Variable Fertilization Decision System and Method Based on Multispectral and Soil Sensors
By fusing data from multispectral remote sensing and soil sensor networks, combined with efficient models and precision fertilization equipment, the problems of single data and insufficient stress diagnosis in existing variable fertilization technologies have been solved, enabling precise multi-source data processing and fertilization, and improving agricultural production efficiency.
Patent Information
- Application Number
- CN202511742866.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-25
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2045-11-25
AI Technical Summary
Existing variable fertilization technologies rely on a single data source, have simple decision-making models, cannot fully reflect the true state of the crop-soil system, and lack stress diagnosis capabilities, resulting in inaccurate fertilization decisions.
Multi-spectral remote sensing units and soil sensor networks are used to acquire multi-source data. By combining modified support vector machines and whale optimization algorithms, a support vector machine model based on RBF-Poly hybrid kernel functions is constructed to dynamically diagnose the causes of crop nutrient stress. Precision fertilization is achieved through a ternary precision variable fertilizer applicator.
It achieves multi-source data fusion, improves the accuracy and real-time nature of fertilization decisions, can dynamically diagnose the true causes of crop nutrient stress, and enables precise application of nitrogen, phosphorus, and potassium fertilizers, reducing fertilizer usage and increasing crop yield.
Smart Images

Figure CN121694099B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of agricultural fertilization technology, specifically to a variable fertilization decision system and method based on multispectral and soil sensors. Background Technology
[0002] Traditional agricultural fertilization methods typically employ uniform fertilization, neglecting the spatial heterogeneity of soil nutrients and crop growth within farmland. This results in low fertilizer utilization efficiency and high agricultural production costs. Currently, variable fertilization technology, as one of the core components of precision agriculture, has been applied in some regions.
[0003] Existing variable fertilization techniques are mainly based on the following three methods:
[0004] One method is variable fertilization based on soil nutrients: This method obtains soil nutrient data through soil sampling analysis or soil sensors, and determines the amount of fertilizer to be applied based on the soil nutrient status; however, this method only focuses on the soil itself and cannot directly reflect the actual nutritional status of crops, ignoring the efficiency of crop absorption and utilization of nutrients.
[0005] Second, variable fertilization based on remote sensing monitoring: This method uses drones or satellite remote sensing to obtain spectral information of crop canopy, judges crop growth based on vegetation index, and then determines the amount of fertilizer. However, the limitation of this method is that it cannot distinguish whether the difference in crop growth is caused by soil nutrient deficiency or other environmental factors, such as water stress, pests and diseases, which may lead to misjudgment.
[0006] Thirdly, variable fertilization based on yield targets: However, this method relies on historical data, lacks real-time capability, and is difficult to adapt to environmental changes in different growing seasons.
[0007] The main problems with existing technologies:
[0008] 1. Single data source: Existing systems often use only a single data source, such as soil or remote sensing, which cannot fully reflect the true condition of the crop-soil system.
[0009] 2. Simple decision-making models: Most decision-making models are based on simple linear regression or empirical formulas, which are difficult to handle multi-source, non-linear, and high-dimensional agricultural data.
[0010] 3. Insufficient stress diagnosis capability: The inability to accurately diagnose the true cause of crop nutrient stress leads to inaccurate fertilization decisions.
[0011] Therefore, those skilled in the art have provided a variable fertilization decision system and method based on multispectral and soil sensors to solve the problems mentioned in the background art. Summary of the Invention
[0012] To address the aforementioned technical problems, this invention provides a variable fertilization decision-making system and method based on multispectral and soil sensors.
[0013] A variable fertilization decision system based on multispectral and soil sensors includes:
[0014] Multispectral remote sensing unit, soil sensor network, data processing center, and variable fertilization execution agency;
[0015] The multispectral remote sensing unit is used to acquire multispectral image data of crop canopy;
[0016] The soil sensor network is used to monitor soil nutrient data in real time;
[0017] The data processing center is used to integrate multi-source data and build coupled decision models;
[0018] The variable fertilization execution mechanism is used to perform variable fertilization operations.
[0019] Preferably, the multispectral remote sensing unit includes a multispectral camera mounted on a UAV, which includes at least a 450nm blue light band, a 560nm green light band, a 660nm red light band, a 720nm / 750nm red edge band, and an 800nm near-infrared band, capable of acquiring the reflectance spectral data of crop canopy and calculating multiple vegetation indices such as NDVI, RENDVI, and mND705.
[0020] Preferably, the soil sensor network uses a visible-near-infrared spectral sensor with a spectral range of 350-2500nm. Based on a method combining a modified support vector machine and a whale optimization algorithm, it detects the total nitrogen, total phosphorus, and total potassium content in the soil in real time and forms a spatial distribution map of soil nutrients.
[0021] Preferably, the data processing center includes a data preprocessing module, a coupled decision model, and a prescription map generation module; the coupled decision model is a soil fertility supply capacity-crop fertility demand coupled decision model, which adopts a support vector machine model based on RBF-Poly hybrid kernel function and uses the whale optimization algorithm to optimize the model parameters.
[0022] Preferably, the coupled decision model can dynamically diagnose the true cause of crop nutrient stress and distinguish between stress caused by soil nutrient deficiency or other environmental factors, specifically achieved through the following methods:
[0023] Establish a correlation analysis model between multispectral vegetation index and soil nutrient content;
[0024] Construct time-series-based trajectories of crop growth and soil nutrient changes;
[0025] The decision tree algorithm is used to classify and identify the types of nutrient stress.
[0026] Preferably, the coupled decision model adopts a support vector machine model based on the RBF-Poly hybrid kernel function, and the expression of the hybrid kernel function is:
[0027]
[0028] in,
[0029]
[0030]
[0031] In the formula, For RBF kernel function, For polynomial kernel functions, It is a mixed kernel function; It is the mixing coefficient. , Degenerates into a pure RBF nucleus. It degenerates into a pure polynomial kernel; These are the parameters of the kernel function; For kernel width parameters; , This is the offset. Let be the order of the polynomial.
[0032] Preferably, the data processing center uses the whale optimization algorithm to optimize the coupled decision model. , , , , The optimization algorithm simulates the bubble-net hunting behavior of humpback whales, including three stages: surrounding prey, bubble-net attack, and searching for prey.
[0033] Preferably, the variable fertilization actuator is a ternary precision variable fertilization machine, which can independently and dynamically control the application amount and ratio of nitrogen, phosphorus and potassium fertilizers, with a fertilization accuracy error of no more than ±5%, and can adjust the fertilization strategy in real time according to the prescription diagram.
[0034] A variable fertilization decision-making method based on multispectral and soil sensors includes the following steps:
[0035] Step 1: Acquire multispectral image data of crop canopy using multispectral remote sensing units;
[0036] Step 2: Monitor soil nutrient data in real time using a soil sensor network;
[0037] Step 3: Preprocess and fuse the multi-source data at the data processing center;
[0038] Step 4: Construct a coupled decision-making model of soil nutrient supply capacity and crop nutrient demand to dynamically diagnose the causes of crop nutrient stress;
[0039] Step 5: Generate a variable fertilization prescription diagram;
[0040] Step 6: Implement precise variable fertilization operations through a variable fertilization implementation agency.
[0041] Preferably, the dynamic diagnosis of the causes of crop nutrient stress in step four includes:
[0042] Identify areas of abnormal crop growth by identifying areas of abnormal crop growth using multispectral vegetation index anomalies;
[0043] Analyze the soil nutrient status corresponding to the abnormal areas;
[0044] Based on historical data and environmental factors, it was determined that nutrient stress was caused by soil nutrient deficiency, water stress, pests and diseases, or other environmental factors.
[0045] Calculate precise fertilization amounts to address stress caused by soil nutrient deficiency;
[0046] For stresses caused by other factors, corresponding agronomic management recommendations are generated.
[0047] The technical effects and advantages of this invention are as follows:
[0048] Multi-source data fusion: By integrating data from UAV multispectral remote sensing and soil sensor networks, crop canopy information and root zone soil nutrient status can be obtained simultaneously, providing more comprehensive data support for fertilization decisions.
[0049] High-precision decision model: It adopts a support vector machine model based on RBF-Poly hybrid kernel function and combines it with whale optimization algorithm for parameter optimization, which can effectively process multi-source, nonlinear and high-dimensional agricultural data and improve the accuracy of nutrient demand prediction.
[0050] Intelligent stress diagnosis: Using a coupled decision model of soil nutrient supply capacity and crop nutrient demand, it can dynamically diagnose the true cause of crop nutrient stress, accurately distinguish whether it is caused by soil nutrient deficiency or other environmental factors, and avoid misjudgment and incorrect fertilization.
[0051] Precise and efficient fertilization: The ternary precision variable fertilizer applicator can independently and dynamically control the application amount and ratio of nitrogen, phosphorus and potassium fertilizers to achieve truly precise fertilization.
[0052] In addition, the present invention can dynamically adjust the fertilization strategy according to the needs of crops at different growth stages, so as to achieve precise nutrient management throughout the entire growth period. The system is applicable to a variety of agricultural production scenarios such as field crops, orchards, and vegetable bases, and has wide applicability. Attached Figure Description
[0053] Figure 1 This is a schematic diagram of the structure of the variable fertilization decision system and method based on multispectral and soil sensors provided in the embodiments of this application;
[0054] Figure 2 This is a flowchart illustrating the workflow of the variable fertilization decision system and method based on multispectral and soil sensors provided in the embodiments of this application. Detailed Implementation
[0055] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and design various embodiments with various modifications suitable for a particular purpose.
[0056] Example 1
[0057] Please see Figures 1-2 This embodiment provides a variable fertilization decision-making system and method based on multispectral and soil sensors, including...
[0058] A variable fertilization decision system based on multispectral and soil sensors includes:
[0059] Multispectral remote sensing unit, soil sensor network, data processing center, and variable fertilization execution agency;
[0060] Multispectral remote sensing units are used to acquire multispectral image data of crop canopies;
[0061] Soil sensor networks are used to monitor soil nutrient data in real time;
[0062] The data processing center is used to integrate multi-source data and build coupled decision models;
[0063] Variable fertilization execution agencies are used to carry out variable fertilization operations.
[0064] The multispectral remote sensing unit includes a multispectral camera mounted on an unmanned aerial vehicle (UAV), which includes at least the 450nm blue light band, 560nm green light band, 660nm red light band, 720nm / 750nm red edge band and 800nm near-infrared band. It can acquire the reflectance spectral data of crop canopy and calculate multiple vegetation indices such as NDVI, RENDVI, and mND705.
[0065] The soil sensor network uses a visible-near-infrared spectral sensor with a spectral range of 350-2500nm. Based on a method combining a modified support vector machine and a whale optimization algorithm, it can detect the total nitrogen, total phosphorus and total potassium content in the soil in real time and form a spatial distribution map of soil nutrients.
[0066] The data processing center includes a data preprocessing module, a coupled decision model, and a prescription map generation module. The coupled decision model is a soil fertility supply capacity-crop fertility demand coupled decision model, which adopts a support vector machine model based on RBF-Poly hybrid kernel function and uses the whale optimization algorithm to optimize the model parameters.
[0067] Coupled decision models can dynamically diagnose the true causes of crop nutrient stress and distinguish between stress caused by soil nutrient deficiency or other environmental factors. This is achieved through the following methods:
[0068] Establish a correlation analysis model between multispectral vegetation index and soil nutrient content;
[0069] Construct time-series-based trajectories of crop growth and soil nutrient changes;
[0070] The decision tree algorithm is used to classify and identify the types of nutrient stress.
[0071] The coupled decision model adopts a support vector machine model based on the RBF-Poly hybrid kernel function. The expression of the hybrid kernel function is as follows:
[0072]
[0073] in,
[0074]
[0075]
[0076] In the formula, For RBF kernel function, For polynomial kernel functions, It is a mixed kernel function; It is the mixing coefficient. , Degenerates into a pure RBF nucleus. It degenerates into a pure polynomial kernel; These are the parameters of the kernel function; For kernel width parameters; , This is the offset. Let be the order of the polynomial.
[0077] The data processing center uses the whale optimization algorithm to optimize the coupled decision model. , , , , The optimization algorithm simulates the bubble-net hunting behavior of humpback whales, including three stages: surrounding prey, bubble-net attack, and searching for prey.
[0078] The variable fertilization actuator is a ternary precision variable fertilization machine, which can independently and dynamically control the application rate and ratio of nitrogen, phosphorus and potassium fertilizers. The fertilization accuracy error does not exceed ±5%, and the fertilization strategy can be adjusted in real time according to the prescription diagram.
[0079] A variable fertilization decision-making method based on multispectral and soil sensors includes the following steps:
[0080] Step 1: Acquire multispectral image data of crop canopy using multispectral remote sensing units;
[0081] Step 2: Monitor soil nutrient data in real time using a soil sensor network;
[0082] Step 3: Preprocess and fuse the multi-source data at the data processing center;
[0083] Step 4: Construct a coupled decision-making model of soil nutrient supply capacity and crop nutrient demand to dynamically diagnose the causes of crop nutrient stress;
[0084] Step 5: Generate a variable fertilization prescription diagram;
[0085] Step 6: Implement precise variable fertilization operations through a variable fertilization implementation agency.
[0086] Step four involves dynamically diagnosing the causes of crop nutrient stress, including:
[0087] Identify areas of abnormal crop growth by identifying areas of abnormal crop growth using multispectral vegetation index anomalies;
[0088] Analyze the soil nutrient status corresponding to the abnormal areas;
[0089] Based on historical data and environmental factors, it was determined that nutrient stress was caused by soil nutrient deficiency, water stress, pests and diseases, or other environmental factors.
[0090] Calculate precise fertilization amounts to address stress caused by soil nutrient deficiency;
[0091] For stresses caused by other factors, corresponding agronomic management recommendations are generated.
[0092] Specific examples of the present invention are as follows:
[0093] System Composition and Working Principle
[0094] 1. Multispectral remote sensing unit
[0095] The multispectral remote sensing unit uses either the DJI Phantom 4 Multispectral Edition drone or the OPTINI NY3300PRO multispectral agricultural detector, equipped with a multispectral camera covering six core bands: 450nm blue light, 560nm green light, 660nm red light, 720nm / 750nm red edge, and 800nm near-infrared. This device supports 32x optical zoom, can cover farmland within a 1-kilometer radius, and generates crop health index maps in real time.
[0096] The drone flies at an altitude of 50-100 meters, achieving a spatial resolution of 5-10 centimeters, enabling it to clearly identify subtle changes in the crop canopy. Flights should be taken during clear weather with low winds, ideally between 10:00 AM and 2:00 PM, to minimize the impact of solar altitude and atmospheric conditions on image quality.
[0097] 2. Soil sensor network
[0098] The soil sensor network is deployed in key locations in the field to form a wireless sensor network. The sensors use visible-near infrared spectroscopy (Vis-NIR, 350-2500nm) and are based on modified support vector machine and whale optimization algorithm (WOA-SVM) to achieve rapid and non-destructive detection of total nitrogen, total phosphorus and total potassium content in the soil.
[0099] The spacing between sensor nodes is determined based on the degree of field variation, generally 30-50 meters. In areas with high variation, the spacing can be appropriately increased. The sensor network uses 4G / 5G wireless transmission technology to upload real-time monitoring data to the cloud platform, supporting remote monitoring and data management.
[0100] 3. Data Processing Center
[0101] The data processing center is the intelligent core of the system, and includes the following three modules:
[0102] Data preprocessing module: Performs radiometric, atmospheric, and geometric corrections on multispectral images to eliminate the influence of the sensor itself, atmospheric conditions, and topographic relief on image quality. Then, it calculates various vegetation indices, including NDVI, RENDVI, and mND705. At the same time, it performs quality control and outlier processing on soil sensor data.
[0103] Coupled decision model: Adopting the coupling concept of soil nutrient supply capacity and crop nutrient demand, a support vector machine model based on RBF-Poly hybrid kernel function is constructed. This model can integrate multispectral features and soil nutrient data to predict crop nutrient demand and diagnose the causes of nutrient stress.
[0104] Prescription map generation module: Based on the output of the coupled decision model, combined with farmland boundary information and agricultural machinery operation path, it generates a spatial distribution map containing the application rates and ratios of nitrogen, phosphorus and potassium. The prescription map can be in vector or raster format, and the spatial resolution is determined according to the operation accuracy of the fertilizer applicator, which is generally 5-10 meters.
[0105] 4. Variable-rate fertilization implementation agencies
[0106] The variable fertilizer application mechanism adopts a ternary precision variable fertilizer applicator, which can independently and dynamically adjust the application rate and ratio of nitrogen, phosphorus and potassium fertilizers. The fertilizer applicator consists of a fertilizer tank, a fertilizer discharge device, a control system and a hydraulic system. The fertilizer discharge device adopts electromechanical-hydraulic integration technology, and controls the speed of the fertilizer discharge shaft through a servo motor to achieve precise control of the fertilizer application rate.
[0107] Construction of Coupled Decision Model
[0108] The coupled decision model is the core innovation of this invention. It adopts a support vector machine model based on RBF-Poly hybrid kernel function and uses the Whale Optimization Algorithm (WOA) to optimize the model parameters.
[0109] 1. Data Preprocessing and Feature Extraction
[0110] After preprocessing the multispectral images, various vegetation indices were calculated:
[0111] NDVI = (NIR - R) / (NIR + R)
[0112] RENDVI = (NIR - RE) / (NIR + RE)
[0113] mND705 = (750 - 705) / (750 + 705 - 2×445)
[0114] Wherein, NIR is the near-infrared reflectance, R is the red light reflectance, and RE is the red edge reflectance.
[0115] Simultaneously, texture features are extracted from multispectral images, including texture features based on gray-level co-occurrence matrix (GLCM) and Gabor texture features.
[0116] 2. Hybrid Kernel Function Support Vector Machine Model
[0117] A support vector machine model based on the RBF-Poly hybrid kernel function is adopted. The expression of the hybrid kernel function is as follows:
[0118]
[0119] in,
[0120]
[0121]
[0122] In the formula, For RBF kernel function, For polynomial kernel functions, It is a mixed kernel function; It is the mixing coefficient. , Degenerates into a pure RBF nucleus. It degenerates into a pure polynomial kernel; These are the parameters of the kernel function; For kernel width parameters; , This is the offset. Let be the order of the polynomial.
[0123] 3. Parameter optimization of the whale optimization algorithm
[0124] The Whale Optimization Algorithm (WOA) was used to optimize the model. , , , , The optimized whale optimization algorithm simulates the bubble-web hunting behavior of humpback whales and features fast convergence speed and strong global search capability.
[0125] 4. Diagnostic mechanisms of nutrient stress
[0126] Coupled decision models can dynamically diagnose the true causes of crop nutrient stress, specifically through the following steps:
[0127] Identify areas of abnormal crop growth by identifying areas of abnormal crop growth using multispectral vegetation index anomalies;
[0128] Analyze the soil nutrient status corresponding to the abnormal areas;
[0129] A decision tree classification model is constructed by combining historical data and environmental factors;
[0130] Determine whether nutrient stress is caused by soil nutrient deficiency, water stress, pests and diseases, or other environmental factors;
[0131] For stress caused by soil nutrient deficiency, calculate precise fertilization amounts; for stress caused by other factors, generate corresponding agronomic management recommendations.
[0132] The following uses variable fertilization in soybean fields in Northeast China as an example to illustrate the specific implementation of this invention:
[0133] 1. Data Collection
[0134] During the three key growth stages of soybeans—branching, flowering and pod setting, and grain filling—aerial photography was conducted using a DJI Phantom 4 Multispectral Edition drone at an altitude of 80 meters to acquire multispectral images with a spatial resolution of 5 centimeters. Simultaneously, 30 soil sensor nodes deployed in the field were used to monitor the nitrogen, phosphorus, and potassium content of the soil in real time.
[0135] 2. Data Processing and Feature Extraction
[0136] Multispectral images were preprocessed to calculate vegetation indices such as NDVI, RENDVI, and mND705, and GLCM and Gabor texture features were extracted. At the same time, nitrogen, phosphorus, and potassium content data monitored by soil sensors were obtained.
[0137] 3. Construction and Validation of Coupled Decision Model
[0138] Historical yield data, soil nutrient data, and crop canopy spectral data were collected and divided into training and validation sets. A coupled decision model was constructed using the WOA-SVM algorithm, and the model performance was evaluated using 10-fold cross-validation. The results showed that the model's prediction accuracy (R²) for soybean nitrogen, phosphorus, and potassium requirements reached 0.89, 0.92, and 0.87, respectively.
[0139] 4. Diagnosis of nutrient stress
[0140] Diagnosing nutrient stress in soybean fields using a coupled decision model:
[0141] Three NDVI anomalous regions were identified;
[0142] Analysis revealed that soil nitrogen content was significantly low in two of the regions, and soil potassium content was low in one region but accompanied by water stress characteristics.
[0143] Precise fertilization recommendations are generated for areas with soil nutrient deficiencies, and irrigation recommendations are generated for areas with water stress.
[0144] 5. Prescription diagram generation and variable fertilization
[0145] Based on model predictions and stress diagnosis, a variable fertilizer prescription map for soybean fields was generated. The prescription map divided the farmland into three fertility levels: high, medium, and low, and set different fertilizer application rates and ratios for each level.
[0146] Fertilization plans for zones with different fertility levels:
[0147] It is divided into fertility level, nitrogen fertilizer application rate (kg / ha), phosphorus fertilizer application rate (kg / ha), potassium fertilizer application rate (kg / ha), and fertilizer ratio (N:P2O5:K2O);
[0148] In high-fertility areas, the application rate of nitrogen fertilizer is 45 kg / ha, the application rate of phosphorus fertilizer is 60 kg / ha, the application rate of potassium fertilizer is 75 kg / ha, and the fertilizer ratio is 1:1.33:1.67 (N:P2O5:K2O).
[0149] In medium fertility areas, the application rate of nitrogen fertilizer is 60 kg / ha, the application rate of phosphorus fertilizer is 75 kg / ha, the application rate of potassium fertilizer is 90 kg / ha, and the fertilizer ratio is 1:1.25:1.5 (N:P2O5:K2O).
[0150] In low fertility areas, the application rate of nitrogen fertilizer is 75 kg / ha, the application rate of phosphorus fertilizer is 90 kg / ha, the application rate of potassium fertilizer is 105 kg / ha, and the fertilizer ratio is 1:1.2:1.4 (N:P2O5:K2O).
[0151] Using a ternary precision variable fertilizer applicator, variable fertilization is performed according to the generated prescription diagram. Before operation, the fertilizer applicator is calibrated to ensure that the fertilization accuracy error of each nutrient unit does not exceed ±5%.
[0152] 6. Effectiveness Evaluation
[0153] Compared with traditional uniform fertilization, the variable fertilization method implemented in this invention reduces fertilizer usage by 25.6%, increases soybean yield by 8.3%, and increases fertilizer partial productivity by 45.5%, achieving the goal of cost reduction and efficiency improvement.
[0154] The detailed description of known functions and components is omitted in this disclosure. To ensure the compatibility of the equipment, the operating methods used are consistent with the parameters of commercially available instruments. For example, the ternary precision variable fertilizer applicator can be selected according to existing technology and the usage needs of those skilled in the art.
[0155] Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art and related fields based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described and explained in the present invention, unless otherwise specified or limited, shall be implemented according to conventional means in the art.
Claims
1. A variable fertilization decision system based on multispectral and soil sensors, characterized in that, include: Multispectral remote sensing unit, soil sensor network, data processing center, and variable fertilization execution agency; The multispectral remote sensing unit is used to acquire multispectral image data of crop canopy; The soil sensor network is used to monitor soil nutrient data in real time; The data processing center is used to integrate multi-source data and build coupled decision models; The variable fertilization execution mechanism is used to perform variable fertilization operations; The coupled decision model can dynamically diagnose the true causes of crop nutrient stress and distinguish between stress caused by soil nutrient deficiency or other environmental factors. This is achieved through the following methods: Establish a correlation analysis model between multispectral vegetation index and soil nutrient content; Construct time-series-based trajectories of crop growth and soil nutrient changes; The decision tree algorithm is used to classify and identify the types of nutrient stress.
2. The variable fertilization decision system based on multispectral and soil sensors according to claim 1, characterized in that, The multispectral remote sensing unit includes a multispectral camera mounted on a UAV, which includes at least the 450nm blue light band, 560nm green light band, 660nm red light band, 720nm / 750nm red edge band and 800nm near-infrared band. It can acquire the reflectance spectral data of crop canopy and calculate multiple vegetation indices such as NDVI, RENDVI, and mND705.
3. The variable fertilization decision system based on multispectral and soil sensors according to claim 1, characterized in that, The soil sensor network uses a visible-near-infrared spectral sensor with a spectral range of 350-2500nm. Based on a method combining a modified support vector machine and a whale optimization algorithm, it detects the total nitrogen, total phosphorus, and total potassium content in the soil in real time and forms a spatial distribution map of soil nutrients.
4. The variable fertilization decision system based on multispectral and soil sensors according to claim 1, characterized in that, The data processing center includes a data preprocessing module, a coupled decision model, and a prescription map generation module. The coupled decision model is a soil fertility supply capacity-crop fertility demand coupled decision model, which adopts a support vector machine model based on RBF-Poly hybrid kernel function and uses the whale optimization algorithm to optimize the model parameters.
5. The variable fertilization decision system based on multispectral and soil sensors according to claim 1, characterized in that, The coupled decision model adopts a support vector machine model based on the RBF-Poly hybrid kernel function, and the expression of the hybrid kernel function is as follows: in, In the formula, For RBF kernel function, For polynomial kernel functions, It is a hybrid kernel function; It is the mixing coefficient. , Degenerates into a pure RBF nucleus. It degenerates into a pure polynomial kernel; These are the parameters of the kernel function; For kernel width parameters; , This is the offset. Let be the order of the polynomial.
6. The variable fertilization decision system based on multispectral and soil sensors according to claim 1, characterized in that, The data processing center uses the whale optimization algorithm to optimize the coupled decision model. , , , , The optimization algorithm simulates the bubble-net hunting behavior of humpback whales, including three stages: surrounding prey, bubble-net attack, and searching for prey.
7. The variable fertilization decision system based on multispectral and soil sensors according to claim 1, characterized in that, The variable fertilization actuator is a ternary precision variable fertilization machine, which can independently and dynamically control the application amount and ratio of nitrogen, phosphorus and potassium fertilizers, with a fertilization accuracy error of no more than ±5%, and can adjust the fertilization strategy in real time according to the prescription diagram.
8. A variable fertilization decision-making method based on multispectral and soil sensors, characterized in that, Includes the following steps: Step 1: Acquire multispectral image data of crop canopy using multispectral remote sensing units; Step 2: Monitor soil nutrient data in real time using a soil sensor network; Step 3: Preprocess and fuse the multi-source data at the data processing center; Step 4: Construct a coupled decision-making model of soil nutrient supply capacity and crop nutrient demand to dynamically diagnose the causes of crop nutrient stress; Step 5: Generate a variable fertilization prescription diagram; Step 6: Implement precise variable fertilization operations through a variable fertilization implementation agency.
9. The variable fertilization decision-making method based on multispectral and soil sensors according to claim 8, characterized in that, The dynamic diagnosis of crop nutrient stress causes in step four includes: Identify areas of abnormal crop growth by identifying areas of abnormal crop growth using multispectral vegetation index anomalies; Analyze the soil nutrient status corresponding to the abnormal areas; Based on historical data and environmental factors, it was determined that nutrient stress was caused by soil nutrient deficiency, water stress, pests and diseases, or other environmental factors. For stress caused by soil nutrient deficiency, calculate precise fertilization amounts; for stress caused by other factors, generate corresponding agronomic management recommendations.
Citation Information
Patent Citations
Unmanned aerial vehicle accurate variable rate fertilization method and system
CN120077821A
