Crop multi-level drought monitoring method based on multi-source data fusion

Abstract

The invention discloses a crop multi-level drought monitoring method based on multi-source data fusion. The method comprises the following steps: step 1, establishing a soil-canopy-air three-level drought diagnosis architecture: performing soil layer monitoring by using a multi-layer embedded FDR / TDR sensor; a visible light-multispectral-thermal infrared three-in-one crop growth monitoring device is used for canopy monitoring; based on a four-element weather station, an air temperature and humidity sensor, an anemorumbometer and a net radiation sensor are used for monitoring an air layer; step 2, canopy and plant phenotype measurement and inversion; and step 3, constructing a multi-level model of the fused phenotype. By integrating multi-dimensional information such as soil moisture content, canopy phenotype and environmental meteorology, comprehensive drought diagnosis from soil moisture and canopy evapotranspiration to crop physiology is realized, and a more accurate and timely scientific basis is provided for agricultural drought resistance decision.

Description

Technical Field

[0001] This invention belongs to the field of smart agriculture technology and relates to a crop drought monitoring method, specifically a multi-level diagnostic method for crop drought status that integrates multi-source sensor data from the field. Background Technology

[0002] As global climate change leads to a continuous increase in the frequency and intensity of droughts, effective monitoring methods for agricultural droughts are of great significance for the development of precision agriculture.

[0003] Traditional drought monitoring methods relying on soil moisture content suffer from significant time lag and spatial limitations, failing to reflect the true water stress status of crops in a timely manner. This is especially problematic when there are differences in root water uptake capacity or uneven localized water distribution, potentially leading to misjudgments based solely on soil data. In contrast, crop phenotypic monitoring can directly capture dynamic physiological changes in plants, such as indicators like canopy temperature increases, changes in spectral reflectance, and differences in crop growth rates, providing earlier and more sensitive indications of drought stress. However, single phenotypic parameters are susceptible to environmental interference and lack synergistic analysis with soil and meteorological data, limiting their practical application value. Summary of the Invention

[0004] This invention provides a multi-level crop drought monitoring method based on multi-source data fusion, which combines soil moisture monitoring, meteorological-phenotypic joint inversion and mechanism model simulation to achieve accurate crop drought assessment.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A multi-level crop drought monitoring method based on multi-source data fusion includes the following steps:

[0007] Step 1: Establish a three-tiered drought diagnostic framework: soil-canopy-air.

[0008] Step 1: Soil layer monitoring using multi-layer buried FDR / TDR sensors;

[0009] Steps 1 and 2: Canopy monitoring using RGB sensors, multispectral sensors, and thermal infrared sensors;

[0010] Step 13: Relying on the four-element meteorological station, use air temperature and humidity sensors, wind speed and direction instruments, and net radiation sensors to monitor the air layer;

[0011] Step 2: Canopy and plant phenotypic measurement and inversion:

[0012] Step 2.1 Calculation of Normalized Difference Vegetation Index (NDVI):

[0013]

[0014] In the formula, It is in the near-infrared band. Red band;

[0015] Step 22: Crop Water Deficiency Index (CWSI) Measurement:

[0016] Using the NDVI value as a threshold, soil background is removed, and the remaining pixels are used as the effective canopy temperature. Simultaneously calculate CWSI, using the following formula:

[0017]

[0018] In the formula, To fully irrigate the canopy temperature, The temperature of the canopy during complete drought;

[0019] Steps 2 and 3: Calculation of Leaf Area Index (LAI):

[0020] The leaf area index is calculated based on statistical regression of the normalized vegetation index, using the following formula:

[0021]

[0022] In the formula, the coefficients , , Through ground measurement Calibration;

[0023] Step 24: Plant height measurement:

[0024] By analyzing depth information from vertically captured crop images using the DepthAnythingV2 model, the depth difference between the top of the plant and the ground is calculated as the plant height.

[0025] Step 25: Canopy Coverage (CC) Measurement

[0026] Using SAM to segment plant masks in images, the proportion of canopy pixels is statistically analyzed using the following formula:

[0027]

[0028] Step 26: Blade curl measurement using LCD:

[0029] Based on crop images obtained by shooting, YOLOv8x is used to segment leaves, extract contour detection curl feature points, calculate leaf curl degree based on curvature change and key point distribution, and classify curl level according to detection point threshold;

[0030] Step 3: Constructing a multi-level model that integrates phenotypes:

[0031] Step 3: 1. Construct a soil moisture diagnostic model: The soil moisture diagnostic model is based on real-time monitored soil moisture content data and combined with soil moisture characteristic curves to calculate the effective soil moisture content and drought index, which are used to assess the degree of water stress in the root zone; when the soil moisture content exceeds the threshold, a primary drought warning is triggered, providing basic water status input for subsequent canopy and crop models.

[0032] Step 3.2: Construct a comprehensive crop drought monitoring index (ADI) that includes soil moisture, atmospheric vapor pressure deficit (VPD), and leaf curling. By assigning different weights to the three indicators, calculate the ADI value and determine different drought thresholds for the ADI using previously obtained data. The specific calculation formula for ADI is as follows:

[0033]

[0034] Wherein, SWC represents soil moisture, VPD represents atmospheric vapor pressure deficit, and LCD represents leaf curl. These are different weighting indices for the three;

[0035] Step 3: Constructing a radiative transfer model: Based on ProSail, a forward simulation is constructed. By inputting chlorophyll content, equivalent water thickness, and leaf area index, the reflectance of the canopy in different wavelength bands is simulated. An optimization algorithm is used to iteratively adjust the input parameters to achieve the best match between the simulated reflectance and the measured multispectral data, thereby retrieving the chlorophyll content, equivalent water thickness, and leaf area index.

[0036] Steps 3 and 4: Constructing a dual-source energy balance model: The dual-source energy balance model decomposes evapotranspiration into vegetation transpiration and soil evaporation by fusing thermal imaging canopy temperature and meteorological data, thus quantifying the actual water consumption of crops; combined with crop canopy parameters retrieved by multispectral inversion, stomatal impedance parameters are dynamically calibrated, and high spatiotemporal resolution evapotranspiration is output to identify field water heterogeneity.

[0037] Step 35: Constructing the AquaCrop crop model: The AquaCrop crop model is driven by soil moisture data and actual evapotranspiration output from the dual-source energy balance model to simulate crop growth dynamics under drought stress; the parameters of the AquaCrop crop model are calibrated in real time through phenotypic data to predict biomass accumulation and yield loss, and to generate irrigation optimization schemes, realizing the mechanism link from water stress to yield impact.

[0038] Compared with the prior art, the present invention has the following advantages:

[0039] This invention integrates multi-dimensional information such as soil moisture, canopy phenotype, and environmental meteorology to achieve comprehensive drought diagnosis from soil moisture and canopy evapotranspiration to crop physiology, providing more accurate and timely scientific basis for agricultural drought resistance decision-making. Attached Figure Description

[0040] Figure 1 A flowchart of a multi-level crop drought monitoring method based on multi-source data fusion;

[0041] Figure 2 Design of a prototype for crop growth monitoring equipment;

[0042] Figure 3 for Figure 2 A magnified view of middle A;

[0043] Figure 4 for Figure 2 Enlarged view of C;

[0044] Figure 5 This is the result of automatic plant height calculation;

[0045] Figure 6 The results are canopy segmentation based on SAM.

[0046] Figure 7 This is a display of time-series data acquired from crop growth monitoring equipment;

[0047] Figure 8 The temporal changes in crop height and canopy coverage are automatically calculated based on images and models.

[0048] Figure 9 The process for calculating blade curl;

[0049] Figure 10 The result is the calculated value of the blade curl.

[0050] Figure 11 This is a crop simulation based on a crop model with multi-gradient irrigation. Detailed Implementation

[0051] The technical solution of the present invention will be further described below with reference to the accompanying drawings, but it is not limited thereto. Any modifications or equivalent substitutions to the technical solution of the present invention that do not depart from the spirit and scope of the technical solution of the present invention should be covered within the protection scope of the present invention.

[0052] This invention provides a multi-level crop drought monitoring method based on multi-source data fusion, such as... Figure 1 As shown, the method includes the following steps:

[0053] Step 1: Establish a three-tiered drought diagnostic framework: soil-canopy-air.

[0054] Step 1: Soil Layer Monitoring Equipment and Installation: Soil layer monitoring utilizes multi-layer buried FDR / TDR sensors, installed at 0-20cm (shallow layer), 20-50cm (root active layer), and 50-100cm (deep layer) respectively, to monitor soil volumetric water content (VWC) and temperature changes in real time. During installation, ensure the probes are in close contact with the soil, avoiding areas directly eroded by irrigation sprinklers. After installation, allow the soil to settle naturally for 48 hours before calibration to ensure data accuracy.

[0055] Steps one and two: Canopy monitoring is performed using a three-in-one crop growth monitoring device combining visible light, multispectral imaging, and thermal infrared imaging. This device can be mounted on a height-adjustable (1.5~3m) support to ensure vertical coverage of the target crop. Multispectral bands (450~900nm) are used to retrieve various vegetation indices, while thermal imaging (8~14μm) captures canopy temperature distribution to identify water stress. The time-series data acquired by the crop growth monitoring device are as follows: Figure 7 As shown.

[0056] like Figures 2-4 As shown, the visible light-multispectral-thermal infrared integrated crop growth monitoring device includes the device's internal and external design, sensor modules, data acquisition and transmission modules, data processing and analysis modules, and power supply modules. Each module has a clearly defined function and works collaboratively to achieve high-precision monitoring of crop growth status in the greenhouse.

[0057] 1. Internal and external design of the equipment

[0058] The design of the equipment must balance the rationality of the internal module layout with the durability of the external structure. The internal design emphasizes modular layout to ensure efficient collaboration among various functional components; the external design takes into account the characteristics of the greenhouse environment, adopts dustproof and moisture-proof materials, and enhances the equipment's resistance to environmental interference to adapt to complex environments such as high temperature and high humidity.

[0059] 2. Sensor Module

[0060] The sensor module serves as the data source for the entire system, integrating RGB sensors, multispectral sensors, and thermal infrared sensors. This module is responsible for collecting multimodal data from crops, providing multimodal and comprehensive monitoring data for subsequent data processing.

[0061] 3. Data Acquisition and Transmission Module

[0062] The data acquisition and transmission module is responsible for quickly and stably transmitting the data collected by the sensors to the central processing unit or a remote server. This module supports both wired and wireless transmission methods, ensuring data transmission stability in different scenarios and providing a guarantee for real-time monitoring.

[0063] 4. Data Processing and Analysis Module

[0064] The data processing and analysis module is the core of the system, primarily used for the fusion, processing, and analysis of multimodal data. It extracts and analyzes crop phenotypic parameters through plant height analysis algorithms (based on Depth Amything), SAM-based cover analysis algorithms, and leaf curl analysis algorithms, thereby enhancing the application value of the data. Simultaneously, this module supports storing analysis results in the cloud for further visualization and management.

[0065] 5. Power supply module

[0066] The power supply module provides continuous and stable power support for the equipment and is designed to be compatible with multiple power supply methods, including solar power and backup batteries, to ensure the long-term operation of the equipment in a solar greenhouse environment.

[0067] Step 13: Atmospheric layer monitoring relies on a four-element meteorological station, including air temperature and humidity sensors, anemometers, and net radiation sensors. These are installed 2 meters above the canopy, with the height adjusted as the crop grows. The anemometers are kept at least three times the length of the support pole to avoid turbulence, and the net radiation sensor's mirror is cleaned regularly to reduce errors. Data is uploaded to a cloud platform in real time for calculating evapotranspiration (ETO) and vapor pressure deficit (VPD), providing environmental driving factors for drought diagnosis.

[0068] Step 2: Canopy and plant phenotypic measurement and inversion:

[0069] Step 21: Calculation of Normalized Difference Vegetation Index (NDVI):

[0070]

[0071] In the formula, It is in the near-infrared band. It is the red band.

[0072] Step 22: Crop Water Deficiency Index (CWSI) Measurement:

[0073] Using an NDVI value of 0.2 as a threshold, soil background was removed, and the remaining pixels were used as the effective canopy temperature. It can also calculate CWSI, as follows:

[0074]

[0075] In the formula, To fully irrigate the canopy temperature, The temperature of the canopy is at the point of complete drought.

[0076] Steps 2 and 3: Leaf Area Index (LAI) Calculation:

[0077] Based on the normalized difference vegetation index (NDVI) statistical regression, the leaf area index (LAI) is calculated using the following formula:

[0078]

[0079] In the formula, the coefficients , , Ground-based measurements are required. (e.g., LAI-2200 canopy analyzer) Calibration.

[0080] Step 24: Plant Height (PH) Measurement:

[0081] By analyzing depth information from vertically captured crop images using the DepthAnythingV2 model, the depth difference between the top of the plant and the ground is calculated as the plant height. This model is a large-scale, open-weighted depth estimation model that can estimate the relative or absolute depth map corresponding to the input RGB image, and indirectly estimate the plant height based on the depth map. The automatic plant height calculation result is shown below. Figure 3 As shown.

[0082] Step 25: Canopy Coverage (CC) Measurement:

[0083] Using SAM to segment the plant mask in the image, the proportion of canopy pixels is calculated using the following formula:

[0084]

[0085] Canopy segmentation results based on SAM are as follows Figure 5 As shown in the figure. The temporal variations of crop height and canopy coverage, automatically calculated based on images and models, are as follows: Figure 8 As shown.

[0086] Step 26: Leaf Curling Degree (LCD) Measurement:

[0087] Based on crop images obtained through photography, leaves are segmented using YOLOv8x, and contour detection of curled feature points is performed. Leaf curl degree is calculated based on curvature changes and key point distribution, and curl levels are classified according to detection point thresholds. Let N be the number of detected corner points: N < 1 indicates mild stress, 1 ≤ N < 5 indicates moderate stress, and N ≥ 5 indicates severe stress. The leaf curl degree calculation process is as follows: Figure 9 As shown, the calculation results of the blade curl are as follows: Figure 10 As shown.

[0088] like Figure 9As shown, a dedicated dataset for maize leaf segmentation was constructed based on phenotypic images of maize leaf curling under drought stress. A high-precision leaf segmentation model was trained using the YOLOv8x algorithm, and leaf region masks were obtained through binarization processing. Based on this, corner detection algorithms were used to quantify the corner features of individual leaves, and a graded evaluation system for the degree of leaf curling was established based on a threshold for the number of corner points.

[0089] Step 3: Constructing a multi-level model that integrates phenotypes:

[0090] Step 3.1 Soil Moisture Diagnostic Model: Based on real-time monitored soil moisture content data and combined with soil moisture characteristic curves, the system calculates the available soil water content (AWC) and drought index (SDI) to assess the degree of water stress in the root zone. When the soil moisture content exceeds a threshold (e.g., <60%), the system triggers a primary drought warning, providing basic moisture status input for subsequent canopy and crop models.

[0091] Step 3.2: Construct a comprehensive crop drought monitoring index (ADI) that includes soil moisture, atmospheric vapor pressure deficit (VPD), and leaf curling. Calculate the ADI value by assigning different weights to the three indicators. Use previously obtained data to determine different drought thresholds for the ADI.

[0092] Step 3: Radiative Transfer Model: Based on ProSail, a forward simulation is constructed. By inputting parameters such as chlorophyll content (Cab), equivalent water thickness (Cw), and leaf area index (LAI), the reflectivity of the canopy in different wavelength bands is simulated. An optimization algorithm (such as the least squares method) is used to iteratively adjust the input parameters to achieve the best match between the simulated reflectivity and the measured multispectral data, thereby retrieving key parameters such as LAI, Cab, and Cw.

[0093] Steps 3 and 4: Dual-Source Energy Balance Model: By fusing thermal imaging canopy temperature and meteorological data, evapotranspiration (ET) is decomposed into vegetation transpiration and soil evaporation, quantifying actual crop water consumption. Combined with vegetation indices (such as NDVI) retrieved from multispectral sources and crop canopy parameters such as LAI, stomatal impedance parameters are dynamically calibrated to output high spatiotemporal resolution evapotranspiration for identifying field water heterogeneity.

[0094] Step 3.5, AquaCrop Crop Model: Driven by soil moisture data and actual ET output from the dual-source energy balance model, this step simulates crop growth dynamics under drought stress. AquaCrop crop model parameters are calibrated in real-time using phenotypic data (such as LAI and canopy cover) to predict biomass accumulation and yield loss, and to generate optimized irrigation schemes, thus establishing a mechanistic link between water stress and yield impact. Multi-gradient irrigation simulation based on the crop model is shown below. Figure 11 As shown.

Claims

1. A multi-level crop drought monitoring method based on multi-source data fusion, characterized in that... The method includes the following steps: Step 1: Establish a three-tiered drought diagnostic framework: soil-canopy-air. Step 1: Soil layer monitoring using multi-layer buried FDR / TDR sensors; Steps 1 and 2: Canopy monitoring using RGB sensors, multispectral sensors, and thermal infrared sensors; Step 13: Relying on the four-element meteorological station, use air temperature and humidity sensors, wind speed and direction instruments, and net radiation sensors to monitor the air layer; Step 2: Canopy and plant phenotypic measurement and inversion: Step 2: Calculate the Normalized Difference Vegetation Index (NDVI); Step 22: Crop Water Deficiency Index (CWSI) Measurement: Using the NDVI value as a threshold, soil background is removed, and the remaining pixels are used as the effective canopy temperature. Simultaneously calculate CWSI; Steps 2 and 3: Calculation of Leaf Area Index (LAI): Leaf area index was calculated based on statistical regression of normalized vegetation index. Step 24: Plant height measurement: By analyzing depth information from vertically captured crop images using the DepthAnythingV2 model, the depth difference between the top of the plant and the ground is calculated as the plant height. Step 25: Canopy Coverage (CC) Measurement: Using SAM to segment plant masks in images, the percentage of canopy pixels is statistically analyzed. Step 26: Blade curl measurement using LCD: Based on crop images obtained by shooting, YOLOv8x is used to segment leaves, extract contour detection curl feature points, calculate leaf curl degree based on curvature change and key point distribution, and classify curl level according to detection point threshold; Step 3: Constructing a multi-level model that integrates phenotypes: Step 3:

1. Construct a soil moisture diagnostic model: The soil moisture diagnostic model is based on real-time monitored soil moisture content data and combined with soil moisture characteristic curves to calculate the effective soil moisture content and drought index, which are used to assess the degree of water stress in the root zone; when the soil moisture content exceeds the threshold, a primary drought warning is triggered, providing basic water status input for subsequent canopy and crop models. Step 3.2: Construct a comprehensive crop drought monitoring index (ADI) that includes soil moisture, atmospheric vapor pressure deficit (VPD), and leaf curling. Calculate the ADI value by assigning different weights to the three indicators. Determine different drought thresholds for ADI using previously obtained data. Step 3: Constructing a radiative transfer model: Based on ProSail, a forward simulation is constructed. By inputting chlorophyll content, equivalent water thickness, and leaf area index, the reflectance of the canopy in different wavelength bands is simulated. An optimization algorithm is used to iteratively adjust the input parameters to achieve the best match between the simulated reflectance and the measured multispectral data, thereby retrieving the chlorophyll content, equivalent water thickness, and leaf area index. Steps 3 and 4: Constructing a dual-source energy balance model: The dual-source energy balance model decomposes evapotranspiration into vegetation transpiration and soil evaporation by fusing thermal imaging canopy temperature and meteorological data, thus quantifying the actual water consumption of crops; combined with crop canopy parameters retrieved by multispectral inversion, stomatal impedance parameters are dynamically calibrated, and high spatiotemporal resolution evapotranspiration is output to identify field water heterogeneity. Step 35: Constructing the AquaCrop crop model: The AquaCrop crop model is driven by soil moisture data and actual evapotranspiration output from the dual-source energy balance model to simulate crop growth dynamics under drought stress; the parameters of the AquaCrop crop model are calibrated in real time through phenotypic data to predict biomass accumulation and yield loss, and to generate irrigation optimization schemes, realizing the mechanism link from water stress to yield impact.

2. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In the above, multi-layer buried FDR / TDR sensors are installed at 0~20cm, 20~50cm and 50~100cm respectively to monitor soil volumetric water content and temperature changes in real time.

3. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In the first and second scenarios, canopy monitoring is performed by a visible light-multispectral-thermal infrared integrated crop growth monitoring device. This device includes a sensor module, a data acquisition and transmission module, a data processing and analysis module, and a power supply module. The sensor module integrates an RGB sensor, a multispectral sensor, and a thermal infrared sensor, and is responsible for collecting multimodal data of crops; The data acquisition and transmission module is responsible for transmitting the data acquired by the sensors to the central processing unit or a remote server. The data processing and analysis module is used for the fusion, processing and parsing of multimodal data, and supports storing the analysis results in the cloud.

4. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In the first and third scenarios, the air temperature and humidity sensor and the net radiation sensor are installed 2m above the canopy, with the height adjusted according to crop growth; the anemometer is kept at least three times the distance from the support pole to avoid turbulence.

5. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In step two, the formula for calculating the Normalized Difference Vegetation Index (NDVI) is as follows: In the formula, It is in the near-infrared band. It is the red band.

6. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In step two, the formula for calculating CWSI is: In the formula, To fully irrigate the canopy temperature, The temperature of the canopy is at the point of complete drought.

7. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In steps two and three, the formula for calculating the leaf area index is as follows: In the formula, the coefficients , , Through ground measurement Calibration.

8. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In step two, the formula for calculating canopy coverage (CC) is as follows: 。 9. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In step two six, let N be the number of detected corner points, N < 1 indicates mild stress, 1 ≤ N < 5 indicates moderate stress, and N ≥ 5 indicates severe stress.

10. The crop multi-level drought monitoring method based on multi-source data fusion according to claim 1, characterized in that... In step 33, the ADI calculation formula is as follows: Wherein, SWC represents soil moisture, VPD represents atmospheric vapor pressure deficit, and LCD represents leaf curl. These are different weighting indices for the three.

Images