Mountain rape unmanned aerial vehicle accurate variable fertilization and pesticide spraying integrated cooperation method

Abstract

According to the precise variable fertilization and pesticide spraying integrated collaborative method for the mountain rape unmanned aerial vehicle, heuristic function layered weighting and node gradient continuity inspection in the improved A star algorithm can generate a flight path highly adaptive to mountain topographic features, the path has the advantages of low energy consumption and full coverage, and the flight path has the advantages of being high in adaptability to the mountain topographic features. Operation hindrance caused by complex terrain can be effectively avoided; in an operation parameter calculation link, an improved weighted multiple linear regression model is constructed, a dynamic space-time weight item is utilized to optimize a characteristic variable, a residual error correction mechanism is combined to improve calculation precision, and accurate quantification of fertilization and pesticide spraying parameters is realized. On the basis, collaborative linkage of path planning and parameter calculation is established through a regional weight association mechanism, the adaptability of the path planning and the parameter calculation is remarkably enhanced, the efficiency and accuracy of unmanned aerial vehicle operation in the complex mountain environment are effectively improved, waste of resources such as fertilizer and chemicals is reduced, and the economic benefit is improved. Meanwhile, the dual adaptability to the rape operation growth dynamic state and the mountain topographic feature is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of agricultural resource regulation, and in particular to a mountainous rape unmanned aerial vehicle precision variable fertilization and pesticide spraying integrated collaborative method. BACKGROUND

[0002] In the mountainous rape planting scene, the traditional operation mode is restricted by the complex terrain of the mountainous area, and has the following disadvantages: the path planning and terrain adaptability are insufficient, and it is difficult to match the undulating topography of the mountainous area; the variable control cannot take into account the regional differences and time dynamic changes, and cannot realize differentiated management; the algorithm collaboration is insufficient, and the model environment adaptability is poor, resulting in low overall operation system collaboration efficiency, causing resource verification waste, low operation efficiency and poor precision. At the same time, the existing unmanned aerial vehicle agricultural operation scheme has defects in algorithm application, and cannot meet the integrated collaborative demand of precision fertilization and pesticide spraying in the mountainous environment.

[0003] In the related art, a control cloud platform is built, and a field is artificially selected based on a high-precision map and a path is automatically generated, so as to realize unmanned operation of a rape seeder machine group, and the sowing trajectory information can be stored in real time; at the same time, the navigation control platform is composed of multiple devices, covering data acquisition, planning control and action execution, and together completes the sowing task. However, the technical scheme does not involve variable fertilization and pesticide dispensing, and the adaptability to the complex terrain of the mountainous area is insufficient, the path planning and terrain adaptability are poor, and the variable control ignores the space-time differences.

[0004] Therefore, how to realize the integrated collaborative operation of precision variable fertilization and pesticide spraying becomes a problem to be solved. SUMMARY

[0005] The embodiment of the present application provides a mountainous rape unmanned aerial vehicle precision variable fertilization and pesticide spraying integrated collaborative method, so as to solve the problem that the unmanned aerial vehicle agricultural operation in the prior art has defects in algorithm application, and cannot meet the integrated collaborative demand of precision fertilization and pesticide spraying in the mountainous environment.

[0006] In order to achieve the above purpose, the technical scheme of the embodiment of the present application is as follows: In a first aspect, the present application provides a mountainous rape unmanned aerial vehicle precision variable fertilization and pesticide spraying integrated collaborative method, which comprises: obtaining multi-source heterogeneous data of a mountainous rape planting area; the multi-source heterogeneous data comprises terrain elevation data, rape planting area boundary data, soil nutrient data and crop growth index data; performing disease and pest risk level evaluation and division on the crop growth index data to obtain a sub-regional division result; The multi-source heterogeneous data is cleaned, standardized in format, and missing data is filled in; the preprocessed multi-source heterogeneous data is associated with a unified geographic or sub-regional unit to form a structured dataset. The preprocessed terrain elevation data, preprocessed rapeseed planting area boundary data, and sub-region division results in the structured dataset are input into the improved A-Star algorithm. Through the heuristic function of the improved A-Star algorithm, hierarchical weighting and node slope continuity check, the target flight path with the lowest energy consumption and full coverage is generated. The preprocessed soil nutrient data and preprocessed crop growth index data in the structured dataset are input into an improved weighted multiple linear regression model trained based on the target flight path. The dynamic spatiotemporal weights in the improved weighted multiple linear regression model are used to calculate the amount of fertilizer and pesticides applied to each sub-region in the mountain rapeseed planting area.

[0007] In some embodiments, after calculating the fertilization and pesticide application rates for each sub-region in the mountain rapeseed planting area, the method further includes: Within a preset time window after fertilization and spraying operations are completed, crop height growth, leaf area index changes, and pest and disease reduction rates are collected in each sub-region as comparative data on fertilization and spraying effects. Based on this comparative data, the crop growth response in each sub-region is evaluated to determine whether it meets the expected target. The deviation between the actual growth indicators and the target growth indicators is calculated and used as the target error for fertilization and spraying effects. For sub-regions where the target error exceeds a preset error threshold, the sample weights of historical sample data in the corresponding sub-regions are corrected using a residual correction mechanism. Combining the path energy consumption data, operation coverage integrity data, and the sample weights of each sub-region during the UAV operation, the terrain cost factor threshold in the improved AlphaGo algorithm is dynamically adjusted using a parameter optimization method.

[0008] In some embodiments, the method further includes: based on the target flight path, using sensors mounted on a drone to collect soil nutrient replenishment data and crop growth status update data of the mountain rapeseed planting area, inputting the collected new data into the improved weighted multiple linear regression model, triggering the dynamic spatiotemporal weight term in the model to recalculate the sample weights, and updating the model parameters through weighted least squares.

[0009] In some embodiments, the topographic elevation data, the rapeseed planting area boundary data, the soil nutrient data, and the crop growth index data are obtained through the following methods: A digital elevation model is constructed using a drone equipped with lidar or multi-view photogrammetry combined with RTK positioning to obtain the terrain elevation data. Satellite remote sensing imagery or a drone's visible light camera, combined with manual interpretation and automatic vectorization technology from a geographic information system, is used to delineate the boundaries of the rapeseed planting area in the mountains, determine the spatial range of the drone's operation, and mark the boundaries using GPS positioning equipment to obtain the boundary data of the rapeseed planting area. Deep soil samples from the mountain rapeseed planting area are collected based on a ground-based gridded sampling strategy. Nutrient baseline values ​​are obtained through laboratory chemical analysis, and the current content of key elements for crop growth is quickly determined using a near-infrared spectrometer combined with a pre-calibrated prediction model. Combined with dynamic monitoring and periodic updates, initial soil samples are collected before planting and their initial contents are determined to form soil nutrient data including current and initial contents. Based on a hyperspectral sensor combined with a ground-based fixed-point strategy, vegetation reflectance spectral data, plant height, and leaf area index are obtained to construct the crop growth index data.

[0010] In some embodiments, generating the target flight path with the lowest energy consumption and full coverage through hierarchical weighting and node slope continuity checks using the heuristic function of the improved A* algorithm includes: Based on the improved A* algorithm, the preprocessed terrain elevation data is transformed into a terrain cost factor, and the preprocessed rapeseed planting area boundary data and the sub-region division results are set as the operation coverage factor to construct a path search cost function. The path search direction is guided by the hierarchical weighting of the cost function, and the slope continuity data in the preprocessed terrain elevation data is called to check the slope difference between adjacent nodes, generating a target flight path containing the coordinates of turning points and the relative crop canopy height. The operation coverage factor is used to quantify the cost parameter of regional coverage priority in the path search. The cost function is determined based on the basic distance cost, the terrain cost factor, and the operation coverage factor.

[0011] In some embodiments, the step of inputting preprocessed soil nutrient data and preprocessed crop growth index data from the structured dataset into an improved weighted multiple linear regression model trained based on the target flight path, and using the dynamic spatiotemporal weights in the improved weighted multiple linear regression model to calculate the fertilization and pesticide application rates for each sub-region in the mountain rapeseed planting area, includes: The input data includes the preprocessed soil nutrient data and preprocessed crop growth index data, initial time labels, and initial regional attribute labels. The initial time labels are used to distinguish the timeliness of the data, and the initial regional attribute labels are used to mark whether the corresponding data area is a high-incidence area of ​​pests and diseases or an area with abnormal growth. Based on the initial time labels, the preprocessed soil nutrient data and preprocessed crop growth index data are classified according to the collection time and assigned time weights. Based on the initial regional attribute labels, regional weights are assigned to the classified soil nutrient data and crop growth index data. The input data is weighted based on the time weights and the regional weights to obtain weighted data. The weighted data is substituted into the improved weighted multiple linear regression model, and the fertilizer application rate and the pesticide application rate are obtained by minimizing the loss function containing the spatiotemporal weight term.

[0012] In some embodiments, the improved weighted multiple linear regression model is obtained by training through the following steps: Based on the collected basic data of the mountain rapeseed planting area, quantitative calculation rules for time weight and regional weight are established, and constraints on the range of weight values ​​are set to form a dynamic spatiotemporal weight term for an improved weighted multiple linear regression model. The basic data includes sample soil nutrient data, sample crop growth index data, sample time labels, and sample regional attribute labels. The basic data is used as a training sample set, and a weighted loss function is constructed based on the dynamic spatiotemporal weight term, wherein the weighted loss function is the weighted sum of squares of the prediction errors of each sample. By minimizing the weighted loss function, the model regression parameters are solved using the weighted least squares method to generate an initial model, wherein the output of the initial model is the recommended fertilization and pesticide application rates for each sub-region. The crop growth dynamics data and equipment operation status data collected in real time during the drone's flight path are updated with time weights according to the quantification calculation rules. The regional weights are also dynamically adjusted based on the deviation between the sample region attribute labels and the real-time detection values. Based on the updated dynamic spatiotemporal weights, the initial model is iteratively updated with new sample data to obtain updated regression parameters, and the corresponding updated fertilization and pesticide application rates are output. The adaptability of the updated model parameters is verified based on the matching degree between the actual field growth response and the preset target. The weight update step is repeated, continuously providing feedback updates until the model output meets the preset convergence conditions, thus obtaining the trained improved weighted multiple linear regression model.

[0013] Secondly, the present invention provides an integrated collaborative device for precision variable fertilization and pesticide spraying using unmanned aerial vehicles (UAVs) in mountainous rapeseed areas, the device comprising: The acquisition module is used to acquire multi-source heterogeneous data of rapeseed planting areas in mountainous areas; the multi-source heterogeneous data includes topographic elevation data, rapeseed planting area boundary data, soil nutrient data and crop growth index data; the crop growth index data is used to assess and classify the risk level of pests and diseases, and to obtain the sub-region division results; The forming module is used to clean and standardize the format of the multi-source heterogeneous data and fill in the missing data; and associate the preprocessed multi-source heterogeneous data with a unified geographic or sub-regional unit to form a structured dataset. The generation module is used to input the preprocessed terrain elevation data, preprocessed rapeseed planting area boundary data and the sub-region division results from the structured dataset into the improved A-Star algorithm. Through the heuristic function of the improved A-Star algorithm, hierarchical weighting and node slope continuity check, the target flight path with the lowest energy consumption and full coverage is generated. The calculation module is used to input the preprocessed soil nutrient data and preprocessed crop growth index data from the structured dataset into an improved weighted multiple linear regression model trained based on the target flight path, and to calculate the amount of fertilizer and pesticide spraying in each sub-region of the mountain rapeseed planting area using the dynamic spatiotemporal weight term in the improved weighted multiple linear regression model.

[0014] Thirdly, embodiments of the present invention provide an electronic device, including: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to realize the above-mentioned integrated method for precise variable fertilization and spraying of rapeseed using a drone in mountainous areas.

[0015] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing executable instructions, which, when executed by a processor, enable the aforementioned integrated method for precise variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas.

[0016] The present invention provides a method for integrated and coordinated precision variable fertilization and pesticide spraying using unmanned aerial vehicles (UAVs) in mountainous rapeseed areas. First, multi-source heterogeneous data of the mountainous rapeseed planting area is acquired. Crop growth index data are then assessed for pest and disease risk levels and divided into sub-regions. The multi-source heterogeneous data is cleaned, standardized, and missing data is supplemented. The pre-processed multi-source heterogeneous data is then correlated and integrated to form a structured dataset. The pre-processed terrain elevation data, pre-processed rapeseed planting area boundary data, and sub-region division data from the structured dataset are input into an improved A-Star algorithm. Through the heuristic function of the improved A-Star algorithm, hierarchical weighting, and node slope continuity checks, a target flight path with the lowest energy consumption and full coverage is generated. The pre-processed soil nutrient data and pre-processed crop growth index data from the structured dataset are input into an improved weighted multiple linear regression model trained based on the target flight path. Using the dynamic spatiotemporal weights in the improved weighted multiple linear regression model, the fertilization and pesticide application rates for each sub-region in the mountainous rapeseed planting area are calculated. Thus, the heuristic function layered weighting and node slope continuity check in the improved A* algorithm provided by this invention can generate flight paths highly adapted to mountainous terrain features. These paths combine low energy consumption and full coverage, effectively avoiding operational obstacles caused by complex terrain. In the operational parameter calculation stage, an improved weighted multiple linear regression model is constructed, utilizing dynamic spatiotemporal weight terms to optimize feature variables, and combining a residual correction mechanism to improve calculation accuracy, achieving precise quantification of fertilization and spraying parameters. Based on this, a regional weight association mechanism is established to establish a synergistic linkage between path planning and parameter calculation, significantly enhancing their adaptability and effectively improving the efficiency and accuracy of UAV operations in complex mountainous environments, reducing waste of resources such as fertilizers and pesticides, while simultaneously strengthening the dual adaptability to rapeseed growth dynamics and mountainous terrain features. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the integrated collaborative system for precision variable fertilization and pesticide spraying of rapeseed using drones in mountainous areas, provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the integrated and coordinated method for precise variable fertilization and pesticide spraying using drones in mountainous rapeseed areas, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the composition of the integrated collaborative device for precision variable fertilization and pesticide spraying of rapeseed using drones in mountainous areas, provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of the composition structure of the electronic device provided in the embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] In the following description, references to "some embodiments" refer to a subset of all possible embodiments; however, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. Unless otherwise defined, all technical and scientific terms used in the embodiments of the invention have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments of the invention pertain. The terminology used in the embodiments of the invention is for the purpose of describing the embodiments of the invention only and is not intended to limit the invention.

[0020] The following describes an exemplary application of the integrated precision variable fertilization and spraying device for rapeseed using drones in mountainous areas, as described in this embodiment of the invention. This integrated device can be implemented as a terminal or a server. In one implementation, the integrated device can be implemented as a laptop, tablet, desktop computer, mobile device, or other types of terminal. In another implementation, it can also be implemented as a server. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited in this embodiment of the invention. The following will illustrate an exemplary application of the integrated collaborative equipment for precision variable fertilization and spraying of rapeseed using drones in mountainous areas, when the server is used as the server.

[0021] See Figure 1 , Figure 1This is a schematic diagram of the integrated collaborative system 10 for precision variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas, provided in an embodiment of the present invention. To achieve integrated collaboration of precision variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas, this embodiment of the present invention can provide an integrated collaborative platform for precision variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas. This platform can be implemented as an integrated collaborative application for precision variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas. The integrated collaborative system 10 for precision variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas provided in this embodiment of the present invention includes a terminal 110, a network 120, and a server 130. The server 130 is the server for the integrated collaborative application of precision variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas. The server 130 can constitute the integrated collaborative device for precision variable fertilization and pesticide spraying using a drone in mountainous rapeseed areas according to this embodiment of the present invention. The terminal 110 is connected to the server 130 through the network 120, which can be a wide area network (WAN), a local area network (LAN), or a combination of both.

[0022] In some embodiments, please refer to Figure 1 When integrating fertilization and pesticide spraying in a mountain rapeseed planting area, terminal 110 sends multi-source heterogeneous data of the mountain rapeseed planting area to server 130 via network 120. Server 130, in response to terminal 110 receiving the multi-source heterogeneous data of the mountain rapeseed planting area, assesses and classifies the crop growth index data for pest and disease risk levels, obtaining sub-region division results; cleans and standardizes the format of the multi-source heterogeneous data, and completes missing data; it then integrates and correlates the pre-processed multi-source heterogeneous data to form a structured dataset; and finally, it integrates the pre-processed topographic elevation data from the structured dataset. The pre-processed boundary data and sub-region division results of the rapeseed planting area are input into the improved A* algorithm. Through the heuristic function of the improved A* algorithm, hierarchical weighting and node slope continuity checks, a target flight path with the lowest energy consumption and full coverage is generated. Pre-processed soil nutrient data and pre-processed crop growth index data from the structured dataset are input into an improved weighted multiple linear regression model trained based on the target flight path. Using the dynamic spatiotemporal weights in the improved weighted multiple linear regression model, the fertilization and pesticide application rates for each sub-region in the mountain rapeseed planting area are calculated, obtaining the integrated fertilization and pesticide application results. After obtaining the integrated fertilization and pesticide application results, server 130 sends the results to terminal 110 via network 120.

[0023] This invention provides a method for the integrated and coordinated application of precision variable fertilization and pesticide spraying using unmanned aerial vehicles (UAVs) in mountainous rapeseed areas. (See also...) Figure 2 , Figure 2 This is a flowchart illustrating an integrated method for precise variable-rate fertilization and pesticide spraying using unmanned aerial vehicles (UAVs) in mountainous rapeseed areas, provided by an embodiment of the present invention. Figure 2 The steps shown are explained.

[0024] Step S210: Obtain multi-source heterogeneous data of the rapeseed planting area in the mountainous region; the multi-source heterogeneous data includes topographic elevation data, rapeseed planting area boundary data, soil nutrient data and crop growth index data; assess and classify the crop growth index data for pest and disease risk levels to obtain sub-region division results.

[0025] In this embodiment of the invention, the topographic elevation data, rapeseed planting area boundary data, soil nutrient data, and crop growth index data in step S210 are obtained through the following methods: Step 211: Using a drone equipped with a lidar or based on multi-view photogrammetry combined with RTK positioning, a digital elevation model is constructed to obtain the terrain elevation data.

[0026] In some embodiments, topographic elevation data refers to spatial data used to describe the elevation of various points on the ground and the slope of the mountains within the rapeseed planting area.

[0027] In this invention, the data comes from LiDAR carried by a drone or from multi-view photogrammetry combined with RTK high-precision positioning technology to obtain terrain elevation data with centimeter-level accuracy, accurately depicting the slope, altitude and other terrain features of the mountains. The update frequency is once a year, and an emergency update is triggered when there are significant changes in the mountain terrain (such as large-scale reclamation or geological disasters). The data is obtained through drone LiDAR scanning or satellite remote sensing data, and the terrain is modeled and the elevation is corrected by GIS software.

[0028] Step 212: Using satellite remote sensing imagery or UAV visible light photography camera, combined with the manual interpretation and automatic vectorization technology of geographic information system, delineate the boundary of the rapeseed planting area in the mountainous area, determine the spatial range of UAV operation, and use GPS positioning equipment to mark the boundary to obtain the boundary data of the rapeseed planting area.

[0029] In some embodiments, rapeseed planting area boundary data is used as spatial vector data to accurately define the actual planting area of ​​rapeseed in mountainous areas, clearly distinguishing between rapeseed planting areas and non-planting areas. This data is a key basis for dividing planting areas into sub-regions and determining the scope of operations (such as fertilization and spraying).

[0030] In this invention, high-resolution satellite remote sensing images or UAV visible light photography are used, combined with the manual interpretation and automatic vectorization technology of Geographic Information System (GIS), to accurately delineate the boundaries of rapeseed planting areas in mountainous areas, clarify the spatial range of UAV operations, update the data before the start of the planting season, combine farmers' planting plans and on-site surveys, use GPS positioning equipment to mark the boundaries, import the data into the geographic information system to generate a digital boundary layer, and obtain rapeseed planting area boundary data.

[0031] Step 213: Based on the ground grid sampling strategy, deep soil samples of the mountain rapeseed planting area are collected. Nutrient baseline values ​​are obtained through laboratory chemical analysis. Near-infrared spectrometer combined with a pre-calibrated prediction model is used to quickly determine the current content of key elements for crop growth. Combined with dynamic monitoring and periodic updates, initial soil samples are collected before planting and the initial content is determined to form soil nutrient data including current content and initial content.

[0032] In some embodiments, soil nutrient data are data reflecting the content of essential nutrients and related physicochemical properties in the soil of mountain rapeseed planting areas that are crucial for rapeseed growth.

[0033] In this invention, based on a ground-based gridded sampling strategy, soil samples from the 0-20cm depth of the rapeseed planting area are collected in a checkerboard or serpentine pattern within the planting area. Nutrient baseline values ​​are obtained through laboratory chemical analysis, and the current content of key crop growth elements, such as nitrogen (N), phosphorus (P), and potassium (K), is rapidly determined using a pre-calibrated prediction model via near-infrared spectroscopy. Simultaneously, combined with dynamic monitoring and periodic updates, initial soil samples are collected before the planting season, and the content of elements such as nitrogen (N), phosphorus (P), and potassium (K) is determined using instruments such as spectrometers and ion chromatographs. During the growth period, the surface soil is rapidly detected every 15 days using a portable soil sensor, and a deep profile sampling analysis is performed every 30 days. The data is automatically uploaded to a cloud database and updated accordingly.

[0034] Step 214: Based on the hyperspectral sensor and ground-based fixed-point strategy, acquire vegetation reflectance spectral data, plant height and leaf area index to construct the crop growth index data.

[0035] In some embodiments, crop growth index data refers to quantitative data on the growth status of rapeseed in mountainous areas at different growth stages (such as seedling stage, budding stage, flowering stage, and pod stage). Core indicators include plant height, stem diameter, leaf area index (LAI), chlorophyll content (SPAD value), etc.

[0036] In the risk assessment and regional division of pests and diseases in mountainous rapeseed planting areas, the distinction between ordinary planting areas and important planting areas can be accurately achieved through the synergistic analysis of crop growth index data and soil nutrient data. This can be accomplished through the following methods: Among these factors, soil nutrient imbalance is a significant contributing factor to the high incidence of pests and diseases. For example, excessive nitrogen content in the soil can easily lead to aphid aggregation; insufficient phosphorus / potassium (P / K) content reduces plant disease resistance and increases the incidence of sclerotinia stem rot. Abnormal crop growth is a direct manifestation of pests and diseases. An abnormally low SPAD value may be caused by aphids sucking sap or sclerotinia stem rot damaging the leaves; a significant reduction in the number of branches compared to the normal level for the same growth stage may be the result of insufficient soil phosphorus content combined with downy mildew.

[0037] Based on the above-mentioned correlation between basic soil nutrient induction factors and crop growth manifestation feedback, in the actual pest and disease risk assessment process, the comprehensive risk score of soil nutrient data and crop growth index data can be quantitatively calculated to achieve regional classification; areas with high pest and disease risk, high production value, and requiring priority intervention and protection are classified as important planting areas; areas with low pest and disease risk and production value at a normal level are classified as ordinary planting areas.

[0038] In this invention, multispectral / hyperspectral sensors are deployed on drones or ground monitoring stations to acquire vegetation reflectance spectral data once a week, with a focus on collecting plant height and leaf area. The data is calibrated by combining ground-based fixed-point measurements (3-5 monitoring points per hectare) and synchronized to the operation management platform in real time, providing a basis for variable fertilization and irrigation decisions to reflect the growth status of rapeseed.

[0039] Step S220: Clean the multi-source heterogeneous data, standardize its format, and fill in any missing data; associate the preprocessed multi-source heterogeneous data with a unified geographic or sub-regional unit to form a structured dataset.

[0040] In some embodiments, data association and integration refers to the process of establishing logical associations between multiple types of data, such as terrain, soil, crops, and sub-regions, through sub-region numbering and time labels (time stamps) on the basis of data cleaning and format standardization, to form an integrated dataset.

[0041] In some embodiments, a structured dataset refers to a dataset that integrates cleaned and standardized terrain data, rapeseed planting area boundary data, soil nutrient data, crop growth index data, and sub-region division data through spatial location and time labels. This results in a dataset with a unified data structure that can be directly used as algorithm input and analysis; that is, a dataset that links terrain, soil, crops, and planting area boundaries. This dataset directly relates to the adaptability of path planning and the accuracy of variable control. Spatially, this dataset binds various types of data through coordinates or sub-regions; temporally (including dynamic data), it links crop growth data at different growth stages through timestamps. The data structure is typically a structured tabular data structure of sub-region number - terrain attributes (elevation, slope, aspect) - soil nutrient attributes - crop growth attributes (plant height, leaf area index) - boundary coordinates.

[0042] In this embodiment of the invention, step S220 can be implemented by the following steps 221 to 223: Step 221: The multi-source heterogeneous data is filtered for outliers based on the threshold filtering method, and the filtered outliers are identified and removed to obtain the removed multi-source heterogeneous data.

[0043] In some embodiments, data cleaning refers to identifying and removing outliers and duplicates from various original data to obtain heterogeneous multi-source data after removal.

[0044] Step 222: Based on the coordinate transformation method and the unit conversion method, the removed multi-source heterogeneous data is standardized to obtain standardized multi-source heterogeneous data.

[0045] In some embodiments, data format standardization refers to the process of converting raw data from different sources and in different formats (such as GeoTIFF format for topographic elevation data, SHP format for planting area boundary data, Excel spreadsheet format for soil nutrients, and CSV format for growth indicators) into a unified and standardized format to ensure data compatibility and correlation. In this invention, standardization can be achieved through coordinate transformation and unit conversion methods.

[0046] Step 223: Interpolate and supplement the standardized data to obtain preprocessed multi-source heterogeneous data.

[0047] In some embodiments, interpolation supplementation refers to supplementing missing data in the data after the above cleaning and standardization processes to ensure that the data has a certain degree of continuity.

[0048] Step S230: Input the preprocessed terrain elevation data, preprocessed rapeseed planting area boundary data and the sub-region division results from the structured dataset into the improved A-Star algorithm. Through the heuristic function of the improved A-Star algorithm, hierarchical weighting and node slope continuity check, generate the target flight path with the lowest energy consumption and full coverage.

[0049] In some embodiments, step S230 above can also be implemented through steps 231 to 232: Step 231: Based on the improved A* algorithm, the preprocessed terrain elevation data is converted into a terrain cost factor, and the preprocessed rapeseed planting area boundary data and the sub-region division results are set as the operation coverage factor to construct the cost function of path search.

[0050] Step 232: The path search direction is guided by the hierarchical weighting of the cost function, and the slope continuity data in the preprocessed terrain elevation data is called to check the slope difference between adjacent nodes, generating a target flight path that includes the coordinates of turning points and the relative crop canopy height; wherein, the operation coverage factor is used to quantify the cost parameter of regional coverage priority in the path search; the cost function is determined based on the basic distance cost, terrain cost factor and operation coverage factor.

[0051] In this invention, the heuristic function hierarchical weighting and node slope continuity checking are the core improvements to the A* algorithm. Hierarchical weighting of the heuristic function involves incorporating mountain terrain features (such as slope and elevation difference) and operational requirements (such as sub-region coverage priority) into the heuristic term and assigning them different weights. For example, areas with steep slopes are given higher weights to guide the path away from high-slope areas and reduce energy consumption; sub-regions with abnormal crop growth are given higher weights to prioritize coverage of these areas. Node slope continuity checking involves checking the slope changes of adjacent path nodes during path search, limiting the slope difference between adjacent nodes, avoiding drastic flight attitude adjustments by the UAV due to sudden slope changes, and reducing energy consumption and operational risks.

[0052] Step S240: Input the preprocessed soil nutrient data and preprocessed crop growth index data from the structured dataset into the improved weighted multiple linear regression model trained based on the target flight path, and use the dynamic spatiotemporal weight term in the improved weighted multiple linear regression model to calculate the fertilization amount and pesticide spraying amount of each sub-region in the mountain rapeseed planting area.

[0053] In some embodiments, the improved weighted multiple linear regression model refers to a model trained with historical data and optimized to accurately predict fertilizer and pesticide application rates in various sub-regions of rapeseed-growing areas in mountainous regions. The core improvement lies in the dynamic spatiotemporal weighting term, which assigns higher weights to sub-regions with large soil nutrient deviations, such as nitrogen content far below rapeseed requirements and poor crop growth, making the model more focused on the fertilizer and pesticide application needs of these areas; and assigns higher weights to key growth periods requiring fertilizer and pesticides, making the model's predictions more aligned with the actual needs at different times.

[0054] In some embodiments, step S240 may further include steps 241 to 243: Step 241: Obtain input data, which includes the preprocessed soil nutrient data and the preprocessed crop growth index data, initial time tags and initial regional attribute tags; the initial time tags are used to distinguish the timeliness of the data, and the initial regional attribute tags are used to mark whether the corresponding data area is a high-incidence area of ​​pests and diseases or an area with abnormal growth.

[0055] Step 242: Based on the initial time label, classify the preprocessed soil nutrient data and the preprocessed crop growth index data according to the collection time and assign time weights; based on the initial regional attribute label, assign regional weights to the classified soil nutrient data and crop growth index data; perform weighted processing on the input data based on the time weights and the regional weights to obtain weighted data.

[0056] Step 243: Substitute the weighted data into the improved weighted multiple linear regression model, and solve for the fertilizer application amount and the pesticide application amount by minimizing the loss function containing the spatiotemporal weight term.

[0057] In some embodiments, the improved weighted multiple linear regression model is obtained by training through the following steps: Based on the collected basic data of the mountain rapeseed planting area, quantitative calculation rules for time weight and regional weight are established, and constraints on the range of weight values ​​are set to form a dynamic spatiotemporal weight term for an improved weighted multiple linear regression model. The basic data includes sample soil nutrient data, sample crop growth index data, sample time labels, and sample regional attribute labels. The basic data is used as a training sample set, and a weighted loss function is constructed based on the dynamic spatiotemporal weight term, wherein the weighted loss function is the weighted sum of squares of the prediction errors of each sample. By minimizing the weighted loss function, the model regression parameters are solved using the weighted least squares method to generate an initial model, wherein the output of the initial model is the recommended fertilization and pesticide application rates for each sub-region. The crop growth dynamics data and equipment operation status data collected in real time during the drone's flight path are updated with time weights according to the quantification calculation rules. The regional weights are also dynamically adjusted based on the deviation between the sample region attribute labels and the real-time detection values. Based on the updated dynamic spatiotemporal weights, the initial model is iteratively updated with new sample data to obtain updated regression parameters, and the corresponding updated fertilization and pesticide application rates are output. The adaptability of the updated model parameters is verified based on the matching degree between the actual field growth response and the preset target. The weight update step is repeated, continuously providing feedback updates until the model output meets the preset convergence conditions, thus obtaining the trained improved weighted multiple linear regression model.

[0058] In some embodiments, after performing step S240 above, the method may further include steps 251 to 253: Step 251: Within a preset time window after fertilization and spraying operations are completed, collect data on crop height growth, leaf area index changes, and pest and disease reduction rates in each sub-region as comparative data on fertilization and spraying effects.

[0059] Step 252: Based on the comparison data of fertilization and pesticide application effects, evaluate whether the crop growth response of each sub-region has reached the expected target, calculate the deviation between the actual growth index and the target growth index, and use it as the target error of the fertilization and pesticide application effect.

[0060] Step 253: For the sub-regions where the target error is greater than the preset error threshold, the sample weights of the historical sample data in the corresponding sub-regions are corrected through the residual correction mechanism; combined with the path energy consumption data, operation coverage integrity data and the sample weights of each sub-region during the UAV operation, the terrain cost factor threshold in the improved A-Star algorithm is dynamically adjusted and improved through the parameter optimization method.

[0061] Based on the above embodiments, in some embodiments, the above method may further include the following: based on the target flight path, using a drone equipped with sensors to collect soil nutrient replenishment data and crop growth status update data of the mountain rapeseed planting area, and inputting the collected new data into the improved weighted multiple linear regression model, triggering the dynamic spatiotemporal weight term in the model to recalculate the sample weights, and updating the model parameters through weighted least squares method.

[0062] The present invention provides a method for integrated and coordinated precision variable fertilization and pesticide spraying using unmanned aerial vehicles (UAVs) in mountainous rapeseed areas. First, multi-source heterogeneous data of the mountainous rapeseed planting area is acquired. Crop growth index data are then assessed for pest and disease risk levels and divided into sub-regions. The multi-source heterogeneous data is cleaned, standardized, and missing data is supplemented. The pre-processed multi-source heterogeneous data is then correlated and integrated to form a structured dataset. The pre-processed terrain elevation data, pre-processed rapeseed planting area boundary data, and sub-region division data from the structured dataset are input into an improved A-Star algorithm. Through the heuristic function of the improved A-Star algorithm, hierarchical weighting, and node slope continuity checks, a target flight path with the lowest energy consumption and full coverage is generated. The pre-processed soil nutrient data and pre-processed crop growth index data from the structured dataset are input into an improved weighted multiple linear regression model trained based on the target flight path. Using the dynamic spatiotemporal weights in the improved weighted multiple linear regression model, the fertilization and pesticide application rates for each sub-region in the mountainous rapeseed planting area are calculated. Thus, the heuristic function layered weighting and node slope continuity check in the improved A* algorithm provided by this invention can generate flight paths highly adapted to mountainous terrain features. These paths combine low energy consumption and full coverage, effectively avoiding operational obstacles caused by complex terrain. In the operational parameter calculation stage, an improved weighted multiple linear regression model is constructed, utilizing dynamic spatiotemporal weight terms to optimize feature variables, and combining a residual correction mechanism to improve calculation accuracy, achieving precise quantification of fertilization and spraying parameters. Based on this, a regional weight association mechanism is established to establish a synergistic linkage between path planning and parameter calculation, significantly enhancing their adaptability and effectively improving the efficiency and accuracy of UAV operations in complex mountainous environments, reducing waste of resources such as fertilizers and pesticides, while simultaneously strengthening the dual adaptability to rapeseed growth dynamics and mountainous terrain features.

[0063] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0064] This invention provides another integrated method for precise variable fertilization and pesticide spraying using drones in mountainous rapeseed areas, the method comprising the following steps: Step 1: Collect topographic elevation data, rapeseed planting area boundary data, soil nutrient data (N / P / K), and crop growth indicators (plant height / leaf area). Assess and classify the crop growth indicator data for pest and disease risk levels to obtain sub-region division results. Clean the data, standardize the format, and supplement missing data. Link and integrate the preprocessed topographic elevation data, crop growth indicator data, soil nutrient data, and rapeseed planting area boundary data to form a structured dataset.

[0065] Step 2: Based on the improved A* algorithm, input the preprocessed terrain elevation data and rapeseed planting area boundary data, and generate the target flight path with the lowest energy consumption and full coverage through hierarchical weighting of heuristic functions and node slope continuity checks.

[0066] Step 3: Input the preprocessed soil nutrient data, crop growth index data, and sub-region division results into an improved weighted multiple linear regression model trained based on the target flight path. Utilize the dynamic spatiotemporal weights to calculate the precise fertilization rate (kg / m²) for each sub-region. 2 ) and spraying volume (mL / m 2 ).

[0067] Step 4: The drone flies along the target flight path, collects data in real time, and feeds it back to the improved weighted multiple linear regression model. The parameters are updated through dynamic weight terms to adjust the output of the fertilization or spraying equipment.

[0068] Step 5: Collect and compare data after the operation to evaluate the effect of fertilization and spraying. For areas with excessive error, increase their sample weight through residual correction mechanism. At the same time, optimize the terrain cost factor threshold of the A* algorithm and iteratively improve the model.

[0069] In this embodiment, the collected data is first cleaned. Outliers in the raw data, such as abnormally high or low soil nutrient values ​​due to sensor malfunctions or plant height data that is clearly inconsistent with reality, are identified and removed to ensure the data accurately reflects the actual situation of rapeseed planting in mountainous areas and to avoid interference from outliers in subsequent algorithm calculations. To ensure the accuracy and comprehensiveness of data cleaning, a threshold screening method is mainly used. Threshold screening sets reasonable thresholds based on business experience or statistical methods, identifying and removing outlier data that exceeds the threshold range, thus achieving precise data cleaning. By setting a reasonable data threshold range, data that significantly deviates from the normal range is automatically filtered out. For example, if rapeseed plant height data exceeds 2 meters, it is judged as an anomaly based on actual planting experience and is directly removed.

[0070] Secondly, format standardization is crucial. Since data from different sources may use different coordinate systems and units, all data needs to be uniformly converted to the same coordinate system and standard units to eliminate data integration barriers caused by format differences. Format standardization is achieved through a combination of coordinate transformation and unit conversion using Geographic Information System (GIS) software and data processing tools: for coordinate system transformation, the projection conversion function of GIS is used to uniformly convert rapeseed planting area boundary data and topographic elevation data to the local Gaussian coordinate system; for unit conversion, soil nutrient data is uniformly converted to mg / kg, and plant height data is uniformly converted to cm, thereby eliminating data integration barriers caused by format differences.

[0071] Finally, utilizing the linear relationship of elevation values ​​between adjacent known grid points in the linear interpolation method, the ratio of the node to be interpolated to the adjacent known elevation nodes in spatial location is calculated. Based on this ratio, the elevation is determined, and the standardized topographic elevation data is interpolated to obtain preprocessed topographic elevation data. According to the geometric extension direction of the known boundary nodes on both sides of the missing segment, the coordinates of the missing boundary nodes in the standardized rapeseed planting area boundary data are calculated and supplemented to obtain preprocessed rapeseed planting area boundary data. Using the inverse distance weighted interpolation method, based on the initial nutrient values ​​of known points, the reciprocal distance from each known point to the point to be interpolated is calculated and weighted to obtain the target nutrient value for the missing area. Based on the target nutrient value, the standardized soil nutrient data is interpolated to obtain preprocessed soil nutrient data. Using the Kriging interpolation method and regionalized variable theory, the spatial correlation of rapeseed growth in different plots is quantified by fitting the variogram of known monitoring points, estimating the target plant height and target leaf area for the missing area. Based on the target plant height and target leaf area, the standardized crop growth index data is interpolated to obtain preprocessed crop growth index data.

[0072] Finally, the cleaned, standardized, and completed data are integrated with crop and soil data. Through spatial coordinate matching, topographic features are mapped one-to-one with data such as soil nutrient content, rapeseed plant height, and leaf area in the region, ultimately forming a structured dataset. This provides a unified, standardized, and complete data input for improving the path planning of the A* algorithm and the variable decision-making of the weighted multiple linear regression model.

[0073] Step 2 uses the structured dataset output from Step 1 as a basis to complete flight path planning based on the improved A* algorithm. The input data specifically includes preprocessed terrain elevation data, rapeseed planting area boundary coordinates, and sub-region division results.

[0074] In this embodiment, the slope data provided in step 1 is converted into a terrain cost factor. The conversion process uses weight coefficients of 0.3 / 0.5 / 0.8 corresponding to slope levels (0-15° / 15-30° / >30°). The steeper the slope, the higher the cost in path calculation, causing the algorithm to prioritize gentler areas during planning. Simultaneously, based on the importance indicators of the planting area sub-regions, an operation coverage factor is set. The center coordinates of key areas (such as areas with high pest and disease incidence or abnormal growth) are assigned a gravity weight of 1.2, while ordinary areas are assigned 0.8, ensuring that the path prioritizes coverage of critical operation points. During the node expansion phase, the algorithm calls the slope continuity data from step 1 to check the slope difference between adjacent nodes. If the difference is greater than 10°, the node is skipped to avoid increased energy consumption or operational instability caused by frequent steep slope turns of the drone.

[0075] The core parameters of the improved A* algorithm include the weight coefficients of the terrain cost factor (0.3 / 0.5 / 0.8 for 0-15° / 15-30° / >30°), the gravity weight of the operation coverage factor (1.2 for key areas and 0.8 for ordinary areas), and the node slope difference threshold (10°). The initial values ​​of these parameters are set based on prior knowledge of rapeseed planting in mountainous areas: for example, according to experience in mountain agricultural machinery operations, the energy consumption of drones increases significantly when the slope is >30°, so the highest cost weight is assigned; the operation priority of key areas (such as areas with high incidence of pests and diseases) should be higher than that of ordinary areas, so a higher gravity weight is set.

[0076] Subsequently, the algorithm enters a testing-feedback-iteration adaptation process: using the preprocessed data output from step 1 (terrain elevation, rapeseed planting area boundaries, and sub-region division results) as input, an initial flight path is generated; through actual drone operations, path energy consumption data (such as power consumption per unit distance), coverage integrity (whether key areas are missed), and flight stability (frequency of turning on steep slopes) are collected; if energy consumption is too high (such as frequently crossing areas >30°), the cost factor weight of that slope level is increased; if key areas are not fully covered, their gravity weight is increased; if the drone still experiences turning instability due to a slope difference close to 10°, the threshold is lowered to 8°.

[0077] Ultimately, after multiple rounds of testing and parameter adjustments in mountainous operation scenarios, the algorithm can stably output the flight path with the lowest energy consumption and full coverage. Its parameter system forms a stable mapping with the structured data in step 1 (such as specific slope data corresponding to specific cost weights), ensuring that the path planning is deeply adapted to the mountainous terrain and operation requirements.

[0078] Step 3 also uses the structured dataset output from Step 1 as input, and employs an improved weighted multiple linear regression model to accurately calculate fertilizer and pesticide application rates. The aim is to make the variable parameters more closely match the growth needs and regional differences of rapeseed in mountainous areas. The input data specifically includes preprocessed soil nutrient data, crop growth index data, and corresponding time labels (collection date) and regional attribute labels (such as whether it is a high-incidence area for pests and diseases or an area with abnormal growth). The time labels are used to distinguish the timeliness of the data, while the regional attribute labels are associated with the regional division results based on crop growth indicators and pest and disease risk assessment in Step 1.

[0079] First, the time tags provided in step 1 are used to categorize soil nutrient data and crop growth index data according to their collection time. Data from the last 3 days is assigned a time weight of 1.0, data from 3-7 days is assigned 0.7 (with a decay coefficient of 0.3), and earlier data is assigned a weight of <0.5. This differentiated weighting across time dimensions ensures that the model prioritizes responding to the latest crop growth status. Simultaneously, regional attribute tags are used to assign 1.5 times the regional weight to samples marked as high-incidence areas for pests and diseases, 1.2 times the weight to areas with abnormal growth, and basic weights to ordinary areas, thus enhancing the parameter fitting accuracy in key areas. During parameter calculation, the model substitutes the weighted samples into the regression equation and calculates the fertilizer application rate (kg / m²) for each sub-region by minimizing the weighted loss function. 2 With spraying volume mL / m 2 This forms the variable control parameters.

[0080] Based on the flight path output in step 2 and the variable control parameters generated in step 3, dynamic adaptation and precise control of UAV operations are achieved. The UAV first invokes the flight path planned in step 2 (including specific turning point coordinates, flight altitude, and sub-area coverage order) to initiate flight, ensuring that the aircraft always covers all rapeseed planting areas along the preset trajectory, especially key operation areas enhanced by the algorithm in step 2. These areas have been given priority passage in the path, laying the spatial foundation for subsequent precise operations. Simultaneously, the initial output parameters of the UAV's fertilization / spraying equipment directly adopt the calculation results from step 3, i.e., the fertilizer application rate (kg / m²) corresponding to each sub-area. 2 With spraying volume mL / m 2 These parameters are generated by combining soil nutrients, crop growth indicators and dynamic spatiotemporal weights with an improved weighted multiple linear regression model to ensure that the initial operation is basically matched with the regional needs.

[0081] During flight, the drone collects two types of key data in real time using its onboard sensors: environmental and crop dynamics information, and operational status data, such as the instantaneous output of actual fertilization / pesticide application and equipment operational stability. This real-time data is wirelessly transmitted back to the improved weighted multiple linear regression model, triggering dynamic weight updates. Since the real-time data is collected within the last three days, its weight can be set to 1.0 according to the time weighting rule, significantly higher than the historical data used in step 3. Simultaneously, combined with the attribute labels of the current flight area, the model will adjust the spatiotemporal weights of that area in real time. For example, if real-time monitoring shows that the yellowing of rapeseed in a certain sub-region is more severe than during modeling in step 3, the model will automatically increase the weight of the abnormal growth area in that region and recalculate the fertilization amount, increasing nitrogen fertilizer output; if it finds that the actual density of pests and diseases in a certain area is higher than the predicted value, the weight of its "high-incidence area of ​​pests and diseases" will be increased, and the amount of pesticide spraying will be increased accordingly.

[0082] When the model outputs updated parameters, the equipment control system adjusts the output of the fertilization / pesticide application equipment based on the drone's real-time location before entering the corresponding sub-region, ensuring precise matching between parameter changes and spatial location. For example, if the drone enters a pest outbreak zone after a turning point in the path, the equipment will immediately increase the spraying flow rate upon arrival at that area, avoiding delays or overshoots. Throughout the process, path planning in step 2 provides a spatial framework for the operation, initial parameters in step 3 provide baseline values, and real-time feedback and dynamic adjustment in step 4 strengthen the synergy between path planning and variable control in the solution, making the operational effect more aligned with the real-time needs of rapeseed cultivation in mountainous areas.

[0083] Step 5 is feedback optimization, which involves evaluating the effectiveness and iterating the model based on the job execution results of Step 4 and the various data generated from Steps 1 to 4.

[0084] The comparative data collected after the flight operation mainly includes three categories: First, actual growth effect data, namely rapeseed growth indicators (plant height, leaf area increase), soil nutrient changes (N / P / K content), and pest and disease control rate within 1-2 weeks after the operation. These data need to be compared with the baseline values ​​(pre-operation state) in the preprocessing data of step 1 and the target effect values ​​predicted by the improved weighted multiple linear regression model in step 3. Second, path execution data, including the deviation between the actual flight trajectory of the UAV and the planned path in step 2, the actual energy consumption and coverage integrity in each slope area, used to verify the path adaptability of the improved A-star algorithm. Third, parameter execution data, namely the difference between the actual fertilizer / pesticide application amount of the UAV and the output parameters of steps 3 and 4, to clarify the error range of variable control.

[0085] During the evaluation process, if the error between the actual effect and the predicted value of a certain area is greater than the preset error value, the residual correction mechanism of the improved weighted multiple linear regression model will be triggered. That is, based on the regional attribute label marked in step 1, the weight of the area will be increased in the model's sample set. At the same time, the dynamic adjustment record of the real-time feedback in step 4 will be tracked. If the error is due to the untimely update of the weights, the time decay coefficient will be optimized to enhance the model's fitting accuracy to the key areas.

[0086] For path planning optimization, combining the actual energy consumption data collected in step 4 with the terrain cost factor settings in step 2, if the actual energy consumption of a certain slope level is found to be significantly higher than the planning expectation, the weight of the terrain cost factor for that level will be increased; if the drone experiences a significant turn and a surge in energy consumption at a slope difference, the threshold for adjusting the slope difference will be decreased to avoid unreasonable paths caused by a loose threshold setting.

[0087] Finally, the evaluated and corrected model parameters will be fed back to the dataset update in step 1, the path planning in step 2, and the parameter calculation in step 3, so that the entire closed loop continuously adapts to the dynamic needs of rapeseed planting in mountainous areas, ensuring that algorithm improvements are always based on actual operational data, and improving the long-term stability and accuracy of the solution.

[0088] Figure 3 This is a schematic diagram of the composition of the integrated collaborative device for precision variable fertilization and pesticide spraying of rapeseed using drones in mountainous areas, provided in an embodiment of the present invention. Figure 3 As shown, the integrated collaborative device 300 for precision variable fertilization and spraying of rapeseed using drones in mountainous areas includes: an acquisition module 301, used to acquire multi-source heterogeneous data of rapeseed planting areas in mountainous areas; the multi-source heterogeneous data includes topographic elevation data, rapeseed planting area boundary data, soil nutrient data, and crop growth index data; the crop growth index data is used to assess and classify the risk level of pests and diseases, and to obtain the sub-region division results; a formation module 302, used to clean and standardize the format of the multi-source heterogeneous data, and to fill in missing data; the preprocessed multi-source heterogeneous data is associated with a unified geographic or sub-region unit to form a structured dataset; and a generation module 303, used to... The preprocessed terrain elevation data, preprocessed rapeseed planting area boundary data, and the sub-region division results in the structured dataset are input into the improved A* algorithm. Through the heuristic function of the improved A* algorithm, hierarchical weighting and node slope continuity checks, a target flight path with the lowest energy consumption and full coverage is generated. The calculation module 304 is used to input the preprocessed soil nutrient data and preprocessed crop growth index data in the structured dataset into an improved weighted multiple linear regression model trained based on the target flight path. Using the dynamic spatiotemporal weight term in the improved weighted multiple linear regression model, the fertilizer application and pesticide application rates of each sub-region in the mountain rapeseed planting area are calculated.

[0089] It should be noted that the description of the apparatus in this embodiment is similar to the description of the method embodiment described above, and has similar beneficial effects to the same method embodiment; therefore, it will not be repeated. For technical details not disclosed in this apparatus embodiment, please refer to the description of the method embodiment of this invention for understanding.

[0090] It should be noted that, in the embodiments of the present invention, if the above-mentioned integrated method for precise variable fertilization and spraying of rapeseed using drones in mountainous areas is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, or the part that contributes to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a terminal to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of the present invention are not limited to any specific hardware and software combination.

[0091] Correspondingly, embodiments of the present invention provide an electronic device, Figure 4 This is a schematic diagram of the composition structure of the electronic device provided in the embodiments of the present invention, such as... Figure 4 As shown, the electronic device 400 includes at least a processor 401 and a computer-readable storage medium 402 configured to store executable instructions, wherein the processor 401 generally controls the overall operation of the electronic device 400. The computer-readable storage medium 402 is configured to store instructions and applications executable by the processor 401, and may also cache data to be processed or processed by various modules in the processor 401 and the electronic device 400, and may be implemented using flash memory or random access memory (RAM).

[0092] This invention provides a storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to perform the method provided in this invention, for example... Figure 2 The method shown.

[0093] In some embodiments, the storage medium may be a computer-readable storage medium, such as a ferromagnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disc, or a compact disk-read-only memory (CD-ROM); or it may be a device that includes one or any combination of the above-mentioned memories.

[0094] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0095] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file containing other programs or data, for example, in one or more scripts within a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., files storing one or more modules, subroutines, or code sections). As an example, executable instructions may be deployed to execute on a single electronic device, or on multiple electronic devices located in one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0096] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of the present invention are included within the scope of protection of the present invention.

[0097] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of the invention, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the invention. The sequence numbers of the above-described embodiments of the invention are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0098] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components may be combined, or integrated into another system, or some features may be ignored or not performed.

[0099] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for integrated and coordinated precision variable fertilization and pesticide spraying using unmanned aerial vehicles (UAVs) in mountainous rapeseed areas, characterized in that, The method includes: Acquire multi-source heterogeneous data of rapeseed planting areas in mountainous areas; the multi-source heterogeneous data includes topographic elevation data, rapeseed planting area boundary data, soil nutrient data, and crop growth index data; assess and classify the crop growth index data for pest and disease risk levels to obtain sub-region division results; The multi-source heterogeneous data is cleaned, standardized in format, and missing data is filled in; the preprocessed multi-source heterogeneous data is associated with a unified geographic or sub-regional unit to form a structured dataset. The preprocessed terrain elevation data, preprocessed rapeseed planting area boundary data, and sub-region division results in the structured dataset are input into the improved A-Star algorithm. Through the heuristic function of the improved A-Star algorithm, hierarchical weighting and node slope continuity check, the target flight path with the lowest energy consumption and full coverage is generated. The preprocessed soil nutrient data and preprocessed crop growth index data in the structured dataset are input into an improved weighted multiple linear regression model trained based on the target flight path. The dynamic spatiotemporal weights in the improved weighted multiple linear regression model are used to calculate the amount of fertilizer and pesticides applied to each sub-region in the mountain rapeseed planting area.

2. The method according to claim 1, characterized in that, After calculating the fertilization and pesticide application rates for each sub-region within the mountain rapeseed planting area, the method further includes: Within a preset time window after fertilization and spraying are completed, crop height growth, leaf area index changes and pest and disease reduction rates in each sub-region are collected as comparative data on fertilization and spraying effects. Based on the comparison data of fertilization and pesticide application effects, the crop growth response of each sub-region is evaluated to see if the expected target is achieved, and the deviation between the actual growth index and the target growth index is calculated and used as the target error of fertilization and pesticide application effects. For sub-regions where the target error exceeds a preset error threshold, the sample weights of historical sample data in the corresponding sub-regions are corrected through a residual correction mechanism. Combining the path energy consumption data, operation coverage integrity data, and sample weights of each sub-region during the UAV operation, the terrain cost factor threshold in the improved A-Star algorithm is dynamically adjusted and improved through parameter optimization methods.

3. The method according to claim 2, characterized in that, The method further includes: Based on the target flight path, the UAV equipped with sensors collects soil nutrient replenishment data and crop growth status update data of the mountain rapeseed planting area. The newly collected data is then input into the improved weighted multiple linear regression model, triggering the dynamic spatiotemporal weight term in the model to recalculate the sample weights, and the model parameters are updated by weighted least squares.

4. The method according to claim 1, characterized in that, The topographic elevation data, the rapeseed planting area boundary data, the soil nutrient data, and the crop growth index data are obtained through the following methods: The terrain elevation data is obtained by constructing a digital elevation model using a drone equipped with lidar or by combining multi-view photogrammetry with RTK positioning. Using satellite remote sensing imagery or UAV visible light photography cameras, combined with manual interpretation and automatic vectorization technology of geographic information systems, the boundaries of the rapeseed planting area in the mountains are delineated, the spatial range of UAV operations is determined, and the boundaries are marked using GPS positioning equipment to obtain the boundary data of the rapeseed planting area. Deep soil samples were collected from the mountain rapeseed planting area using a ground-based grid sampling strategy. Nutrient baseline values ​​were obtained through laboratory chemical analysis. The current content of key elements for crop growth was quickly determined using a near-infrared spectrometer combined with a pre-calibrated prediction model. Combined with dynamic monitoring and periodic updates, initial soil samples were collected before planting and the initial content was determined to form soil nutrient data including current and initial contents. Based on hyperspectral sensors combined with ground-based fixed-point strategies, vegetation reflectance spectral data, plant height, and leaf area index are acquired to construct the crop growth index data.

5. The method according to claim 1, characterized in that, The process of generating a target flight path with the lowest energy consumption and full coverage through hierarchical weighting and node slope continuity checks using the heuristic function of the improved A* algorithm includes: Based on the improved A-star algorithm, the preprocessed terrain elevation data is transformed into a terrain cost factor, and the preprocessed rapeseed planting area boundary data and the sub-region division results are set as operation coverage factors to construct a cost function for path search. The path search direction is guided by the hierarchical weighting of the cost function, and the slope continuity data in the preprocessed terrain elevation data is called to check the slope difference between adjacent nodes, generating a target flight path that includes the coordinates of turning points and the relative height of the crop canopy. The operation coverage factor is used to quantify the cost parameter of regional coverage priority in path search; the cost function is determined based on the base distance cost, terrain cost factor and operation coverage factor.

6. The method according to claim 1, characterized in that, The process involves inputting preprocessed soil nutrient data and preprocessed crop growth index data from the structured dataset into an improved weighted multiple linear regression model trained based on the target flight path. Using the dynamic spatiotemporal weights in the improved weighted multiple linear regression model, the fertilizer and pesticide application rates for each sub-region within the mountain rapeseed planting area are calculated, including: The input data includes the pre-processed soil nutrient data and the pre-processed crop growth index data, initial time labels and initial regional attribute labels; the initial time labels are used to distinguish the timeliness of the data, and the initial regional attribute labels are used to mark whether the corresponding data area is a high-incidence area of ​​pests and diseases or an area with abnormal growth. Based on the initial time label, the preprocessed soil nutrient data and the preprocessed crop growth index data are classified according to the collection time and assigned time weights; based on the initial regional attribute label, the classified soil nutrient data and crop growth index data are assigned regional weights; the input data are weighted based on the time weights and the regional weights to obtain weighted data; The weighted data is substituted into the improved weighted multiple linear regression model, and the fertilizer application rate and the pesticide application rate are obtained by minimizing the loss function that includes the spatiotemporal weight term.

7. The method according to claim 1, characterized in that, The improved weighted multiple linear regression model is obtained through the following steps: Based on the collected basic data of the mountain rapeseed planting area, quantitative calculation rules for time weight and regional weight are established and weight range constraints are set to form a dynamic spatiotemporal weight term for the improved weighted multiple linear regression model; the basic data includes sample soil nutrient data, sample crop growth index data, sample time labels and sample regional attribute labels. The basic data is used as the training sample set, and a weighted loss function is constructed based on the dynamic spatiotemporal weight term, wherein the weighted loss function is the weighted sum of squares of the prediction errors of each sample; By minimizing the weighted loss function, the model regression parameters are solved using the weighted least squares method to generate an initial model, wherein the output of the initial model is the recommended fertilization amount and spraying amount for each sub-region. Based on the real-time crop growth dynamic data and equipment operation status data collected by the UAV during the target flight path execution, the time weight is updated according to the quantification calculation rules, and the region weight is dynamically adjusted in combination with the degree of deviation between the sample region attribute label and the real-time detection value. Based on the updated dynamic spatiotemporal weights, the parameters of the initial model are iteratively updated using the newly added sample data to obtain the updated regression parameters, and the corresponding updated fertilization amount and spraying amount are output. The adaptability of the updated model parameters is verified based on the degree of matching between the actual growth response in the field and the preset target. Repeat the weight update step and continue to update the model until the model output meets the preset convergence condition, thus obtaining the trained improved weighted multiple linear regression model.

Images