. A forest disease monitoring system, characterized in that: The system is used to implement a forest disease monitoring method according to any one of claims 1 to 7, and the system comprises: A regional terrain feature acquisition module, which is used to obtain regional terrain features of the forest monitoring area from a GIS system call; A monitoring flight trajectory generation module, which is used to call local data of the UAV, obtain multimodal parameters of the multimodal sensor, and then generate a monitoring flight trajectory according to the multimodal parameters and regional terrain features; A multimodal monitoring data acquisition module, wherein the multimodal monitoring data acquisition module is used to control the UAV to fly in the forest monitoring area using the monitoring flight trajectory, start the multimodal sensor to synchronously collect forest data, and obtain multimodal monitoring data; A regional monitoring information map acquisition module, the regional monitoring information map acquisition module is used to spatially align the multimodal monitoring data to obtain a plurality of regional monitoring information maps; A disease risk area distribution acquisition module, the disease risk area distribution acquisition module is used to identify and select disease areas by collaboratively analyzing the multiple regional monitoring information graphs to obtain the disease risk area distribution; A hierarchical disease diffusion area acquisition module is used to take the disease risk area distribution as the starting point, predict the disease diffusion according to the disease type, and select the hierarchical disease diffusion area.

Images