Grassland drought stress monitoring device and method based on unmanned aerial vehicle multi-source sensor

Abstract

The invention relates to the technical field of unmanned aerial vehicle remote sensing monitoring, and discloses a grassland drought stress monitoring device and method based on an unmanned aerial vehicle multi-source sensor, and the device comprises an unmanned aerial vehicle platform, a multi-source sensor module, a data processing module and a ground workstation. The method corresponds to the device. According to the grassland drought stress monitoring device and method based on the unmanned aerial vehicle multi-source sensor, through space-time registration of the spectrum camera, the LiDAR, the thermal imaging lens and the soil moisture sensor and through combination of RTK sub-meter positioning compensation geometric deformation, the accuracy and timeliness of grassland drought monitoring are ensured.

Description

Technical Field

[0001] The present application relates to the field of UAV remote sensing monitoring technology, and specifically to a grassland drought stress monitoring device and method based on UAV multi-source sensors. Background Art

[0002] In existing technologies, grassland drought monitoring usually relies on a single sensor (such as thermal imaging or multispectral cameras), which cannot effectively integrate three-dimensional point clouds, soil moisture and meteorological parameters. For example, traditional methods calculate CWSI (canopy water stress index) through NDVI and land surface temperature (LST), but do not consider the impact of vegetation height changes on canopy structure, resulting in a drought level misjudgment rate of >15% (such as NDVI distortion in sandstorms). Humidity (RH>80%) causes condensation on the thermal imaging lens, resulting in LST measurement errors of >2°C; sandstorms (visibility <50m) cause the LiDAR point cloud density to drop by >50%, and the vegetation height inversion error is >15%. Therefore, grassland drought monitoring in existing technologies has problems with insufficient reliability of monitoring results and delayed response. Summary of the Invention

[0003] The purpose of this application is to provide a grassland drought stress monitoring device and method based on unmanned aerial vehicle multi-source sensors to solve the technical problems raised in the above background technology.

[0004] To achieve the above objectives, this application discloses the following technical solutions:

[0005] In a first aspect, the present application discloses a grassland drought stress monitoring device based on a multi-source sensor of an unmanned aerial vehicle, comprising:

[0006] UAV platform: equipped with a multi-rotor UAV or a fixed-wing UAV, the multi-rotor UAV and the fixed-wing UAV are equipped with an RTK module for sub-meter positioning and a multi-sensor collaborative operation gimbal;

[0007] Multi-source sensor module: includes: a spectral camera, a thermal imaging lens, a lidar, a soil moisture sensor, and an IMU inertial navigation module. The spectral camera covers visible light, near-infrared, and short-wave infrared bands. The thermal imaging lens is used to monitor surface temperature. The lidar is used to obtain three-dimensional point cloud data. The soil moisture sensor is integrated into the mounting equipment of the multi-rotor drone and the fixed-wing drone, and obtains soil surface moisture content and temperature through hovering or landing measurements.

[0008] Data processing module: used for output transmission and processing, the data processing module includes:

[0009] Data transmission unit: transmits sensor raw data to the ground workstation in real time through wireless transmission;

[0010] Flight attitude calculation unit: Based on the IMU inertial navigation module and RTK data, it compensates for the geometric deformation and point cloud distortion caused by the movement of the UAV sensor;

[0011] Multi-source data spatiotemporal registration unit: performs spatial consistency processing on sensor raw data through timestamp synchronization and georeferencing;

[0012] Drought stress collaborative analysis unit: integrates spectral index, thermal infrared data, point cloud features and soil moisture parameters to obtain grassland drought stress conditions;

[0013] Dynamic calibration unit: corrects sensor performance deviations through real-time environmental parameters and compensates for measurement errors under complex meteorological conditions through redundant sensor data;

[0014] The device enables each sensor in the multi-source sensor module to be plug-and-play through a standardized interface protocol, and the data transmission unit adopts an edge compression algorithm for data transmission.

[0015] Preferably, the synchronous trigger frequency of the spectral camera and the thermal imaging lens is satisfy The point cloud density ρ of the laser radar satisfies ρ ≥ 1000 points / m 3 , and the soil moisture sensor performs in-situ measurement of surface soil moisture content through electromagnetic induction or neutron scattering method when hovering.

[0016] Preferably, the data processing module adopts a spatiotemporal registration algorithm, and the spatiotemporal registration algorithm specifically includes:

[0017] Based on IMU inertial data and combined with RTK positioning coordinates, the timestamps of multi-sensor raw data are synchronized;

[0018] Build a 3D terrain model using LiDAR point clouds and correct geometric projection errors of multispectral and thermal imaging data;

[0019] Weather station data are introduced to correct the radiation error of thermal infrared data.

[0020] Preferably, the data processing module suppresses meteorological interference by an adaptive filtering algorithm, wherein the adaptive filtering algorithm comprises:

[0021] When the relative humidity RH satisfies RH>80%, Kalman filtering is enabled for the thermal imaging data;

[0022] When the visibility VB satisfies VB<50 meters, the point cloud density is corrected by fusing LiDAR point cloud data with millimeter-wave radar data.

[0023] Preferably, the dynamic calibration unit is configured with an environment adaptive compensation algorithm, and the environment adaptive compensation algorithm specifically includes:

[0024] The operating temperature of the sensor is controlled within a range of 15-35° C. by a PID algorithm, and the housing of each sensor in the multi-source sensor module is encapsulated with a polyimide composite material.

[0025] Preferably, the control of the sensor operating temperature at 15-35°C by the PID algorithm is implemented based on a constant temperature control module, and the constant temperature control module includes a three-level temperature protection mechanism, which specifically includes:

[0026] Level 1: PTC heating plate at temperature T a Start when <10℃, heating power P h ≤5W;

[0027] Second level: semiconductor refrigeration chip at temperature T a >38℃, start, cooling power P c ≤10W;

[0028] Level 3: When the temperature changes by more than ±5°C, the sensor is triggered to temporarily sleep and automatically restart.

[0029] Preferably, the drought stress collaborative analysis unit obtains the water stress index CWSI through a comprehensive stress index algorithm; the formula of the stress index algorithm is:

[0030]

[0031] Where Tc is the canopy temperature; Tc max and Tc min are the theoretical maximum and minimum values of canopy temperature respectively; θ v is the volumetric water content of soil; θ v0 is the historical background soil moisture content; H is the vegetation height; H0 is the benchmark vegetation height.

[0032] Preferably, the drought stress collaborative analysis unit obtains the difference water index NDWI by a difference water index algorithm, and the difference water index algorithm is specifically:

[0033] D1: Calculate the initial moisture index NDWI0 using the initial moisture index calculation formula. The initial moisture index calculation formula is:

[0034] Among them, R Red-Edge is the red edge band reflectivity; R SWIR is the reflectivity in the shortwave infrared band;

[0035] D2: Based on the dynamic calibration unit, the ambient temperature T is corrected. The correction formula is: T compensation =1+0.01×(T-25°C), where 15°C≤T≤35°C;

[0036] D3: Calculate the difference moisture index NDWI, NDWI = NDWI0 × T compensation .

[0037] Preferably, the data processing module further comprises a dynamic warning unit for classifying drought stress and performing dynamic warning based on a historical drought database and real-time monitoring data;

[0038] The drought stress classification specifically includes:

[0039] S11-Classification model construction: Random forest, support vector machine and deep learning algorithm are used to build a drought stress classification model based on input features, including water stress index CWSI, difference water index NDWI, soil moisture gradient Vegetation height change rate Meteorological factors;

[0040] S12-Classification results: When NDWI≤0.3 and CWSI≤0.2, the grassland drought stress is classified as healthy grassland; when 0.3<NDWI≤0.6 or 0.2<CWSI≤0.4, the grassland drought stress is classified as mild drought; when NDWI>0.6 or CWSI>0.4, the grassland drought stress is classified as severe drought;

[0041] The dynamic early warning based on the historical drought database and real-time monitoring data specifically includes:

[0042] S21- Input early warning features, including water stress index CWSI, difference water index NDWI, soil moisture gradient Vegetation height change rate Meteorological factors and grassland drought stress classification results;

[0043] S22-Use the sliding window method to analyze the water stress index CWSI or difference water index NDWI in the last 30 days and calculate the drought trend coefficient C trend , Among them, D 当前值 is the current value of water stress index CWSI or difference water index NDWI, D 历史均值 is the average value of water stress index CWSI or difference water index NDWI in the past year, D 历史标准差 is the standard deviation of the water stress index CWSI or the difference water index NDWI in the past year;

[0044] S23-Set the benchmark threshold based on the grassland drought stress classification results, combined with the drought trend coefficient C trend Dynamically modify the warning threshold, the correction result is: warning threshold = reference threshold × (1 + k × C trend ), k is the sensitivity coefficient;

[0045] S24-When the real-time water stress index CWSI or difference water index NDWI exceeds the dynamic threshold, a drought warning is triggered.

[0046] In a second aspect, the present application discloses a grassland drought stress monitoring method based on a multi-source sensor of an unmanned aerial vehicle, which is applied to the grassland drought stress monitoring device based on a multi-source sensor of an unmanned aerial vehicle as described above. The method comprises the following steps in sequence:

[0047] Setting up a drone platform, wherein the drone platform is configured with a multi-rotor drone or a fixed-wing drone, wherein the multi-rotor drone and the fixed-wing drone are equipped with an RTK module for sub-meter positioning and a multi-sensor collaborative operation gimbal;

[0048] Controlling the multi-sensor gimbal to synchronously collect raw sensor data from a spectral camera, thermal imaging lens, lidar, soil moisture sensor, and IMU inertial navigation module, wherein the spectral camera covers visible light, near-infrared, and short-wave infrared bands, the thermal imaging lens is used to monitor ground surface temperature, the lidar is used to obtain three-dimensional point cloud data, and the soil moisture sensor is integrated into the mounting equipment of the multi-rotor drone and the fixed-wing drone to obtain soil surface moisture content and temperature through hovering or landing measurements;

[0049] The IMU inertial navigation module and RTK data compensate for the geometric deformation and point cloud distortion caused by the movement of the drone;

[0050] Spatial consistency processing of sensor raw data through timestamp synchronization and georeferencing;

[0051] Correct sensor performance deviations through real-time environmental parameters and compensate for measurement errors under complex meteorological conditions through redundant sensor data;

[0052] Transmit sensor raw data to the ground workstation in real time via wireless transmission;

[0053] The ground workstation fuses spectral index, thermal infrared data, point cloud features and soil moisture parameters to obtain grassland drought stress conditions.

[0054] Beneficial effects: The grassland drought stress monitoring device and method based on drone multi-source sensors of the present application ensure the accuracy and timeliness of grassland drought monitoring through the spatiotemporal alignment of spectral cameras, LiDAR, thermal imaging lenses, and soil moisture sensors, combined with RTK sub-meter positioning to compensate for geometric deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 This is a structural block diagram of a grassland drought stress monitoring device based on a drone multi-source sensor provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] The following is a clear and complete description of the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0058] In this document, the term "comprising" is intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0059] In a first aspect, this embodiment provides a Figure 1 The grassland drought stress monitoring device shown here, based on a multi-source sensor from a drone, achieves high-precision monitoring of grassland drought stress through multi-source sensor collaboration, dynamic calibration, and data fusion. The device includes a drone platform, a multi-source sensor module, a data processing module, and a ground workstation.

[0060] Specifically

[0061] The drone platform is equipped with a multi-rotor drone or a fixed-wing drone with long flight time and high payload capacity. These drones are equipped with RTK modules (such as DJI's RTK system, which provides ±1cm horizontal positioning accuracy) for sub-meter positioning, meeting the spatial resolution requirements of grassland drought monitoring. They are also equipped with a multi-sensor collaborative gimbal (such as the DJI Zenmuse series), which uses MEMS gyroscopes and electric motors to compensate for changes in the drone's attitude (such as pitch, roll, and yaw) in real time, ensuring sensor imaging stability. During operation, the drone platform first plans the flight path. This path is generally achieved by dividing the monitoring area according to grassland type and drought risk level.

[0062] The multi-source sensor module includes a (multispectral / hyperspectral) camera, a thermal imaging lens, a lidar (LiDAR), a soil moisture sensor, and an inertial navigation unit (IMU) (e.g., a fiber optic gyro IMU). The spectral camera covers the visible (VIS), near-infrared (NIR), and shortwave infrared (SWIR) bands and is used to extract vegetation indices (e.g., NDVI, EVI) and water sensitivity indices (e.g., CWSI). The thermal imaging lens is used to monitor land surface temperature (LST) and identify drought areas through canopy temperature anomalies (e.g., high temperatures are associated with soil moisture deficiency). The lidar is used to acquire three-dimensional point cloud data (e.g., vegetation height and canopy density) to assist in inferring biomass distribution. The soil moisture sensor is integrated into the mounting equipment of the multi-rotor drone and the fixed-wing drone, and obtains soil surface moisture content and temperature through hovering or landing measurements. During the data acquisition process, a multi-resolution dataset is generated by combining high-altitude wide-area coverage (drone cruising) with low-altitude detailed inspection of key areas (e.g., hovering to obtain soil moisture data).

[0063] Data processing module: used for output transmission and processing. The data processing module includes:

[0064] Data transmission unit: transmits sensor raw data to the ground workstation in real time via wireless transmission (such as 5G or satellite communication);

[0065] Flight attitude calculation unit: Based on the IMU inertial navigation module and RTK data, the geometric deformation and point cloud distortion of the sensor caused by the movement of the drone are compensated. In some existing technologies, the compensation for geometric deformation is specifically: the real-time attitude of the sensor (Euler angle: pitch, roll, heading) is jointly calculated by IMU data (angular velocity, acceleration) and RTK positioning coordinates; the compensation for point cloud distortion is specifically: using attitude data to correct the vertical direction error of the LiDAR point cloud caused by the pitch angle of the drone.

[0066] Multi-source data spatiotemporal registration unit: This unit processes the spatial consistency of sensor raw data through timestamp synchronization and georeferencing. For example, it uses GPS timestamps (e.g., UTC time) to align multi-sensor data (e.g., the trigger time difference between a multispectral camera and a thermal imaging lens is ≤ 5ms). It also converts the sensor coordinate system (e.g., the camera coordinate system) into a unified WGS-84 geographic coordinate system, and uses RTK positioning and LiDAR point cloud to construct a DEM (digital elevation model) to assist in alignment.

[0067] Drought stress collaborative analysis unit: integrates spectral index, thermal infrared data, point cloud features and soil moisture parameters to obtain grassland drought stress conditions; algorithms such as support vector machine (SVM) and random forest can input multi-source features into the classification model to output drought stress conditions.

[0068] Dynamic calibration unit: Corrects sensor performance deviations through real-time environmental parameters (such as temperature and humidity), and compensates for measurement errors under complex meteorological conditions through redundant sensor data; such as atmospheric correction algorithms and redundant data fusion using weighted averaging methods.

[0069] The device enables plug-and-play of each sensor in the multi-source sensor module through a standardized interface protocol (such as the ROS2 adaptation layer), and the data transmission unit adopts an edge compression algorithm (compression ratio ≥ 5:1) for data transmission to retain the original physical unit accuracy.

[0070] Traditional UAV remote sensing systems mostly use a trigger frequency below 5Hz, which results in spatiotemporal misalignment of multi-source data (e.g., the time difference between thermal imaging and spectral data is greater than 0.2 seconds), affecting the accuracy of collaborative analysis. In this embodiment, the synchronous trigger frequency of the spectral camera and the thermal imaging lens is satisfy In this way, the synchronization frequency is increased to 10Hz, making the spatiotemporal consistency error ≤0.1 second, ensuring that the temperature-spectral feature matching accuracy of thermal infrared and spectral data is improved by more than 30%, and supporting continuous monitoring under high-speed flight (such as 15m / s for fixed-wing UAVs), reducing the number of repeated flights.

[0071] Secondly, traditional low-density point clouds (e.g., 200 points / m2) can only roughly distinguish between vegetation and bare land, and cannot capture canopy structure details (e.g., layered leaf distribution). In this embodiment, the point cloud density ρ of the lidar satisfies ρ ≥ 1000 points / m 3, using pulsed LiDAR (such as Riegl VQ-1560i), by increasing the laser emission frequency (≥100kHz) and the number of scanning lines (such as 1024 lines), ≥1000 point cloud data per square meter can be achieved. In this way, vegetation height (error ≤5cm), canopy coverage (accuracy ±2%) and terrain micro-undulations (such as ±10cm height difference) can be accurately extracted. The individual height of single herbaceous plants (such as forage) can be distinguished, supporting three-dimensional canopy structure modeling (such as leaf area index LAI inversion error ≤10%), and supporting high-precision DEM construction in complex terrain (such as areas with slope >30°), thereby improving the reliability of spatial distribution analysis of drought stress.

[0072] Furthermore, traditional soil moisture sensors (such as TDR time-domain reflectometry) require fixed installation and cannot be used for in-situ measurement by drones. Mobile sensors (such as handheld sensors) are inefficient and cannot cover large areas. In this embodiment, the soil moisture sensor uses electromagnetic induction or neutron scattering while hovering to perform in-situ measurement of surface soil moisture content.

[0073] In order to achieve timestamp synchronization and geographic registration, in this embodiment, the data processing module adopts a spatiotemporal registration algorithm, which specifically includes:

[0074] Based on the IMU inertial data and combined with the RTK positioning coordinates, the timestamp synchronization of the multi-sensor raw data is performed. It is feasible that the drone RTK module provides the UTC timestamp (accuracy ±1ms), and the multi-sensor data is synchronized through the hardware trigger signal (such as the flight control GPIO) to ensure that the time alignment error is ≤5ms.

[0075] A three-dimensional terrain model is constructed using LiDAR point clouds, and the geometric projection errors of multispectral and thermal imaging data are corrected. Multi-sensor data are projected into a unified WGS-84 coordinate system using RTK positioning coordinates. A digital surface model is constructed in combination with LiDAR point clouds. Weather station data (such as wind speed and humidity) is introduced to correct the radiation error of thermal infrared data, and terrain shadows and perspective distortion of multispectral / thermal imaging data are corrected. The registration accuracy reaches sub-pixel level (≤0.05 pixel).

[0076] It is feasible that, in this embodiment, the data processing module suppresses meteorological interference by an adaptive filtering algorithm, and the adaptive filtering algorithm includes:

[0077] When the relative humidity RH satisfies RH>80%, Kalman filtering is enabled for the thermal imaging data (noise threshold ≤ 0.5℃);

[0078] When the visibility VB satisfies VB<50 meters (such as in dusty weather), the point cloud density is corrected by fusing LiDAR point cloud data with millimeter-wave radar data (error ≤ 5%).

[0079] Traditional drone sensors rely on passive heat dissipation and cannot cope with extreme environments (such as desert areas with temperatures greater than 50°C or high-altitude cold areas with temperatures less than -20°C), resulting in an increase in dark current of the CMOS sensor (noise increase by 30%). To achieve environmental adaptive compensation, the dynamic calibration unit is equipped with an environmental adaptive compensation algorithm, which specifically includes:

[0080] The sensor operating temperature is controlled at 15-35°C by PID algorithm, and the housing of each sensor in the multi-source sensor module is encapsulated with polyimide composite material, whose thermal expansion coefficient is ≤5×10 -6 / ℃. It should be noted that the PID algorithm is a prior art. The PID algorithm is used to control the sensor operating temperature at 15-35℃ based on a constant temperature control module. The constant temperature control module includes a three-level temperature protection mechanism, which specifically includes:

[0081] Level 1: PTC heating plate at temperature T a Start when <10℃, heating power P h ≤5W;

[0082] Second level: semiconductor refrigeration chip at temperature T a >38℃, start, cooling power P c ≤10W;

[0083] Level 3: When the temperature changes by more than ±5°C, the sensor is triggered to temporarily sleep and automatically restart.

[0084] The sensor is equipped with a constant temperature chamber, housing both a PTC heater and a semiconductor cooler. Using a PID algorithm to control the temperature and the chamber's configuration, the sensor's operating temperature remains stable at 15-35°C. This improves the spectrometer's signal-to-noise ratio by 20% and reduces thermal imaging temperature drift from ±5°C to ±0.5°C.

[0085] As a preferred implementation of this embodiment, the drought stress collaborative analysis unit obtains a water stress index CWSI through a comprehensive stress index algorithm. The lower the water stress index CWSI, the more severe the drought stress. The formula of the stress index algorithm is:

[0086]

[0087] Where Tc is the canopy temperature, obtained by thermal imaging lens; Tc max and Tc min are the theoretical maximum and minimum values of canopy temperature, respectively, and are calculated by air humidity RH and wind speed V: Tc min =T air-(RH×0.05)-(V×0.1), Tc max =T air +(RH×0.03)+(V×0.05), Tc min Reflects the minimum temperature of the canopy under ideal moisture conditions (such as high humidity and strong winds that enhance evaporative cooling), Tc max The canopy temperature under the most severe water stress is simulated (e.g. low humidity and weak wind inhibiting transpiration and heat dissipation). In the calculation, the coefficients (e.g. 0.05, 0.1) are calibrated through field experiments to ensure that they are consistent with the actual canopy temperature change trend. For example: for every 1m / s increase in wind speed, Tc min Reduce by 0.1℃ (because wind speed accelerates heat loss); when humidity increases by 1%, Tc max Increased by 0.03°C (due to high humidity inhibiting canopy heat dissipation); θ v is the volumetric water content of soil (percentage); θ v0 is the historical background soil moisture content (such as the average value of the same period in the past five years), reflecting the water demand for normal grassland growth; H is the vegetation height (m), which is inverted through LiDAR point cloud, for example, by using a clustering algorithm (such as DBSCAN) to distinguish surface points from vegetation points, and the maximum Z coordinate of the vegetation point cloud is extracted as the vegetation height; H0 is the benchmark vegetation height (m), which is set according to the grassland type (such as pasture, grassland).

[0088] Through the design of the water stress index CWSI, using Tc max and Tc min The meteorological parameterization can eliminate the influence of regional climate differences on the stress index (such as automatic adjustment of Tc baseline between arid and humid areas), and through θ v and θ v0 Finally, by setting the benchmark vegetation height, the CWSI can dynamically adapt to different grassland types and avoid the "one-size-fits-all" threshold problem.

[0089] Furthermore, the drought stress collaborative analysis unit obtains a differential water index (NDWI) through a differential water index algorithm. A larger differential water index indicates a higher degree of drought stress. The differential water index algorithm is specifically as follows:

[0090] D1: Calculate the initial moisture index NDWI0 using the initial moisture index calculation formula. The initial moisture index calculation formula is:

[0091] Among them, R Red-Edge is the red edge band reflectivity; R SWIR is the reflectivity in the shortwave infrared band;

[0092] D2: Based on the dynamic calibration unit, the ambient temperature T is corrected. The correction formula is: Tcompensation =1+0.01×(T-25°C), where 15°C≤T≤35°C;

[0093] D3: Calculate the difference moisture index NDWI, NDWI = NDWI0 × T compensation .

[0094] The differential water index calculation in this embodiment corrects the NDWI by using the vegetation height parameter to eliminate the impact of terrain undulation on the water index. The temperature compensation factor corrects the misjudgment of water stress caused by high temperature, thereby realizing the use of vegetation height and temperature compensation to form a spatial-meteorological-physiological multidimensional drought response model, which is superior to the traditional differential water index calculation.

[0095] Preferably, the data processing module further comprises a dynamic early warning unit for classifying drought stress and providing dynamic early warning based on a historical drought database and real-time monitoring data;

[0096] The drought stress classification specifically includes:

[0097] S11-Classification model construction: Random forest, support vector machine and deep learning algorithm are used to build a drought stress classification model based on input features, including water stress index CWSI (weight 30%), difference water index NDWI (weight 25%), soil moisture gradient (weight 20%), vegetation height change rate (Δt is the monitoring time interval, weight 15%), meteorological factors (temperature, humidity, weight 10%);

[0098] S12-Classification results: When NDWI≤0.3 and CWSI≤0.2, the grassland drought stress is classified as healthy grassland; when 0.3<NDWI≤0.6 or 0.2<CWSI≤0.4, the grassland drought stress is classified as mild drought; when NDWI>0.6 or CWSI>0.4, the grassland drought stress is classified as severe drought;

[0099] The dynamic early warning based on the historical drought database and real-time monitoring data specifically includes:

[0100] S21- Input early warning features, including water stress index CWSI, difference water index NDWI, soil moisture gradient Vegetation height change rate Meteorological factors and grassland drought stress classification results;

[0101] S22-Use the sliding window method to analyze the water stress index CWSI or difference water index NDWI in the last 30 days and calculate the drought trend coefficient C trend , Among them, D 当前值 is the current value of water stress index CWSI or difference water index NDWI, D 历史均值 is the average value of water stress index CWSI or difference water index NDWI in the past year, D 历史标准差 is the standard deviation of the water stress index CWSI or the difference water index NDWI in the past year;

[0102] S23- According to the drought stress classification results of grassland, a benchmark threshold is set (e.g., the healthy threshold is NDWI ≤ 0.3), combined with the drought trend coefficient C trend Dynamically modify the warning threshold, the correction result is: warning threshold = reference threshold × (1 + k × C trend ), k is the sensitivity coefficient (e.g. k = 0.2, which needs to be adjusted according to business needs);

[0103] S24-When the real-time water stress index CWSI or difference water index NDWI exceeds the dynamic threshold, a drought warning is triggered. The warning threshold is dynamically adjusted according to the grassland type, and the error rate is ≤5%.

[0104] Based on this, a sliding window method analyzes the last 30 days of NDWI or CWSI data to calculate the drought trend coefficient. The threshold is then dynamically adjusted based on the historical mean and standard deviation, ensuring that the threshold updates in sync with the current environmental state. This avoids the drawback of traditional fixed thresholds, which are unable to adapt to seasonal, cyclical, or sudden drought events. A grading system based on complementary NDWI and CWSI indicators avoids the limitations of a single indicator and improves warning accuracy, achieving an error rate of ≤5%.

[0105] In this embodiment, the interface corresponding to the standardized interface protocol supports unified conversion of RS-232, USB 3.0, and Ethernet protocols, and provides a hardware abstraction layer (HAL) to implement dynamic loading of sensor drivers. The technical requirements corresponding to the standardized interface protocol include:

[0106] During the data transmission process, the sensor raw data (such as thermal imaging temperature Tc, LiDAR point cloud height H, soil moisture θ v ) and calculate intermediate parameters (such as Tc in the CWSI formula max and Tc min ) uses a layered encapsulation format to make the algorithm directly callable;

[0107] The data packet header contains the algorithm identifier (such as CWSI, PID control parameters), timestamp (accuracy ±1ms), sensor ID and checksum (CRC-32);

[0108] HAL provides a dynamic loading interface for the algorithm library (such as CWSI, Kalman filter, and PID control), and supports switching algorithms at runtime based on environmental conditions (such as loading the Kalman filter driver when RH>80%).

[0109] Algorithm configuration parameters (such as PID Kp=0.5Kp=0.5, Ki=0.1Ki=0.1, Kd=0.05Kd=0.05) are updated in real time through the ROS2 parameter server;

[0110] The RS-232 interface supports the Modbus RTU protocol (baud rate ≤ 115200bps) for compatible access to low-speed sensors (such as traditional soil moisture sensors);

[0111] The Ethernet interface is compatible with the IEEE 802.3 1000BASE-T standard and supports real-time transmission of LiDAR point clouds (≥100MB / s) and CWSI calculation results.

[0112] The USB 3.0 interface complies with the USB 3.2 Gen 2 standard (bandwidth ≥ 10 Gbps) and is used for direct transmission of raw data from high-speed cameras (such as multispectral cameras ≥ 100 fps).

[0113] In the embodiments provided herein, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any appropriate combination thereof. For hardware implementation, the processor can be implemented in one or more of the following units: an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a processor, a controller, a microcontroller, a microprocessor, other electronic units designed to implement the functions described herein, or a combination thereof. For software implementation, part or all of the processes of the embodiments can be completed by instructing the relevant hardware through a computer program. When implemented, the above program can be stored in a computer-readable storage medium or transmitted as one or more instructions or codes on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein the communication media include any medium that facilitates the transmission of a computer program from one place to another. The storage medium can be any available medium that a computer can access. The computer-readable storage medium can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer.

[0114] Finally, it should be noted that the above is only a preferred embodiment of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent replacements for some of the technical features therein. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A grassland drought stress monitoring device based on a multi-source sensor of an unmanned aerial vehicle, characterized in that: include: UAV platform: equipped with a multi-rotor UAV or a fixed-wing UAV, the multi-rotor UAV and the fixed-wing UAV are equipped with an RTK module for sub-meter positioning and a multi-sensor collaborative operation gimbal; Multi-source sensor module: includes: a spectral camera, a thermal imaging lens, a lidar, a soil moisture sensor, and an IMU inertial navigation module. The spectral camera covers visible light, near-infrared, and short-wave infrared bands. The thermal imaging lens is used to monitor surface temperature. The lidar is used to obtain three-dimensional point cloud data. The soil moisture sensor is integrated into the mounting equipment of the multi-rotor drone and the fixed-wing drone, and obtains soil surface moisture content and temperature through hovering or landing measurements. Data processing module: used for output transmission and processing, the data processing module includes: Data transmission unit: transmits sensor raw data to the ground workstation in real time through wireless transmission; Flight attitude calculation unit: Based on the IMU inertial navigation module and RTK data, it compensates for the geometric deformation and point cloud distortion caused by the movement of the UAV sensor; Multi-source data spatiotemporal registration unit: performs spatial consistency processing on sensor raw data through timestamp synchronization and georeferencing; Drought stress collaborative analysis unit: integrates spectral index, thermal infrared data, point cloud features and soil moisture parameters to obtain grassland drought stress conditions; Dynamic calibration unit: corrects sensor performance deviations through real-time environmental parameters and compensates for measurement errors under complex meteorological conditions through redundant sensor data; The device enables each sensor in the multi-source sensor module to be plug-and-play through a standardized interface protocol, and the data transmission unit adopts an edge compression algorithm for data transmission.

2. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 1 is characterized in that: Synchronous trigger frequency of the spectral camera and the thermal imaging lens satisfy The point cloud density ρ of the laser radar satisfies ρ ≥ 1000 points / m 3 , and the soil moisture sensor performs in-situ measurement of surface soil moisture content through electromagnetic induction or neutron scattering method when hovering.

3. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 1 is characterized in that: The data processing module adopts a spatiotemporal registration algorithm, which specifically includes: Based on IMU inertial data and combined with RTK positioning coordinates, the timestamps of multi-sensor raw data are synchronized; Build a 3D terrain model using LiDAR point clouds and correct geometric projection errors of multispectral and thermal imaging data; Weather station data are introduced to correct the radiation error of thermal infrared data.

4. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 3 is characterized in that: The data processing module suppresses meteorological interference through an adaptive filtering algorithm, and the adaptive filtering algorithm includes: When the relative humidity RH satisfies RH>80%, Kalman filtering is enabled for the thermal imaging data; When the visibility VB satisfies VB<50 meters, the point cloud density is corrected by fusing LiDAR point cloud data with millimeter-wave radar data.

5. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 1 is characterized in that: The dynamic calibration unit is configured with an environment adaptive compensation algorithm, and the environment adaptive compensation algorithm specifically includes: The operating temperature of the sensor is controlled within a range of 15-35° C. by a PID algorithm, and the housing of each sensor in the multi-source sensor module is encapsulated with a polyimide composite material.

6. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 5 is characterized in that: The sensor operating temperature is controlled at 15-35°C by the PID algorithm based on a constant temperature control module. The constant temperature control module includes a three-level temperature protection mechanism. The three-level temperature protection mechanism specifically includes: Level 1: PTC heating plate at temperature T a Start when <10℃, heating power P h ≤5W; Second level: semiconductor refrigeration chip at temperature T a Start when it is >38℃, cooling power P c ≤10W; Level 3: When the temperature changes by more than ±5°C, the sensor is triggered to temporarily sleep and automatically restart.

7. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 1 is characterized in that: The drought stress collaborative analysis unit obtains the water stress index CWSI through a comprehensive stress index algorithm; the formula of the stress index algorithm is: Where Tc is the canopy temperature; Tc max and Tc min are the theoretical maximum and minimum values of canopy temperature respectively; θ v is the volumetric water content of soil; θ v0 is the historical background soil moisture content; H is the vegetation height; H0 is the benchmark vegetation height.

8. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 7 is characterized in that: The drought stress collaborative analysis unit obtains the difference water index NDWI through a difference water index algorithm, and the difference water index algorithm is specifically as follows: D1: Calculate the initial moisture index NDWI0 using the initial moisture index calculation formula. The initial moisture index calculation formula is: Among them, R Red-Edge is the red edge band reflectivity; R SWIR is the reflectivity in the short-wave infrared band; D2: Based on the dynamic calibration unit, the ambient temperature T is corrected. The correction formula is: T compensation =1+0.01×(T-25°C), where 15°C≤T≤35°C; D3: Calculate the difference moisture index NDWI, NDWI = NDWI0 × T compensation .

9. The grassland drought stress monitoring device based on drone multi-source sensors according to claim 8, characterized in that: The data processing module also includes a dynamic early warning unit for classifying drought stress and providing dynamic early warning based on a historical drought database and real-time monitoring data; The drought stress classification specifically includes: S11-Classification model construction: Random forest, support vector machine and deep learning algorithm are used to build a drought stress classification model based on input features, including water stress index CWSI, difference water index NDWI, soil moisture gradient Vegetation height change rate Meteorological factors; S12-Classification results: When NDWI≤0.3 and CWSI≤0.2, the grassland drought stress is classified as healthy grassland; when 0.3<NDWI≤0.6 or 0.2<CWSI≤0.4, the grassland drought stress is classified as mild drought; when NDWI>0.6 or CWSI>0.4, the grassland drought stress is classified as severe drought; The dynamic early warning based on the historical drought database and real-time monitoring data specifically includes: S21- Input early warning features, including water stress index CWSI, difference water index NDWI, soil moisture gradient Vegetation height change rate Meteorological factors and grassland drought stress classification results; S22-Use the sliding window method to analyze the water stress index CWSI or difference water index NDWI in the last 30 days and calculate the drought trend coefficient C trend , Among them, D current value is the current value of the water stress index CWSI or the difference water index NDWI, D historical mean is the mean of the water stress index CWSI or the difference water index NDWI in the past year, and D historical standard deviation is the standard deviation of the water stress index CWSI or the difference water index NDWI in the past year; S23-Set the benchmark threshold based on the grassland drought stress classification results, combined with the drought trend coefficient C trend Dynamically modify the warning threshold, the correction result is: warning threshold = reference threshold × (1 + k × C trend ), k is the sensitivity coefficient; S24-When the real-time water stress index CWSI or difference water index NDWI exceeds the dynamic threshold, a drought warning is triggered.

10. A method for monitoring grassland drought stress based on a multi-source sensor of an unmanned aerial vehicle, applied to the grassland drought stress monitoring device based on a multi-source sensor of an unmanned aerial vehicle according to any one of claims 1 to 9, characterized in that: The method comprises the following steps in sequence: Setting up a drone platform, wherein the drone platform is configured with a multi-rotor drone or a fixed-wing drone, wherein the multi-rotor drone and the fixed-wing drone are equipped with an RTK module for sub-meter positioning and a multi-sensor collaborative operation gimbal; Controlling the multi-sensor gimbal to synchronously collect raw sensor data from a spectral camera, thermal imaging lens, lidar, soil moisture sensor, and IMU inertial navigation module, wherein the spectral camera covers visible light, near-infrared, and short-wave infrared bands, the thermal imaging lens is used to monitor ground surface temperature, the lidar is used to obtain three-dimensional point cloud data, and the soil moisture sensor is integrated into the mounting equipment of the multi-rotor drone and the fixed-wing drone to obtain soil surface moisture content and temperature through hovering or landing measurements; Compensate for the geometric deformation and point cloud distortion of the sensor caused by the movement of the drone based on the IMU inertial navigation module and RTK data; Spatial consistency processing of sensor raw data through timestamp synchronization and georeferencing; Correct sensor performance deviations through real-time environmental parameters and compensate for measurement errors under complex meteorological conditions through redundant sensor data; Transmit sensor raw data to the ground workstation in real time via wireless transmission; The ground workstation fuses spectral index, thermal infrared data, point cloud features and soil moisture parameters to obtain grassland drought stress conditions.

Images