Intelligent construction method for water collection and corrosion prevention plough layer of slope cropland

Through hyperspectral remote sensing, GIS and Internet of Things technology, a water-catching corrosion-proof tillage layer on sloping farmland has been built, which solves the problems of low resource utilization efficiency and ineffective pest management in traditional sloping farmland management, and achieves targeted agriculture and sustainable development.

CN118941955AInactive Publication Date: 2024-11-12LIAONING PROVINCIAL DRYLAND AGRI & FORESTRY RES INST (LIAONING PROVINCIAL SOIL & WATER CONSERVATION RES INST LIAONING PROVINCIAL ARID AREA AFFORESTATION RES INST)

Patent Information

Application Number
CN202410985729.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The lack of accurate assessment of traditional sloping farmland management has led to inefficient resource utilization, soil erosion and water resources, and ineffective pest management, affecting agricultural production and ecological environment.

Method used

Using hyperspectral remote sensing, GIS, machine learning and IoT technologies, integrating data analysis and intelligent decision-making, building a water-collection and corrosion-proof tillage layer on sloping farmland, and generating color-coded topographic maps through multi-source geospatial data collection and processing, identifying soil types and Water flow path, optimize crop selection, irrigation and fertilization plans, monitor crop growth status in real time, and dynamically adjust farming strategies.

Benefits of technology

It improves the accuracy and efficiency of sloping farmland management, optimizes water resource management and soil protection, strengthens pest management, realizes dynamic optimization and intelligent decision-making, promotes sustainable agricultural development, and reduces production costs and environmental negative impacts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941955A_ABST
    Figure CN118941955A_ABST
Patent Text Reader

Abstract

The invention particularly relates to a slope cropland water collection anti-corrosion plough layer intelligent construction method, which comprises the following steps: collecting multi-source geographic space data, constructing a digital terrain model by using GIS software, calculating to obtain a slope and a slope direction, and identifying and dividing different types of soil by using a hyperspectral remote sensing technology and a machine learning classification algorithm. Fusing the terrain parameters, the soil type information and the slope cropland boundary data, and performing preliminary tillage mode planning; utilizing a machine learning model, based on historical data and field investigation, utilizing a training model to predict optimal farming modes under different topographic conditions, including crop selection, planting density, irrigation and fertilization schemes; prevention and treatment suggestions are provided through big data analysis; deploying an unmanned aerial vehicle and a ground sensor network, and continuously monitoring soil humidity and crop growth state key indexes; monitoring data is fed back to the algorithm model in real time, a farming strategy is optimized, and precise irrigation, fertilization and pest and disease management are achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of smart agriculture, and specifically relates to a method for intelligently constructing a water-collecting and erosion-proof tillage layer on sloping farmland. Background Art

[0002] Sloping farmland management is particularly critical because such terrain is susceptible to soil erosion, leading to soil degradation, reduced productivity, and deterioration of the ecological environment. Sloping farmland usually accounts for a considerable proportion of the world's total cultivated land area, especially in mountainous and hilly areas, so its sustainable management is crucial to maintaining global food security. Traditional sloping farmland management often relies on empirical farming patterns and lacks accurate assessment of terrain, soil properties, and climatic conditions, resulting in inefficient resource utilization. For example, inappropriate crop selection, planting density, and irrigation and fertilization strategies may lead to increased soil erosion, water waste, and fertilizer residue pollution. In addition, pest and disease management is often ineffective due to the lack of real-time monitoring and precise intervention, increasing the abuse of pesticides. In recent years, the development of emerging technologies such as remote sensing technology, geographic information system (GIS), machine learning, and the Internet of Things (IoT) has provided new ways to solve the above problems. Hyperspectral remote sensing can provide fine classification of soil types, while GIS can perform terrain analysis and spatial planning. Machine learning algorithms can learn from large amounts of historical data to predict the optimal farming pattern, while IoT sensor networks can monitor farmland environments in real time to achieve precision agriculture.

[0003] Precision agriculture aims to improve agricultural production efficiency, reduce resource waste and reduce pressure on the environment through intelligent and automated technologies. For sloping farmland, the goal of precision agriculture is to build an effective water-collecting and erosion-proof arable layer, that is, to reduce soil erosion by optimizing farming patterns, while increasing crop yields and soil health.

[0004] In view of the limitations of traditional sloping farmland management and the progress of modern technology, this paper aims to propose an innovative method for intelligent construction of sloping farmland water collection and erosion prevention arable layer, using hyperspectral remote sensing, GIS, machine learning and Internet of Things technology to achieve refined management of sloping farmland. This method aims to improve agricultural production efficiency and reduce resource waste through data analysis and intelligent decision-making, while protecting and restoring the health of sloping ecosystems, and contributing to the realization of sustainable agricultural development. Summary of the invention

[0005] The purpose of the present invention is to provide a method for intelligently constructing a water-collecting and erosion-proof tillage layer on sloping farmland to solve the problems raised in the above-mentioned background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A method for intelligently constructing a water-collecting and erosion-proof plough layer on sloping farmland comprises the following steps:

[0008] Collect multi-source geospatial data to obtain 3D terrain information and surface features of sloping farmland, use GIS software to build digital terrain models, calculate key terrain parameters such as slope and aspect, use hyperspectral remote sensing technology and machine learning classification algorithms to analyze spectral features in RS data, identify and classify different types of soil, integrate terrain parameters and soil type information with sloping farmland boundary data, generate color-coded sloping farmland topographic maps through the GIS platform, use GIS spatial analysis to identify areas of contour farming or micro-terracing, analyze terrain features, and conduct preliminary farming pattern planning;

[0009] Using machine learning models, based on historical data and field research, the trained models are used to predict the most optimized farming patterns under different terrain conditions, including crop selection, planting density, irrigation and fertilization schemes;

[0010] Integrate IoT sensors and AI image recognition technology to monitor crop growth and signs of pests and diseases in real time, and provide prevention and treatment recommendations through big data analysis;

[0011] Deploy drones and ground sensor networks to continuously monitor key indicators of soil moisture and crop growth status;

[0012] Feed monitoring data into the algorithm model in real time to optimize farming strategies and achieve precise irrigation, fertilization, and pest and disease management.

[0013] Furthermore, the key terrain parameters of slope and aspect are calculated, including:

[0014] The slope value of each pixel and the aspect value of each pixel are calculated, and the slope map is classified according to different thresholds. The area with a slope less than 5% is considered flat, 5%-10% is a gentle slope, and greater than 10% is steep;

[0015] Analyze aspect maps to determine changes in sunlight timing and intensity;

[0016] Combine slope and aspect data to conduct hydrological network analysis, identify potential water flow paths and confluence points, and assess soil erosion risks;

[0017] The calculated slope and aspect parameters are combined with soil type information to develop appropriate tillage and management strategies for different soil types, especially in areas prone to erosion.

[0018] Furthermore, hyperspectral remote sensing technology and machine learning classification algorithms are used to analyze the spectral features in RS data and identify and classify different types of soil. Specifically, hyperspectral images covering the target area are obtained from the hyperspectral camera and preprocessed. The reflectance curve of the soil is extracted from the processed hyperspectral images. Soil samples are collected and analyzed in the laboratory to determine their types and key attributes, including pH value, texture, organic matter content, etc., as training labels for the machine learning model. A feature selection algorithm is used to select the most discriminative bands from the high-dimensional spectral data. A deep neural network (DNN) supervised learning algorithm is used to train the classification model based on spectral features and soil type labels. K-fold cross-validation or leave-one-out method is used to evaluate the generalization ability of the model. The best model is found by adjusting the model parameters. The trained model is applied to the hyperspectral data of the entire area to generate a soil type classification map. The soil type information is superimposed and analyzed with the calculated slope and aspect terrain parameters to ensure that each sloping farmland plot has complete terrain and soil properties. GIS is used to identify the areas most suitable for the growth of specific crops and high-risk areas prone to soil erosion.

[0019] Furthermore, terrain parameters, soil type information and sloping farmland boundary data are integrated, including:

[0020] Create an attribute table for each sloping farmland plot, record the plot ID, area, average slope, dominant slope aspect, soil type key information, if the soil data contains additional attributes, including soil pH, organic matter content, sand / clay ratio, should also be associated;

[0021] Small plots of sloping farmland are merged into larger management units, ensuring that the aggregated plots retain the statistical averages of the topography and soil properties of the original plots;

[0022] Create a comprehensive sloping farmland map where plots are represented by different colors or patterns to represent different soil types and slope classes.

[0023] Further, preliminary farming pattern planning is carried out, including:

[0024] Consider the effects of slope and aspect to identify areas suitable for contour farming or micro-terracing, with flat areas used for conventional farming, gently sloping areas for contour farming, and steep areas for micro-terracing;

[0025] Determine soil suitability for crop cultivation based on soil characteristics and terrain conditions;

[0026] Develop crop rotation plans based on soil health and erosion control needs, as well as soil improvements where needed;

[0027] Based on terrain parameters, determine ridge width, furrow width and ridge height, and adjust row and plant spacing to ensure maximum crop coverage;

[0028] Plant vegetation in marginal areas prone to erosion and plant trees on or around sloping farmland;

[0029] Combined with IoT sensor data, precision irrigation is implemented to provide water at the right time and in the right amount based on soil moisture and crop needs.

[0030] Furthermore, the steps of using machine learning models to predict the optimal farming mode include data preparation, feature engineering, model training, prediction and optimization, and implementation and monitoring, where model training uses a neural network model to predict the optimal farming mode under different terrain conditions, including crop selection, planting density, irrigation and fertilization schemes. The neural network model is constructed as follows: the input layer receives remote sensing images and other feature data, the convolution layer is used to extract image features, including multiple convolution layers and pooling layers, the fully connected layer is used to process non-image features and integrate all features, and the output layer predicts crop selection, planting density, irrigation and fertilization schemes; the mathematical expression of the operation is:

[0031] y=(x*w)+b;

[0032] Where y is one of the control targets, including crop selection, planting density, irrigation and fertilization schemes, x is the input image, w is the weight matrix, and b is the bias term;

[0033] The activation function is

[0034] ReLU(f(x)=max(0,x));

[0035] The loss function uses cross entropy loss for classification tasks and mean square error for regression tasks.

[0036] Furthermore, a convolutional neural network model is used to predict the optimal crop selection under different terrain conditions. The prediction model is:

[0037]

[0038] in,

[0039] is the predicted probability distribution of crop selection;

[0040] X is the input data, including terrain images and soil types;

[0041] W1 and W2 are weight matrices, corresponding to the weights of the convolutional layer and the fully connected layer, respectively;

[0042] b1 and b2 are bias terms;

[0043] σ is the activation function, which is ReLU;

[0044] Softmax is an activation function used for multi-classification tasks that converts the output into a probability distribution.

[0045] Furthermore, a hybrid model is used to predict the optimal planting density, irrigation and fertilization scheme. The hybrid model includes a convolutional neural network model and a recurrent neural network. The convolutional neural network model processes image data, and the recurrent neural network model processes non-image data. Non-image data includes soil type, historical climate data, and crop requirements. The mathematical expression of the convolutional neural network model for image feature extraction is:

[0046] f image =σ(W conv *X image +b cone );

[0047] in,

[0048] X image is the input of terrain image;

[0049] W conv and b conv They are the weights and biases of the convolutional layer respectively;

[0050] σ is the activation function;

[0051] The mathematical expression of the recurrent neural network processing of non-image features is:

[0052] f non-image =σ(W fc ·X non-image +b fc )

[0053] in,

[0054] X non-image Includes non-image data such as soil type, historical climate data, and crop requirements;

[0055] W fc and b fc are the weights and biases of the fully connected layer;

[0056] The mathematical expression of feature fusion is:

[0057] f combined =Concat(f image ,f non-image )

[0058] in,

[0059] Concat represents the concatenation of feature vectors;

[0060] For planting density, irrigation, and fertilization, there are three independent output layers, each of which is responsible for predicting the corresponding agricultural parameters. For planting density, the mathematical expression of the output layer is:

[0061]

[0062] For irrigation, the mathematical expression of the output layer is:

[0063]

[0064] For the fertilization scheme, the mathematical expression of the output layer is:

[0065]

[0066] in,

[0067] W density , W irrigation and W fertilization are the weights of each output layer;

[0068] b density , b irrigation and b fertilization are the biases of each output layer.

[0069] Beneficial effects:

[0070] The technical solution of this application is aimed at the intelligent construction method of the water collection and erosion prevention arable layer of sloping farmland. By integrating hyperspectral remote sensing, machine learning, geographic information system (GIS) and Internet of Things (IoT) technology, it realizes the refined management and soil and water conservation of sloping farmland. Its beneficial effects are particularly reflected in the following aspects:

[0071] 1. Improved the accuracy and efficiency of sloping farmland management

[0072] The application of hyperspectral remote sensing technology makes the identification of soil types more accurate and can distinguish different types of soil and their properties, providing a scientific basis for subsequent crop selection, irrigation and fertilization plans.

[0073] The introduction of machine learning models makes it possible to analyze complex environmental factors. By learning from historical data, the model can predict the optimal farming pattern, including crop selection, planting density, irrigation and fertilization plans, greatly improving the scientificity and efficiency of farming decisions.

[0074] 2. Optimized water resource management and soil protection

[0075] Digital terrain models and hydrological network analysis help identify potential water flow paths and confluence points, assess soil erosion risks, and provide a basis for designing contour farming and micro-terracing, which reduces runoff velocity, increases surface water infiltration, and effectively controls soil erosion.

[0076] The implementation of precision irrigation strategies, combined with IoT sensor data, can provide timely and appropriate watering based on soil moisture and crop needs, avoiding over-irrigation, reducing water waste and soil structure damage.

[0077] 3. Strengthened pest and disease management and crop health

[0078] The integration of AI image recognition technology and IoT sensors can monitor crop growth conditions and signs of pests and diseases in real time, provide prevention and treatment recommendations through big data analysis, reduce the risk of pest and disease outbreaks, and ensure crop health.

[0079] 4. Realized dynamic optimization and intelligent decision-making

[0080] The dynamic optimization mechanism enables farming strategies to be adjusted based on real-time monitoring data of seasonal changes, crop growth cycles and environmental conditions, improving the flexibility and responsiveness of agricultural production.

[0081] Model iteration and optimization ensure that the accuracy of the prediction model improves over time, constantly adapts to new agricultural environments, and enhances the reliability of agricultural decision-making.

[0082] 5. Promoted sustainable agricultural development

[0083] Comprehensive planning and dynamic adjustment mechanism, combined with GIS-assisted planning, generate visual farming pattern planning maps, which help to formulate farming patterns that meet local terrain and soil conditions, promote the sustainability of agricultural production, and reduce negative impacts on the environment.

[0084] 6. Reduced agricultural inputs and costs

[0085] Through precise irrigation, fertilization and pest and disease management, unnecessary resource consumption is reduced, the efficiency of agricultural inputs is improved, and agricultural production costs are reduced.

[0086] 7. Improved agricultural production efficiency and crop yields

[0087] Optimized farming patterns and dynamic adjustment strategies, combined with intelligent decision support, help improve crop yield and quality, increase agricultural output, and improve agricultural economic benefits.

[0088] To sum up, the technical solution of this application realizes the construction of an intelligent method for constructing a water-collecting and erosion-proof tillage layer on sloping farmland through advanced technical means. It not only effectively controls soil and water loss, but also improves agricultural production efficiency and resource utilization, which is of great significance to promoting the sustainable development of agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 This is a flow chart of the application method. DETAILED DESCRIPTION

[0090] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0091] The present invention provides a method for intelligently constructing a water-collecting and anti-erosion tillage layer on sloping farmland. Figure 1 As shown, the steps include:

[0092] Collect multi-source geospatial data to obtain 3D terrain information and surface features of sloping farmland, use GIS software to build digital terrain models, calculate key terrain parameters such as slope and aspect, use hyperspectral remote sensing technology and machine learning classification algorithms to analyze spectral features in RS data, identify and classify different types of soil, integrate terrain parameters and soil type information with sloping farmland boundary data, generate color-coded sloping farmland topographic maps through the GIS platform, use GIS spatial analysis to identify areas of contour farming or micro-terracing, analyze terrain features, and conduct preliminary farming pattern planning;

[0093] Using machine learning models, based on historical data and field research, the trained models are used to predict the most optimized farming patterns under different terrain conditions, including crop selection, planting density, irrigation and fertilization schemes;

[0094] Integrate IoT sensors and AI image recognition technology to monitor crop growth and signs of pests and diseases in real time, and provide prevention and treatment recommendations through big data analysis;

[0095] Deploy drones and ground sensor networks to continuously monitor key indicators of soil moisture and crop growth status;

[0096] Feed monitoring data into the algorithm model in real time to optimize farming strategies and achieve precise irrigation, fertilization, and pest and disease management.

[0097] Among them, we collected multi-source geospatial data, obtained the three-dimensional terrain information and surface features of sloping farmland, used GIS software to build a digital terrain model, and calculated the key terrain parameters of slope and aspect, including:

[0098] The slope value of each pixel and the aspect value of each pixel are calculated, and the slope map is classified according to different thresholds. The area with a slope less than 5% is considered flat, 5%-10% is a gentle slope, and greater than 10% is steep;

[0099] Analyze aspect maps to determine changes in sunlight timing and intensity;

[0100] Combine slope and aspect data to conduct hydrological network analysis, identify potential water flow paths and confluence points, and assess soil erosion risks;

[0101] The calculated slope and aspect parameters are combined with soil type information to develop appropriate tillage and management strategies for different soil types, especially in areas prone to erosion.

[0102] In the embodiment, collecting multi-source geospatial data to obtain three-dimensional terrain information and surface features of sloping farmland specifically includes the following steps:

[0103] Data Collection and Processing

[0104] High-resolution satellite imagery

[0105] Diversified sources: Imagery is acquired from multiple satellite platforms, such as the Sentinel series, Landsat, MODIS, or commercial satellites such as Planet Labs and Maxar Technologies to ensure data diversity and timeliness.

[0106] Data fusion: Use multi-temporal images to improve resolution and reduce cloud cover effects through image fusion techniques, such as principal component analysis (PCA) and wavelet transform.

[0107] Drone oblique photography data

[0108] High-precision positioning: Using RTK-GPS technology, the images taken by the drone can be guaranteed to have sub-meter accuracy.

[0109] Multi-view acquisition: Shooting from different heights and angles to obtain stereoscopic information for 3D modeling.

[0110] LiDAR data

[0111] Point cloud processing: Filter the LiDAR point cloud data, remove vegetation and non-ground objects, and extract pure ground point clouds for building accurate DEM (digital elevation model).

[0112] High-density sampling: Ensure that the point density of the LiDAR data is high enough to capture complex terrain features such as gullies and ridges.

[0113] Digital terrain model construction

[0114] DEM generation: Use LiDAR point cloud data to build a high-precision DEM as the basis for the digital terrain model.

[0115] DSM & DOM: Generate digital surface model (DSM) and digital orthophoto (DOM) simultaneously for analyzing land cover and terrain characteristics.

[0116] Terrain parameter calculation

[0117] Slope and Aspect: Automatically calculate the slope and aspect of each pixel from the DEM, providing key terrain parameters for subsequent analysis.

[0118] Terrain curvature analysis: Calculates the profile and plan curvature of terrain for use in assessing soil erosion potential and water flow patterns.

[0119] Hyperspectral remote sensing and soil type identification

[0120] Spectral feature extraction: Extract soil reflectance curves from hyperspectral remote sensing data to distinguish different types of soil.

[0121] Machine learning classification: Use support vector machines (SVM), random forests (RF), or other supervised learning algorithms to identify soil types based on training samples.

[0122] Data fusion and mapping

[0123] Multi-source data integration: Integrate DEM, DOM, DSM with soil type classification results and sloping farmland boundary data on the GIS platform.

[0124] Property Mapping: Assign specific colors or textures to each soil type and terrain parameter for clear presentation on a map.

[0125] Quality control and update mechanism

[0126] Regular updates: Establish a data update cycle, such as re-collecting and processing data every year or after every major terrain change.

[0127] Data verification: Use ground-based measured data to verify remote sensing analysis results to ensure data accuracy and reliability.

[0128] Through the above-mentioned refinement steps, it can be ensured that the collected geospatial data are not only rich but also of reliable quality, providing solid data support for the subsequent intelligent construction method of the sloping farmland water collection and erosion prevention tillage layer.

[0129] Use GIS software to construct a digital terrain model and calculate key terrain parameters such as slope and aspect, including:

[0130] Digital Terrain Model (DTM) Construction

[0131] Data sources: Use high-resolution satellite images, drone oblique photography data, LiDAR data and other multi-source geospatial data as the basis for building DTM. LiDAR data are particularly suitable for obtaining high-precision terrain surface information because they can penetrate the vegetation layer and directly measure the ground height.

[0132] Data processing: First, the collected raw data is preprocessed, including radiation correction, geometric correction, stitching and registration, to eliminate image distortion and improve data consistency. Then, the point cloud or raster data processing module of GIS software is used to extract terrain features from LiDAR point cloud or DEM (digital elevation model).

[0133] Calculating slope and aspect

[0134] Slope calculation: Slope refers to the degree of inclination of any point on the ground, usually expressed as a percentage or angle. In GIS software, the slope value of each pixel can be calculated from the DTM using the "Slope" tool. This step will generate a slope map that can be used to visualize and quantify the degree of inclination of the entire study area.

[0135] Aspect calculation: Aspect refers to the inclination direction of any point on the ground, usually expressed as azimuth, which is the angle rotated clockwise from the north. Similarly, in the GIS environment, the "Aspect" tool can be used to calculate the aspect value of each pixel from the DTM. The aspect map can reveal the lighting conditions and precipitation paths on the terrain, which is crucial for crop planting and water resource management.

[0136] Parameter analysis and application

[0137] Slope classification: Slope maps are classified according to different thresholds, for example, areas with slopes less than 5% are considered flat, 5%-10% are gently inclined, and greater than 10% are steep. Such classification helps identify areas susceptible to erosion and guide farming strategies.

[0138] Slope aspect analysis: Analyze the aspect map to determine the changes in sunlight time and intensity, which is important for crop selection and placement. South-facing slopes may be more suitable for growing sun-loving crops in temperate regions, while north-facing slopes may be more suitable for shade-tolerant crops.

[0139] Hydrological analysis: Combined with slope and aspect data, hydrological network analysis can be performed to identify potential water flow paths and confluence points, assess soil erosion risks, and provide a scientific basis for water collection and erosion prevention measures.

[0140] Integration and Optimization

[0141] Integration with soil type data: Combining the calculated slope and aspect parameters with soil type information allows for the development of appropriate tillage and management strategies for different soil types, especially in areas prone to erosion.

[0142] Dynamic optimization: As the seasons change and the crop growth cycle varies, the impact of slope and aspect on farming will also change. Therefore, DTM and related parameters should be updated regularly to reflect the latest terrain characteristics, so as to optimize farming patterns and erosion prevention measures.

[0143] Hyperspectral remote sensing technology and machine learning classification algorithms are used to analyze the spectral features in RS data and identify and classify different types of soil. Specifically, hyperspectral images covering the target area are obtained from the hyperspectral camera, preprocessed, and the soil reflectance curve is extracted from the processed hyperspectral images. Soil samples are collected and analyzed in the laboratory to determine their types and key attributes, including pH value, texture, organic matter content, etc., as training labels for the machine learning model. A feature selection algorithm is used to select the most discriminative bands from high-dimensional spectral data. A deep neural network (DNN) supervised learning algorithm is used to train the classification model based on spectral features and soil type labels. K-fold cross-validation or leave-one-out method is used to evaluate the generalization ability of the model. The best model is found by adjusting the model parameters. The trained model is applied to the hyperspectral data of the entire area to generate a soil type classification map. The soil type information is superimposed and analyzed with the calculated slope and aspect terrain parameters to ensure that each sloping cultivated land plot has complete terrain and soil properties. GIS is used to identify the areas most suitable for the growth of specific crops and high-risk areas prone to soil erosion.

[0144] In the embodiment, hyperspectral remote sensing technology and machine learning classification algorithms are used to analyze the spectral features in RS data to identify and classify different types of soil, as follows:

[0145] Hyperspectral remote sensing data processing and feature extraction

[0146] Data acquisition: Obtain hyperspectral images covering the target area from hyperspectral satellites (such as Hyperion, PRISMA) or hyperspectral cameras carried by drones, ensuring that the data covers the entire visible light to near-infrared band to capture the spectral characteristics of the soil.

[0147] Preprocessing: radiation correction, atmospheric correction and geometric correction are performed to eliminate the influence of atmospheric effects, changes in the sun's position and terrain shadows to ensure the accuracy of spectral data.

[0148] Feature extraction: Extract soil reflectance curves from processed hyperspectral images. The reflectance differences of these curves at different wavelengths reflect the physical and chemical properties of the soil, such as organic matter content, water content, mineral composition, etc.

[0149] Machine Learning Classification

[0150] Training data preparation: Soil samples are collected and analyzed in the laboratory to determine their type and key properties such as pH, texture, organic matter content, etc., which serve as training labels for the machine learning model.

[0151] Feature selection: Use feature selection algorithms (such as recursive feature elimination (RFE) and feature importance of random forests) to pick out the most discriminative bands from high-dimensional spectral data.

[0152] Model training: Use supervised learning algorithms such as support vector machine (SVM), random forest (RF), deep neural network (DNN), etc. to train classification models based on spectral features and soil type labels.

[0153] Cross-validation: Use k-fold cross-validation or leave-one-out cross-validation to evaluate the generalization ability of the model to ensure that the model performs well on unseen data.

[0154] Model optimization: By adjusting model parameters (such as the kernel function of SVM, C parameter, number of trees of RF, etc.), find the best model configuration to improve classification accuracy.

[0155] Soil type identification and mapping

[0156] Classification result output: Apply the trained model to the hyperspectral data of the entire area to generate a soil type classification map.

[0157] Post-processing: Perform post-classification processing, such as majority voting or morphological operations, to reduce classification noise and smooth classification boundaries.

[0158] Calibration and verification: The classification results are compared with the soil types measured on the ground for calibration and verification to ensure the accuracy of the classification.

[0159] Soil type mapping: Using GIS software, the soil type classification results are mapped, a specific color or pattern is assigned to each soil type, and a detailed soil type distribution map is generated.

[0160] Combining terrain parameters with soil type information

[0161] Attribute association: Soil type information is superimposed and analyzed with previously calculated terrain parameters such as slope and aspect to ensure that each sloping farmland plot has complete terrain and soil attributes.

[0162] Spatial analysis: Using GIS’s multi-criteria decision analysis (MCDA) tool, factors such as soil type, slope, and aspect are considered to identify areas most suitable for growing specific crops and areas at high risk of soil erosion.

[0163] Through the above steps, hyperspectral remote sensing technology and machine learning algorithms can be used to accurately identify and classify different types of soil, providing a scientific basis for the fine management of sloping farmland and the construction of water-collecting and erosion-resistant arable layers.

[0164] The terrain parameters, soil type information and sloping farmland boundary data are integrated, including:

[0165] Create an attribute table for each sloping farmland plot, record the plot ID, area, average slope, dominant slope aspect, soil type key information, if the soil data contains additional attributes, including soil pH, organic matter content, sand / clay ratio, should also be associated;

[0166] Small plots of sloping farmland are merged into larger management units, ensuring that the aggregated plots retain the statistical averages of the topography and soil properties of the original plots;

[0167] Create a comprehensive sloping farmland map where plots are represented by different colors or patterns to represent different soil types and slope classes.

[0168] In the embodiment, the fusion of terrain parameters, soil type information and sloping farmland boundary data includes the following steps:

[0169] Data preprocessing

[0170] Data cleaning: Check the consistency and completeness of all data sets, exclude noise and outliers, and ensure that all data are in the same coordinate system.

[0171] Projection conversion: If the data comes from different geographic spatial reference systems, perform necessary projection conversion so that the analysis can be performed under a unified framework.

[0172] Data fusion steps

[0173] Overlay analysis:

[0174] Use the functions of GIS software, such as “Spatial Join” of ArcGIS or “JoinAttributes byLocation” of QGIS, to overlay the sloping farmland boundary data with terrain parameters (slope, aspect).

[0175] For soil type data, a similar overlay method is used to ensure that each piece of sloping farmland has corresponding soil type information.

[0176] Attribute association:

[0177] Create an attribute table for each sloping farmland plot to record key information such as the plot’s ID, area, average slope, dominant slope aspect, soil type, etc.

[0178] If the soil data contains additional attributes, such as soil pH, organic matter content, sand / clay ratio, etc., these should also be associated.

[0179] Spatial aggregation:

[0180] When necessary, small plots of sloping farmland are merged into larger management units, which helps simplify subsequent farming planning and management.

[0181] Ensure that the aggregated plots still retain the statistical averages of the topographic and soil properties of the original plots.

[0182] Generate a comprehensive map:

[0183] Using the mapping capabilities of GIS, create a comprehensive sloping farmland map in which plots are represented by different colors or patterns to represent different soil types and slope grades.

[0184] Additional layers such as contour lines, hydrographic networks, roads, and buildings can be added to provide a more comprehensive geographic context.

[0185] Interactive query function:

[0186] Develop GIS applications or WebGIS interfaces that enable users to view detailed properties of parcels by clicking or selecting specific areas.

[0187] The interface should include search functionality allowing filtering and querying by parcel ID, soil type, or terrain parameters.

[0188] Data quality control

[0189] Soil and topographic data are regularly updated to reflect the impacts of land use changes and soil improvement measures.

[0190] Implement quality control processes to regularly review and validate data for accuracy and consistency.

[0191] Application Examples

[0192] Develop specific crop rotation plans based on soil type and slope to optimize soil health and reduce erosion.

[0193] Adjust irrigation plans according to soil permeability and water retention capacity to avoid soil structure damage and nutrient loss caused by excessive irrigation.

[0194] Generate color-coded sloping farmland topographic maps through the GIS platform, use GIS spatial analysis to identify areas of contour farming or micro-terracing, analyze terrain characteristics, and conduct preliminary farming pattern planning;

[0195] Conduct preliminary farming pattern planning, including:

[0196] Consider the effects of slope and aspect to identify areas suitable for contour farming or micro-terracing, with flat areas used for conventional farming, gently sloping areas for contour farming, and steep areas for micro-terracing;

[0197] Determine soil suitability for crop cultivation based on soil characteristics and terrain conditions;

[0198] Develop crop rotation plans based on soil health and erosion control needs, as well as soil improvements where needed;

[0199] Based on terrain parameters, determine ridge width, furrow width and ridge height, and adjust row and plant spacing to ensure maximum crop coverage;

[0200] Plant vegetation in marginal areas prone to erosion and plant trees on or around sloping farmland;

[0201] Combined with IoT sensor data, precision irrigation is implemented to provide water at the right time and in the right amount based on soil moisture and crop needs.

[0202] Conducting preliminary tillage pattern planning is an important step after completing terrain analysis, soil type identification and attribute association. It involves determining the most appropriate tillage method based on terrain characteristics, soil conditions and hydrological conditions. The following are the detailed steps of preliminary tillage pattern planning:

[0203] Analyze terrain and soil properties

[0204] Topographic considerations: Consider the effects of slope and aspect to identify areas suitable for contour farming or micro-terracing. Flat areas (slope <5%) may be suitable for conventional farming, gently sloping areas (5%-10%) may use contour farming, and steep areas (>10%) may require micro-terracing or other soil and water conservation measures.

[0205] Soil property assessment: Determine which soil types are suitable for specific crops and which soils require amendments such as lime or organic matter based on information such as soil type, pH, organic matter content and texture (sand / clay ratio).

[0206] Determining crop suitability

[0207] Crop selection: Choose the types of crops that are best suited to growing in a particular area, based on soil and topographic conditions. For example, south-facing slopes may be better suited to sun-loving crops, while north-facing slopes may be better suited to shade-tolerant crops.

[0208] Crop rotation plan: Develop a crop rotation plan based on soil health and erosion control needs to ensure soil fertility is maintained and pests and diseases are managed.

[0209] Design tillage parameters

[0210] Ridge width, furrow width and ridge height: Based on terrain parameters, determine the optimal ridge width, furrow width and ridge height to promote effective water management and minimize soil erosion. For example, steep areas may require wider ridges and deeper furrows to slow runoff.

[0211] Row and plant spacing: Adjust row and plant spacing to ensure maximum crop cover, thereby reducing soil exposure and lowering the risk of erosion.

[0212] Implement soil and water conservation measures

[0213] Contour tillage: Tillage along contour lines on slopes to reduce runoff velocity and soil erosion.

[0214] Micro-terracing: Building micro-terraces in steep areas to increase the infiltration rate of surface water and reduce soil erosion.

[0215] Vegetation buffer strip: Planting vegetation in marginal areas prone to erosion to stabilize the soil and reduce soil erosion.

[0216] Soil and water conservation forest belt: Planting trees on or around sloping farmland to intercept rainwater and reduce surface runoff.

[0217] Consider water harvesting and irrigation strategies

[0218] Water harvesting measures: Design water harvesting structures, such as small reservoirs or cisterns, to collect and store rainwater for use in irrigation during dry periods.

[0219] Precision irrigation: Combined with IoT sensor data, precision irrigation is implemented to water crops at the right time and in the right amount according to soil moisture and crop needs, avoiding soil structure damage and nutrient loss caused by over-irrigation.

[0220] Comprehensive planning and dynamic adjustment

[0221] GIS-assisted planning: Use the GIS platform to integrate terrain, soil and crop information to generate visual farming pattern planning maps.

[0222] Dynamic Optimization: Dynamically adjust farming strategies to adapt to the ever-changing agricultural environment based on real-time monitoring data of seasonal changes, crop growth cycles and environmental conditions.

[0223] Through the above steps, a preliminary farming pattern plan can be developed that is in line with local terrain and soil conditions, can effectively control soil erosion, and promote crop growth. This will not only help improve agricultural production efficiency, but also ensure the sustainable use of land resources.

[0224] Using machine learning models, based on historical data and field research, the trained models are used to predict the most optimized farming patterns under different terrain conditions, including crop selection, planting density, irrigation and fertilization schemes;

[0225] Using machine learning models to predict the most optimal farming patterns under different terrain conditions involves the following key steps:

[0226] Data collection and preprocessing

[0227] 1. Historical farming data: Collect historical data such as past farming patterns, crop yields, soil health, irrigation and fertilization records, etc.

[0228] 2. Topography and soil data: Integrate the terrain parameters (such as slope and aspect) and soil type information obtained from the previous analysis.

[0229] 3. Climate data: including meteorological data such as rainfall, temperature, wind speed, etc., which can be obtained from weather stations or climate databases.

[0230] 4. Preprocessing: clean data, handle missing values, standardize or normalize numerical features, and encode categorical features.

[0231] Feature Engineering

[0232] 1. Feature selection: Select features that have a significant impact on the farming pattern based on domain knowledge and statistical tests.

[0233] 2. Feature construction: Create new features, such as soil health index, climate suitability index, etc., to better reflect the farming environment.

[0234] Model training and selection

[0235] 1. Model selection: Use deep learning models (convolutional neural network CNN, recurrent neural network RNN).

[0236] 2. Cross-validation: Use k-fold cross-validation to evaluate model performance and prevent overfitting.

[0237] 3. Hyperparameter tuning: Apply grid search or random search to find the best parameter combination for the model.

[0238] Model Training

[0239] 1. Training set and test set split: The data is divided into training set and test set. The training set is used for model training, and the test set is used to evaluate the generalization ability of the model.

[0240] 2. Model training: Use the training set data to train the selected model and adjust the weights to minimize the loss function.

[0241] 3. Model evaluation: Evaluate the model on the test set, using metrics such as accuracy, F1 score, mean squared error, etc.

[0242] Prediction and Optimization

[0243] 1. Model application: Apply the model to newly collected terrain, soil and climate data to predict optimal farming patterns.

[0244] 2. Result interpretation: Analyze the model output and understand the impact of different variables on the prediction results.

[0245] 3. Dynamic optimization: Regularly update model inputs and adjust farming strategies based on crop growth cycles and environmental changes.

[0246] Implementation and Monitoring

[0247] 1. Implementation plan: Based on the forecast results, develop a specific farming plan, including crop selection, planting density, irrigation and fertilization schemes.

[0248] 2. Real-time monitoring: Combine IoT sensors and AI image recognition to monitor farming results in real time and collect feedback data.

[0249] 3. Model iteration: Based on real-time monitoring data and actual farming results, the model is continuously iterated and optimized to improve prediction accuracy.

[0250] Through the above steps, a machine learning-based prediction system can be established, which can predict the optimal farming pattern according to terrain, soil and climate conditions, and then guide precision agriculture practices to increase crop yields while reducing resource waste and environmental damage.

[0251] The steps of using machine learning models to predict the optimal farming mode include data preparation, feature engineering, model training, prediction and optimization, and implementation and monitoring. Model training uses a neural network model to predict the optimal farming mode under different terrain conditions, including crop selection, planting density, irrigation and fertilization schemes. The model predicts the optimal farming mode under different terrain conditions, including:

[0252] Data preparation

[0253] 1. Historical data collection: Collect historical farming data, including crop types, planting density, irrigation frequency, fertilizer application, soil moisture, crop yield and health status, etc.

[0254] 2. Topography and soil data: Integrate information such as slope, aspect, soil type, pH value, and organic matter content.

[0255] 3. Climate data: including meteorological data such as rainfall, temperature, and humidity.

[0256] 4. Preprocessing: clean data, handle missing values, encode categorical features, and standardize or normalize numerical features.

[0257] Feature Engineering

[0258] 1. Feature extraction: Extracting terrain features from remote sensing images, such as using CNN to extract from high-resolution satellite imagery or oblique photography data taken by drones.

[0259] 2. Feature construction: Create new features based on domain knowledge, such as soil health index, crop suitability score, etc.

[0260] Model building and training

[0261] 1. CNN model design:

[0262] The input layer receives remote sensing images and other feature data.

[0263] The convolutional layer is used to extract image features and may include multiple convolutional layers and pooling layers.

[0264] The fully connected layers are used to process non-image features and integrate all features.

[0265] The output layer predicts crop selection, planting density, irrigation, and fertilization schedules.

[0266] 2. Training process:

[0267] Train the model using labeled historical data.

[0268] The weights are updated using back-propagation algorithm and gradient descent method.

[0269] Cross-validation techniques were used to evaluate model performance and prevent overfitting.

[0270] 3. Mathematical expression:

[0271] Convolution operation: The mathematical expression of the operation is:

[0272] y=(x*w)+b;

[0273] Where y is one of the control objectives, including crop selection, planting density, irrigation and fertilization schemes, x is the input image, w is the weight matrix, and b is the bias term.

[0274] Activation function: ReLU (f(x) = max(0, x)).

[0275] Loss function: Cross entropy loss is used for classification tasks, and mean squared error is used for regression tasks.

[0276] Prediction and Optimization

[0277] 1. Model application: Apply the model to new terrain and soil data to predict the optimal farming pattern.

[0278] 2. Dynamic optimization: Regularly update model inputs and adjust farming strategies based on crop growth cycles and environmental changes.

[0279] Implementation and Monitoring

[0280] 1. Implementation plan: Develop a specific farming plan based on the forecast results.

[0281] 2. Real-time monitoring: Combine IoT sensors and AI image recognition to monitor farming results in real time and collect feedback data.

[0282] 3. Model iteration: Continuously iterate and optimize the model based on real-time monitoring data and actual farming results.

[0283] Using convolutional neural network models to predict the optimal crop selection under different terrain conditions, crop selection is a key decision point when using machine learning models to predict the optimal farming mode under different terrain conditions. Based on historical data and field research, we can train a model to predict which crop is the best choice under given terrain and soil conditions. The following is a specific implementation example, including the mathematical formula corresponding to crop selection:

[0284] Data preparation

[0285] First, historical farming data, terrain data (such as slope, aspect), soil data (such as pH value, organic matter content), climate data (such as rainfall, temperature), and crop growth and yield data are collected. These data are used to train and validate the model.

[0286] Feature Engineering

[0287] Preprocess remote sensing images to extract key terrain features.

[0288] Encode soil and climate data as numerical features.

[0289] Create a feature matrix containing all input features.

[0290] Model building

[0291] Build a neural network model that includes convolutional layers and fully connected layers. The convolutional layers are used to process image data, and the fully connected layers are used to process non-image feature data. The output of the model is the probability distribution of crop selection.

[0292] Mathematical expressions

[0293] For the crop selection model, the mathematical expression can be written as:

[0294] The prediction model is:

[0295]

[0296] in,

[0297] is the predicted probability distribution of crop selection;

[0298] X is the input data, including terrain images and soil types;

[0299] W1 and W2 are weight matrices, corresponding to the weights of the convolutional layer and the fully connected layer, respectively;

[0300] b1 and b2 are bias terms;

[0301] σ is the activation function, which is ReLU;

[0302] Softmax is an activation function used for multi-classification tasks that converts the output into a probability distribution.

[0303] Training the model

[0304] The model is trained using labeled historical data, and the weights and biases are adjusted to minimize the loss function. The loss function could be the cross entropy loss, as it is suitable for multi-class classification problems.

[0305] Prediction and Optimization

[0306] Apply the trained model to a new dataset to predict the optimal crop selection. The output of the model is a probability distribution of different crop types, from which the crop with the highest probability is selected as the recommended crop to be planted.

[0307] Implementation and Monitoring

[0308] Once the optimal crop selection is determined, planting can begin. At the same time, IoT sensors and AI image recognition technology are used to monitor crop growth conditions and signs of pests and diseases in real time, and preventive and treatment recommendations are provided through big data analysis. This will help adjust farming strategies and achieve precise irrigation, fertilization, and pest and disease management.

[0309] example

[0310] Assume we have the following input data:

[0311] Terrain image, size is n*m pixels.

[0312] Soil type information, including pH, organic matter content, etc.

[0313] Historical climate data, such as average temperature and precipitation.

[0314] Past crop growth and yield data.

[0315] Our model first processes the terrain image through a convolutional layer to extract key features. The extracted image features are then combined with non-image features (soil type, climate data) and fed into a fully connected layer for prediction.

[0316] Assuming we have K different crop types, the output of the model will be a vector of length K, representing the probability of each crop being selected. We select the crop with the highest probability as the recommended crop to be planted.

[0317] This model can be continuously optimized as new data is added, thereby continuously improving the accuracy of crop selection and thus improving the efficiency and sustainability of agricultural production.

[0318] Furthermore, using models to predict the optimal planting density, irrigation, and fertilization schemes involves transforming multiple input data into specific outputs. To integrate these different types of data, we use a hybrid model in which CNN is used to process image data, while fully connected layers or recurrent neural networks (RNNs) are used to process sequence or non-image data.

[0319] Specifically, a hybrid model is used to predict the optimal planting density, irrigation and fertilization scheme. The hybrid model includes a convolutional neural network model and a recurrent neural network. The convolutional neural network model processes image data, and the recurrent neural network model processes non-image data. The non-image data includes soil type, historical climate data, and crop requirements. The hybrid model is used to predict the optimal planting density, irrigation and fertilization scheme, which specifically includes the following steps:

[0320] Data preparation and feature engineering

[0321] Collect historical farming data, including but not limited to crop yield, growth cycle, soil moisture, fertilizer application amount, irrigation frequency, etc.

[0322] Integrate topographic and soil data, including slope, aspect, soil type, pH, organic matter content, etc.

[0323] Get climate data such as rainfall, temperature, humidity, etc.

[0324] Preprocess the data, including cleaning, filling missing values, feature encoding, and standardization.

[0325] Extract remote sensing image features and use CNN for image feature extraction.

[0326] Construct non-image features, such as features based on historical data and expert knowledge.

[0327] Model building

[0328] Build a hybrid model combining CNN and RNN (or LSTM):

[0329] The CNN part is used to process image data (such as terrain images) and extract terrain features.

[0330] The RNN / LSTM part is used to process time series data (such as historical climate data) and non-image data.

[0331] The fully connected layer is used to integrate all features, including image features and non-image features.

[0332] The output layer is responsible for predicting planting density, irrigation, and fertilization schemes.

[0333] Model Training

[0334] Divide the dataset into training set, validation set and test set.

[0335] Use the training set data to train the model, and update the weights and biases through the back-propagation algorithm and optimizer (such as Adam). Evaluate the model performance on the validation set, adjust the hyperparameters, and prevent overfitting.

[0336] Finally, the generalization ability of the model is evaluated on the test set.

[0337] Prediction and Optimization

[0338] Apply the model to new data to predict optimal farming patterns.

[0339] Monitor crop growth conditions and environmental changes in real time, and collect data through IoT sensors.

[0340] Feed monitoring data back to the model to dynamically adjust prediction results and farming strategies.

[0341] Mathematical Expression Examples

[0342] The above steps are converted into specific mathematical expressions to predict planting density, irrigation and fertilization schemes:

[0343] The mathematical expression of image feature extraction by the convolutional neural network model is:

[0344] f image =σ(W conv *X image +b conv );

[0345] in,

[0346] X image is the input of terrain image;

[0347] W conv and b conv They are the weights and biases of the convolutional layer respectively;

[0348] σ is the activation function;

[0349] The mathematical expression of the recurrent neural network processing of non-image features is:

[0350] f non-image =σ(W fc ·X non-image +b fc )

[0351] in,

[0352] X non-image Includes non-image data such as soil type, historical climate data, and crop requirements;

[0353] W fc and b fc are the weights and biases of the fully connected layer;

[0354] The mathematical expression of feature fusion is:

[0355] f combined =Concat(f image ,f non-image )

[0356] in,

[0357] Concat represents the concatenation of feature vectors;

[0358] For planting density, irrigation, and fertilization, there are three independent output layers, each of which is responsible for predicting the corresponding agricultural parameters. For planting density, the mathematical expression of the output layer is:

[0359]

[0360] For irrigation, the mathematical expression of the output layer is:

[0361]

[0362] For the fertilization scheme, the mathematical expression of the output layer is:

[0363]

[0364] in,

[0365] W density , W irrigation and W fertilization are the weights of each output layer;

[0366] b density , b irrigation and b fertilization are the biases of each output layer.

[0367] It is worth noting that since planting density, irrigation and fertilization schemes are usually continuous values, we do not need to use a softmax function to obtain a probability distribution, but instead use a linear activation function or a slight modification such as ReLU+, depending on whether the output needs to be non-negative.

[0368] In practical applications, the predicted continuous values ​​may also need to be further converted into a form that farmers can understand, for example, recommendations such as "medium density", "irrigate twice a week" or "high nitrogen and low phosphorus" convert continuous values ​​into discrete categories such as "low density", "medium density", "high density". In addition, the predicted irrigation and fertilization schemes may need to be converted into specific recommendations, such as "irrigate twice a week, each irrigation volume is X liters / m2" or "apply Y kg of nitrogen fertilizer per hectare".

[0369] In addition, in order to optimize the model, we need a loss function to measure the gap between the predicted value and the true value. For the prediction of continuous values, the commonly used loss function is Mean Squared Error (MSE) or Root Mean Squared Error (RMSE).

[0370] Ultimately, the model will go through a back-propagation algorithm and an optimizer such as Adam or SGD to update the weights and biases to minimize the loss function.

[0371] Integrate IoT sensors and AI image recognition technology to monitor crop growth and signs of pests and diseases in real time, and provide prevention and treatment recommendations through big data analysis;

[0372] Deploy drones and ground sensor networks to continuously monitor key indicators of soil moisture and crop growth status;

[0373] Feed monitoring data into the algorithm model in real time to optimize farming strategies and achieve precise irrigation, fertilization, and pest and disease management.

[0374] Integrating IoT sensors and AI image recognition technology to monitor crop growth and signs of pests and diseases in real time is a key link in the refined management of modern agriculture. This process involves the collaborative work of multiple levels of technology, aiming to provide prevention and treatment recommendations through big data analysis to achieve continuous monitoring of crop health and timely intervention. Here are the specific details:

[0375] IoT sensor network deployment

[0376] 1. Sensor type: Deploy various IoT sensors, including but not limited to:

[0377] Soil Moisture Sensor: Monitors soil moisture content to optimize irrigation schedules.

[0378] Meteorological sensors: measure temperature, humidity, light intensity, wind speed and direction to predict weather changes and the possibility of pest and disease outbreaks.

[0379] Crop Growth Sensor: Assess crop growth status by monitoring chlorophyll fluorescence, canopy temperature and crop height.

[0380] Pest and disease sensors: Use chemical sensors and acoustic wave detectors to detect early signs of pests and diseases.

[0381] 2. Sensor network architecture: Use low-power wide area network (LPWAN) technology, such as LoRaWAN or NB-IoT, to ensure stable transmission of sensor data to the central data platform.

[0382] 3. Data collection frequency: Set a reasonable data collection interval to balance data accuracy and energy consumption, such as collecting data once every hour or every day.

[0383] AI Image Recognition Technology Application

[0384] 1. Drone aerial photography: Use drones to take aerial photos regularly or on demand to obtain high-resolution crop images for early identification of signs of pests and diseases.

[0385] 2. Image analysis algorithms: Use deep learning models, such as convolutional neural networks (CNNs), to automatically identify signs of pests and diseases, nutrient deficiencies, or water stress in images.

[0386] 3. Image data fusion: Fuse drone images with satellite remote sensing data and ground sensor data to provide multi-angle crop health assessment.

[0387] Big data analysis and decision support

[0388] 1. Data integration and cleaning: Data from various sensors and image recognition systems are centrally stored and preprocessed to ensure data quality.

[0389] 2. Real-time monitoring platform: Develop a real-time monitoring interface to display key indicators such as crop growth status, soil moisture, and pest and disease risks.

[0390] 3. Early warning system: Set thresholds, and when abnormal conditions are detected, automatically trigger an early warning to notify farm managers.

[0391] 4. Intelligent decision support: Combined with machine learning model predictions, it provides optimization suggestions for irrigation, fertilization and pest and disease management.

[0392] Precision Agriculture Strategy Implementation

[0393] 1. Precision irrigation: Adjust irrigation plans based on soil moisture sensors and meteorological data to avoid over- or under-irrigation.

[0394] 2. Intelligent fertilization: Develop personalized fertilization plans based on crop needs and soil nutritional conditions to reduce fertilizer waste.

[0395] 3. Pest and disease management: Based on AI image recognition results, take targeted biological or chemical control measures to reduce the use of pesticides.

[0396] Dynamic optimization and feedback mechanism

[0397] 1. Model iteration: Based on monitoring data feedback, continuously optimize the machine learning model to improve the accuracy of prediction and decision support.

[0398] 2. Dynamic adjustment: Adjust farming strategies in real time based on crop growth cycles and environmental changes to ensure efficient use of resources.

[0399] 3. Effectiveness evaluation: Regularly evaluate the actual effects of agricultural strategies, including crop yield, quality, and resource consumption, to verify and improve existing models.

[0400] Through the above steps, a comprehensive intelligent agricultural management system can be built to achieve real-time monitoring, precise intervention and continuous optimization, ultimately achieving the goal of increasing crop yields, reducing production costs and protecting the ecological environment.

[0401] The technical solution of this application is aimed at the intelligent construction method of the water collection and erosion prevention arable layer of sloping farmland. By integrating hyperspectral remote sensing, machine learning, geographic information system (GIS) and Internet of Things (IoT) technology, it realizes the refined management and soil and water conservation of sloping farmland. Its beneficial effects are particularly reflected in the following aspects:

[0402] 1. Improved the accuracy and efficiency of sloping farmland management

[0403] The application of hyperspectral remote sensing technology makes the identification of soil types more accurate and can distinguish different types of soil and their properties, providing a scientific basis for subsequent crop selection, irrigation and fertilization plans.

[0404] The introduction of machine learning models makes it possible to analyze complex environmental factors. By learning from historical data, the model can predict the optimal farming pattern, including crop selection, planting density, irrigation and fertilization plans, greatly improving the scientificity and efficiency of farming decisions.

[0405] 2. Optimized water resource management and soil protection

[0406] Digital terrain models and hydrological network analysis help identify potential water flow paths and confluence points, assess soil erosion risks, and provide a basis for designing contour farming and micro-terracing, which reduces runoff velocity, increases surface water infiltration, and effectively controls soil erosion.

[0407] The implementation of precision irrigation strategies, combined with IoT sensor data, can provide timely and appropriate watering based on soil moisture and crop needs, avoiding over-irrigation, reducing water waste and soil structure damage.

[0408] 3. Strengthened pest and disease management and crop health

[0409] The integration of AI image recognition technology and IoT sensors can monitor crop growth conditions and signs of pests and diseases in real time, provide prevention and treatment recommendations through big data analysis, reduce the risk of pest and disease outbreaks, and ensure crop health.

[0410] 4. Realized dynamic optimization and intelligent decision-making

[0411] The dynamic optimization mechanism enables farming strategies to be adjusted based on real-time monitoring data of seasonal changes, crop growth cycles and environmental conditions, improving the flexibility and responsiveness of agricultural production.

[0412] Model iteration and optimization ensure that the accuracy of the prediction model improves over time, constantly adapts to new agricultural environments, and enhances the reliability of agricultural decision-making.

[0413] 5. Promoted sustainable agricultural development

[0414] Comprehensive planning and dynamic adjustment mechanism, combined with GIS-assisted planning, generate visual farming pattern planning maps, which help to formulate farming patterns that meet local terrain and soil conditions, promote the sustainability of agricultural production, and reduce negative impacts on the environment.

[0415] 6. Reduced agricultural inputs and costs

[0416] Through precise irrigation, fertilization and pest and disease management, unnecessary resource consumption is reduced, the efficiency of agricultural inputs is improved, and agricultural production costs are reduced.

[0417] 7. Improved agricultural production efficiency and crop yields

[0418] Optimized farming patterns and dynamic adjustment strategies, combined with intelligent decision support, help improve crop yield and quality, increase agricultural output, and improve agricultural economic benefits.

[0419] To sum up, the technical solution of this application realizes the construction of an intelligent method for constructing a water-collecting and erosion-proof tillage layer on sloping farmland through advanced technical means. It not only effectively controls soil and water loss, but also improves agricultural production efficiency and resource utilization, which is of great significance to promoting the sustainable development of agriculture.

Claims

1. An intelligent construction method for water collection and erosion prevention plough layer on sloping farmland, It is characterized in that Includes steps: Collect multi-source geospatial data to obtain 3D terrain information and surface features of sloping farmland, use GIS software to build digital terrain models, calculate key terrain parameters such as slope and aspect, use hyperspectral remote sensing technology and machine learning classification algorithms to analyze spectral features in RS data, identify and classify different types of soil, integrate terrain parameters and soil type information with sloping farmland boundary data, generate color-coded sloping farmland topographic maps through the GIS platform, use GIS spatial analysis to identify areas of contour farming or micro-terracing, analyze terrain features, and conduct preliminary farming pattern planning; Using machine learning models, based on historical data and field research, the trained models are used to predict the most optimized farming patterns under different terrain conditions, including crop selection, planting density, irrigation and fertilization schemes; Integrate IoT sensors and AI image recognition technology to monitor crop growth conditions and signs of pests and diseases in real time, and provide prevention and treatment recommendations through big data analysis; Deploy drones and ground sensor networks to continuously monitor key indicators of soil moisture and crop growth status; Feed monitoring data into the algorithm model in real time to optimize farming strategies and achieve precise irrigation, fertilization, and pest and disease management.

2. According to claim 1, a method for intelligently constructing a water-collecting and erosion-proof plough layer on sloping farmland, It is characterized in that The key terrain parameters of slope and aspect are calculated, including: The slope value of each pixel and the aspect value of each pixel are calculated, and the slope map is classified according to different thresholds. The area with a slope less than 5% is considered flat, 5%-10% is a gentle slope, and greater than 10% is steep; Analyze aspect maps to determine changes in sunlight timing and intensity; Combine slope and aspect data to conduct hydrological network analysis, identify potential water flow paths and confluence points, and assess soil erosion risks; The calculated slope and aspect parameters are combined with soil type information to develop appropriate tillage and management strategies for different soil types, especially in areas prone to erosion.

3. According to claim 1, a method for intelligently constructing a water-collecting and erosion-proof plough layer on sloping farmland, It is characterized in that Hyperspectral remote sensing technology and machine learning classification algorithms are used to analyze the spectral features in RS data and identify and classify different types of soil. Specifically, hyperspectral images covering the target area are obtained from the hyperspectral camera and preprocessed. The reflectance curve of the soil is extracted from the processed hyperspectral images. Soil samples are collected and analyzed in the laboratory to determine their type and key attributes, including pH value, texture, and organic matter content, as training labels for the machine learning model. A feature selection algorithm is used to select the most discriminative bands from high-dimensional spectral data. A deep neural network supervised learning algorithm is used to train the classification model based on spectral features and soil type labels. K-fold cross-validation or leave-one-out method is used to evaluate the generalization ability of the model. The best model is found by adjusting the model parameters. The trained model is applied to the hyperspectral data of the entire area to generate a soil type classification map. The soil type information is superimposed and analyzed with the calculated slope and aspect terrain parameters to ensure that each sloping farmland has complete terrain and soil properties. GIS is used to identify the areas most suitable for growing specific crops and high-risk areas prone to soil erosion.

4. According to claim 1, a method for intelligently constructing a water-collecting and anti-erosion plough layer on sloping farmland, It is characterized in that The terrain parameters, soil type information and sloping farmland boundary data are integrated, including: Create an attribute table for each sloping farmland plot, record the plot ID, area, average slope, dominant slope aspect, soil type key information, if the soil data contains additional attributes, including soil pH, organic matter content, sand / clay ratio, should also be associated; Small plots of sloping farmland are merged into larger management units, ensuring that the aggregated plots retain the statistical averages of the topography and soil properties of the original plots; Create a comprehensive sloping farmland map where plots are represented by different colors or patterns to represent different soil types and slope classes.

5. According to claim 1, a method for intelligently constructing a water-collecting and erosion-proof plough layer on sloping farmland, It is characterized in that Conduct preliminary farming pattern planning, including: Consider the effects of slope and aspect to identify areas suitable for contour farming or micro-terracing, with flat areas used for conventional farming, gently sloping areas for contour farming, and steep areas for micro-terracing; Determine soil suitability for crop cultivation based on soil characteristics and terrain conditions; Develop crop rotation plans based on soil health and erosion control needs, as well as soil improvements where needed; Based on terrain parameters, determine ridge width, furrow width and ridge height, and adjust row and plant spacing to ensure maximum crop coverage; Plant vegetation in marginal areas prone to erosion and plant trees on or around sloping farmland; Combined with IoT sensor data, precision irrigation is implemented to provide water at the right time and in the right amount based on soil moisture and crop needs.

6. The method for intelligently constructing a water-collecting and erosion-proof plough layer on sloping farmland according to claim 1, It is characterized in that The steps of using machine learning models to predict the optimal farming mode include data preparation, feature engineering, model training, prediction and optimization, and implementation and monitoring. Model training uses a neural network model to predict the optimal farming mode under different terrain conditions, including crop selection, planting density, irrigation and fertilization schemes. The neural network model is constructed as follows: the input layer receives remote sensing images and other feature data, the convolution layer is used to extract image features, including multiple convolution layers and pooling layers, the fully connected layer is used to process non-image features and integrate all features, and the output layer predicts crop selection, planting density, irrigation and fertilization schemes; the mathematical expression of the operation is: y=(x*w)+b; Where y is one of the control targets, including crop selection, planting density, irrigation and fertilization schemes, x is the input image, w is the weight matrix, and b is the bias term; The activation function is ReLU (f(x) = max(0, x)); The loss function uses cross entropy loss for classification tasks and mean square error for regression tasks.

7. The method for intelligently constructing a water-collecting and anti-erosion plough layer on sloping farmland according to claim 6, It is characterized in that The convolutional neural network model is used to predict the optimal crop selection under different terrain conditions. The prediction model is: in, is the predicted probability distribution of crop selection; X is the input data, including terrain images and soil types; W 1 and W 2 are weight matrices, corresponding to the weights of the convolutional layer and the fully connected layer respectively; b 1 and b 2 is the bias term; σ is the activation function, which is ReLU; Softmax is an activation function used for multi-classification tasks that converts the output into a probability distribution.

8. The method for intelligently constructing a water-collecting and anti-erosion plough layer on sloping farmland according to claim 6, It is characterized in that A hybrid model is used to predict the optimal planting density, irrigation and fertilization scheme. The hybrid model includes a convolutional neural network model and a recurrent neural network model. The convolutional neural network model processes image data, and the recurrent neural network model processes non-image data. Non-image data includes soil type, historical climate data, and crop requirements. The mathematical expression of the convolutional neural network model for image feature extraction is: f image =σ(W conv *X image +b conv ); in, X image is the input of terrain image; W conv and b conv They are the weights and biases of the convolutional layer respectively; σ is the activation function; The mathematical expression of the recurrent neural network processing of non-image features is: f non-image =σ(W fc ·X non-image +b fc ) among them, X non-image Includes non-image data such as soil type, historical climate data, and crop requirements; W fc and b fc are the weights and biases of the fully connected layer; The mathematical expression of feature fusion is: f combined =Concat(f image ,f non-image ) in, Concat represents the concatenation of feature vectors; For planting density, irrigation and fertilization schemes, there are three independent output layers, each of which is responsible for predicting the corresponding agricultural parameters. For planting density, the mathematical expression of the output layer is: For irrigation, the mathematical expression of the output layer is: For the fertilization scheme, the mathematical expression of the output layer is: in, W density , W irrigation and W fertilization are the weights of each output layer; b density 、b irrigation and b fertilization are the biases of each output layer.

Citation Information

Patent Citations

  • 3S technology based method for returning farmland to forests

    CN105160465A

  • Intelligent accurate water irrigation control system and method for solar greenhouse fruit and vegetable cultivation

    CN113141940A

  • Satellite-borne GNSS-R soil humidity inversion method

    CN114894819A

  • Landslide risk point classification method based on multi-modal decision fusion

    CN115761335A

  • Geographic information system (GIS)-based territorial space planning drawing method

    CN116469303A

Cited By

  • Garden nursery stock intelligent management method based on stereoscopic perception

    CN119228069A

  • Sky-ground net integrated high-standard farmland intelligent monitoring management system

    CN119443870A

  • Method for predicting maize hybrid expression based on multi-environment data

    CN119516393A

  • Wild vegetable planting method capable of reducing cost and adapting to multiple environments

    CN119655122A

  • A land data identification method and system based on IoT drone

    CN119785251A