An unmanned aerial vehicle-based crop canopy temperature acquisition system and control method

Abstract

The application discloses a crop canopy temperature acquisition system and control method based on a UAV, relates to the technical field of intelligent agricultural monitoring, and comprises a multi-rotor UAV platform, a sensor nacelle provided with a thermal infrared camera and a depth camera, an environment sensing module, an airborne computing unit and a ground station system, a true solar time is calculated according to the latitude and longitude of a target field to set an acquisition window, and an adverse wind flight route is planned; after flying to a waypoint, the standard deviation of wind speed is calculated in real time, and acquisition is triggered after hovering stably; thermal infrared images and depth images are synchronously acquired in multiple frames, soil background is removed through image registration and canopy mask, the mean value is obtained after the multi-frame temperature values of the canopy area are counted and abnormal values are removed; the canopy temperature is corrected according to a compensation model called according to real-time wind speed data; the canopy temperature data with geographic coordinates are generated in combination with positioning data, a temperature distribution map is fused and generated, the problem of temperature measurement distortion caused by rotor airflow interference is effectively solved, and the canopy temperature acquisition precision is improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural intelligent monitoring technology, specifically to a crop canopy temperature acquisition system and control method based on unmanned aerial vehicles (UAVs). Background Technology

[0002] Crop canopy temperature is closely related to water stress and is an effective indicator reflecting crop water surplus or deficit. Calculating the Crop Water Stress Index (CWSI) using canopy temperature data can guide precision irrigation, achieving water conservation, increased yield, and improved quality. Rapidly and accurately obtaining crop canopy temperature information has significant practical implications for improving agricultural water resource utilization.

[0003] Currently, crop canopy temperature is mainly collected using the following methods:

[0004] (a) Fixed sensor monitoring. Infrared temperature sensors are deployed in the field to measure crop canopy temperature at fixed points. This method has a limited number of sensors, low spatial resolution, and difficulty in reflecting the spatial heterogeneity of canopy temperature in the field; at the same time, the installation location of fixed sensors is limited, making it difficult to adapt to changes in crop canopy height at different growth stages.

[0005] (II) Ground-based mobile data collection robots. In recent years, with the development of intelligent agricultural equipment, ground-based mobile robot solutions for canopy temperature collection have emerged. These solutions typically employ wheeled, tracked, or wheel-legged hybrid locomotives equipped with thermal infrared cameras to move and collect data in the field. Among them, wheel-legged hybrid robots are equipped with multi-degree-of-freedom locomotives and multi-degree-of-freedom robotic arms to improve terrain mobility.

[0006] Although ground-based mobile data acquisition robots have achieved a certain degree of automation in canopy temperature acquisition, the applicant has discovered through research and practice that this type of solution has the following fundamental technical defects:

[0007] (i) Poor terrain mobility. Farmland has complex terrain such as ridges, ditches, uneven ground, and densely planted crops. Wheeled or tracked robots have poor mobility when crossing ridges and navigating narrow rows, and are prone to getting stuck or overturning. Although wheeled-legged hybrid robots are theoretically more adaptable, their multi-degree-of-freedom walking mechanism is complex, difficult to control, and expensive. In practical applications, stability is difficult to guarantee, making it difficult to operate effectively in vast hilly and mountainous areas and complex fields.

[0008] (ii) Low operational efficiency and poor spatiotemporal consistency of data. Ground robots walk slowly and it takes several hours to complete large-area data collection. Canopy temperature is significantly affected by solar radiation, and data collected at different times lack comparability due to different lighting conditions, making it impossible to stitch together an effective field temperature distribution map.

[0009] (iii) High equipment cost. Multi-degree-of-freedom walking mechanisms and robotic arms significantly increase the mechanical complexity, manufacturing cost, and maintenance difficulty of the system, making it difficult to promote and apply among ordinary farmers and agricultural cooperatives.

[0010] In summary, ground-based mobile data collection robot solutions are constrained by factors such as terrain mobility, operational efficiency, data spatiotemporal consistency, and cost. Summary of the Invention

[0011] In response to the problems mentioned in the background art, the present invention provides a crop canopy temperature acquisition system and control method based on unmanned aerial vehicles (UAVs). The purpose is to solve the technical problems that the crop canopy temperature acquisition scheme of ground mobile acquisition robots is constrained by factors such as terrain passability, operation efficiency, data spatiotemporal consistency, and cost.

[0012] To achieve the above-mentioned objectives, this invention provides a crop canopy temperature acquisition system based on unmanned aerial vehicles (UAVs), the system comprising:

[0013] A multi-rotor unmanned aerial vehicle (UAV) platform, which has autonomous flight and hovering capabilities, is equipped with a flight controller and a positioning module. The positioning module is used to acquire positioning data of the multi-rotor UAV platform at various data collection points.

[0014] A sensor pod is installed under the fuselage of the UAV in the multi-rotor UAV platform. The sensor pod is equipped with a thermal infrared camera and a depth camera. The thermal infrared camera is used to acquire thermal infrared images of the target crop canopy, and the depth camera is used to acquire depth images of the target crop canopy.

[0015] An environmental perception module is installed on the fuselage of the UAV or on the sensor pod. The environmental perception module includes at least a wind speed sensor, which is used to collect wind speed data near the surface of the target crop canopy.

[0016] An airborne computing unit is mounted on the fuselage of the UAV, and the airborne computing unit is communicatively connected to the thermal infrared camera, the depth camera, the wind speed sensor and the flight controller.

[0017] The ground station system is communicatively connected to the airborne computing unit;

[0018] The airborne computing unit is configured to: receive the thermal infrared image and the depth image; perform image registration between the thermal infrared image and the depth image acquired at the same time, so that their pixel coordinates correspond one-to-one; identify the canopy region based on the depth value of each pixel in the registered depth image, generate a canopy mask, wherein the pixel value of the canopy region in the canopy mask is a first value, and the pixel value of the non-canopy region in the canopy mask is a second value; superimpose the canopy mask onto the registered thermal infrared image, retain the pixels in the thermal infrared image whose pixel value corresponds to the first value as the temperature data of the canopy region, and remove the soil background pixels in the thermal infrared image whose pixel value corresponds to the second value; compensate and correct the temperature data of the canopy region based on the wind speed data to obtain corrected canopy temperature data; associate the corrected canopy temperature data with the positioning data to generate canopy temperature data with geographic coordinates; and send the canopy temperature data with geographic coordinates to the ground station system.

[0019] The ground station system includes a waypoint planning module and a time window calculation module. The waypoint planning module is used to divide the target field into multiple grids and generate a data collection waypoint at the center of each grid. The time window calculation module is used to calculate the true solar time based on the latitude and longitude coordinates of the target field and set the data collection time window.

[0020] The flight controller controls the multi-rotor UAV platform to fly to each of the data acquisition waypoints and the data acquisition time window planned by the ground station system to perform the canopy temperature data acquisition task.

[0021] Preferably, the pixel value of the canopy region in the canopy mask is 1, and the pixel value of the non-canopy region in the canopy mask is 0; after the airborne computing unit superimposes the canopy mask onto the registered thermal infrared image, it sets the pixels with pixel values ​​of 0 in the thermal infrared image to invalid values, and retains the pixels with pixel values ​​of 1 as the temperature data of the canopy region.

[0022] Preferably, the sensor pod is mounted under the fuselage of the UAV platform via a telescopic boom. The telescopic boom includes at least two interconnected tubular components and is equipped with a drive mechanism. The drive mechanism drives the telescopic boom to extend and retract to adjust its extension length. The onboard computing unit controls the extension length of the telescopic boom based on the current flight altitude of the multi-rotor UAV platform, so that the distance between the thermal infrared camera and the depth camera and the top of the target crop canopy is maintained within a preset temperature measurement distance range.

[0023] Preferably, the sensor pod is equipped with a three-axis stabilization gimbal, and the thermal infrared camera and the depth camera are mounted on the three-axis stabilization gimbal. The three-axis stabilization gimbal has a built-in inertial measurement unit (IMU) for recording the attitude data of the three-axis stabilization gimbal in real time when the thermal infrared camera acquires multiple frames of thermal infrared images. Based on the attitude data, the airborne computing unit uses the attitude of the first frame of the multiple frames of thermal infrared images as a reference attitude, calculates the attitude deviation of each subsequent frame of thermal infrared images relative to the reference attitude, calculates a homography transformation matrix based on the attitude deviation, projects each subsequent frame of thermal infrared images onto the viewpoint corresponding to the reference attitude using the homography transformation matrix, and fuses the projected multiple frames of thermal infrared images to obtain a stabilized thermal infrared image. The airborne computing unit performs image registration based on the stabilized thermal infrared image and the depth image.

[0024] Preferably, the number of wind speed sensors is 2 to 4, and each wind speed sensor is evenly arranged along the circumference of the sensor pod; the airborne computing unit receives the wind speed data collected by each wind speed sensor, calculates the average value of each wind speed data as the current wind speed value, and determines the current wind direction based on the spatial distribution gradient of each wind speed data.

[0025] The present invention also provides a crop canopy temperature acquisition and control method based on the aforementioned UAV-based crop canopy temperature acquisition system, the method comprising:

[0026] Step S1: The ground station system calculates the true solar time based on the latitude and longitude coordinates of the target field to obtain the collection time window; the ground station system divides the target field into multiple grids, generates collection waypoints at the center of each grid, and sets the connection direction of each collection waypoint to the headwind direction based on the wind direction data of the collection day to generate a flight path;

[0027] Step S2: The multi-rotor UAV platform flies to the target data collection point according to the flight route, determines the hovering height according to the crop canopy height, and keeps the distance between the thermal infrared camera and the depth camera on the sensor pod and the top of the canopy within the preset temperature measurement distance range, and hovers at the hovering height;

[0028] Step S3: After hovering and stabilizing, the onboard computing unit controls the thermal infrared camera and the depth camera to be triggered synchronously, continuously acquiring multiple frames of thermal infrared images and multiple frames of depth images corresponding to the thermal infrared images.

[0029] Step S4: The airborne computing unit performs image registration on each frame of the thermal infrared image and the depth image acquired at the same time, so that the pixel coordinates of the two correspond one-to-one; reads the depth value of each pixel in the registered depth image, identifies the canopy region based on the depth value of each pixel, and generates a canopy mask. The pixel value of the canopy region in the canopy mask is a first value, and the pixel value of the non-canopy region in the canopy mask is a second value; the canopy mask is superimposed on the registered thermal infrared image, retains the pixels in the thermal infrared image whose pixel value corresponds to the first value, and uses the retained pixels as the temperature value of the canopy region; removes the soil background pixels in the thermal infrared image whose pixel value corresponds to the second value; calculates the mean and standard deviation of the temperature value of the same pixel in the canopy region across multiple frames, removes temperature values ​​that deviate from the mean by more than 2 times the standard deviation, and takes the mean of the remaining temperature values ​​to obtain the initial canopy temperature Traw of the target acquisition waypoint;

[0030] Step S5: The airborne computing unit acquires the wind speed data v recorded by the wind speed sensor at the acquisition time, calls the pre-stored wind speed-temperature deviation compensation model ΔT=f(v) to compensate and correct the initial canopy temperature Traw, and obtains the corrected canopy temperature Tcorrected=Traw+ΔT.

[0031] Step S6: The airborne computing unit acquires the positioning data obtained by the positioning module at the target acquisition waypoint, associates the corrected canopy temperature Tcorrected with the positioning data, generates canopy temperature data with geographic coordinates, and sends it to the ground station system; the ground station system fuses the canopy temperature data with geographic coordinates at each acquisition waypoint with spatial coordinates to generate a canopy temperature distribution map of the target field.

[0032] Preferably, in step S2, the multi-rotor UAV platform ascends from a position below the hovering height to the hovering height, and begins to perform hovering stability judgment after reaching the hovering height. The hovering stability judgment includes: the onboard computing unit receives wind speed data collected by the wind speed sensor in real time, calculates the standard deviation σv of the wind speed data within a preset sliding window, and when the standard deviation σv is continuously lower than a preset threshold for a preset stabilization time, it is determined that the hovering has stabilized and the acquisition action in step S3 is triggered; when the standard deviation σv is continuously higher than the preset threshold for a preset timeout time, it is determined that there is turbulence at the current acquisition waypoint, the onboard computing unit records the acquisition waypoint and controls the UAV platform to skip the acquisition waypoint.

[0033] Preferably, the wind speed-temperature deviation compensation model ΔT=f(v)=a·v²+b·v+c, where a, b, and c are compensation coefficients determined through offline calibration experiments; the compensation coefficients a, b, and c are determined as follows: in a controlled environment, artificial airflows of different wind speeds blow across the crop canopy, and the canopy temperature is simultaneously measured using the thermal infrared camera and a ground reference thermometer. The temperature difference between the thermal infrared camera measurement and the ground reference thermometer measurement at each wind speed is recorded, and the compensation coefficients a, b, and c are obtained through polynomial fitting.

[0034] Preferably, the airborne computing unit compensates and corrects the initial canopy temperature Traw in the following manner: when the flight direction of the multi-rotor UAV platform at the data acquisition point is consistent with the wind direction indicated by the wind direction data, Tcorrected = Traw + (a·v² + b·v + c) + kdirection, where kdirection is the direction compensation coefficient; when the flight direction of the multi-rotor UAV platform at the data acquisition point is opposite to the wind direction indicated by the wind direction data, Tcorrected = Traw + (a·v² + b·v + c), and the direction compensation coefficient is not applied.

[0035] Preferably, the method for calculating true solar time by the ground station system is as follows: The longitude difference between the local true solar time and Beijing time is calculated based on the longitude of the target field; a mean time difference correction value is calculated based on the ordinal number of the collection date; based on the longitude difference and the mean time difference correction value, true solar time 11:00 and true solar time 14:00 are converted to their corresponding Beijing times, and the converted Beijing time interval is used as the collection time window; the ground station system obtains the weather forecast for the day of collection; when the forecast is sunny, the Beijing time interval is used as the collection time window; when the forecast is cloudy, the ground station system obtains the radiation intensity data monitored in real time by the solar radiation sensor, determines the continuous period with the highest radiation intensity within the Beijing time interval, and uses this continuous period as the collection time window; when the forecast is cloudy, rainy, or the wind speed is greater than a preset wind speed threshold, the collection task for that day is cancelled.

[0036] One or more technical solutions provided by this invention have at least the following technical effects or advantages:

[0037] Overcoming terrain limitations and achieving full terrain coverage, this invention uses a multi-rotor UAV as a data collection platform, fundamentally solving the problem of complex terrain such as ridges, ditches, and dense planting areas that are difficult for ground robots to traverse. It can achieve canopy temperature collection in various terrain conditions such as hills, mountains, and paddy fields.

[0038] Significantly improves operational efficiency. The flight speed of drones is much higher than that of ground robots. Combined with grid-based waypoint planning and automated data collection processes, temperature data collection can be completed over large areas of fields in a short time, making it particularly suitable for large-scale monitoring tasks requiring rapid response.

[0039] To ensure data consistency in time and space, the true solar time constraint acquisition window ensures that all acquisition points complete measurements under the same solar radiation conditions. The high-speed movement of drones compresses the acquisition time for the entire field into a short period, effectively eliminating the incomparability of temperatures caused by time differences.

[0040] This invention systematically addresses airflow interference to ensure temperature measurement accuracy. Through a multi-pronged approach involving headwind flight strategy, dynamic hovering stability assessment, wind speed compensation model, and multi-frame anomaly removal, it systematically reduces and compensates for the impact of rotor airflow on temperature measurement from three levels: flight strategy, data acquisition process, and data correction. This improves temperature measurement accuracy to meet the requirements of CWSI calculations.

[0041] The structure is simple and the cost is low. Compared with the complex walking mechanisms and multi-degree-of-freedom robotic arms of ground robots, the hardware structure of this invention is simpler, eliminating the need for complex walking mechanisms, reducing equipment costs and maintenance difficulty, and facilitating widespread application. Attached Figure Description

[0042] The accompanying drawings, which are provided to further illustrate embodiments of the invention and constitute a part of this invention, are not intended to limit the scope of the invention.

[0043] Figure 1 This is a schematic diagram of a crop canopy temperature acquisition system based on unmanned aerial vehicles (UAVs).

[0044] Figure 2 is a flowchart of a crop canopy temperature acquisition and control method. Detailed Implementation

[0045] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, where there is no conflict, the embodiments of the present invention and the features thereof can be combined with each other.

[0046] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0047] Example 1;

[0048] Please refer to Figure 1 , Figure 1 This is a schematic diagram of a crop canopy temperature acquisition system based on a drone. Embodiment 1 of the present invention provides a crop canopy temperature acquisition system based on a drone, the system comprising:

[0049] A multi-rotor unmanned aerial vehicle (UAV) platform, which has autonomous flight and hovering capabilities, is equipped with a flight controller and a positioning module. The positioning module is used to acquire positioning data of the multi-rotor UAV platform at various data collection points.

[0050] A sensor pod is installed under the fuselage of the UAV in the multi-rotor UAV platform. The sensor pod is equipped with a thermal infrared camera and a depth camera. The thermal infrared camera is used to acquire thermal infrared images of the target crop canopy, and the depth camera is used to acquire depth images of the target crop canopy.

[0051] An environmental perception module is installed on the fuselage of the UAV or on the sensor pod. The environmental perception module includes at least a wind speed sensor, which is used to collect wind speed data near the surface of the target crop canopy.

[0052] An airborne computing unit is mounted on the fuselage of the UAV, and the airborne computing unit is communicatively connected to the thermal infrared camera, the depth camera, the wind speed sensor and the flight controller.

[0053] The ground station system is communicatively connected to the airborne computing unit;

[0054] The airborne computing unit is configured to: receive the thermal infrared image and the depth image; perform image registration between the thermal infrared image and the depth image acquired at the same time, so that their pixel coordinates correspond one-to-one; identify the canopy region based on the depth value of each pixel in the registered depth image, generate a canopy mask, wherein the pixel value of the canopy region in the canopy mask is a first value, and the pixel value of the non-canopy region in the canopy mask is a second value; superimpose the canopy mask onto the registered thermal infrared image, retain the pixels in the thermal infrared image whose pixel value corresponds to the first value as the temperature data of the canopy region, and remove the soil background pixels in the thermal infrared image whose pixel value corresponds to the second value; compensate and correct the temperature data of the canopy region based on the wind speed data to obtain corrected canopy temperature data; associate the corrected canopy temperature data with the positioning data to generate canopy temperature data with geographic coordinates; and send the canopy temperature data with geographic coordinates to the ground station system.

[0055] The ground station system includes a waypoint planning module and a time window calculation module. The waypoint planning module is used to divide the target field into multiple grids and generate a data collection waypoint at the center of each grid. The time window calculation module is used to calculate the true solar time based on the latitude and longitude coordinates of the target field and set the data collection time window.

[0056] The flight controller controls the multi-rotor UAV platform to fly to each of the data acquisition waypoints and the data acquisition time window planned by the ground station system to perform the canopy temperature data acquisition task.

[0057] The applicant's research found that although UAV thermal infrared remote sensing technology has been initially applied in agriculture, mainly for large-scale crop growth monitoring and disaster assessment, directly using UAVs for canopy temperature collection presents the following problems:

[0058] (i) Rotor airflow interference leads to temperature measurement distortion. When multi-rotor UAVs hover or fly at low speeds, the downwash airflow generated by the rotors directly impacts the crop canopy, forcing convection heat transfer and accelerating heat loss from the leaf surface, resulting in the canopy temperature measured by the thermal infrared camera being significantly lower than the true value. Rotor airflow can cause temperature measurement deviations of 0.5~2℃. For CWSI calculations, an error of 0.5℃ can lead to misjudgment of the water stress level.

[0059] (ii) There is a contradiction between flight altitude and spatial resolution. When the flight altitude is too low (<10m), the downwash airflow impact is strong, resulting in large temperature measurement errors; when the flight altitude is too high (>50m), the spatial resolution of the thermal infrared image is insufficient, making it difficult to extract fine temperature information of a single plant or a single row of canopy.

[0060] (iii) Lack of dedicated canopy temperature acquisition methods. Existing UAV remote sensing technologies mostly adopt conventional aerial photography methods and have not systematically designed methods for the specific task of canopy temperature acquisition. There is a lack of targeted solutions in terms of acquisition time windows, flight parameter optimization, and airflow interference compensation.

[0061] This embodiment provides a crop canopy temperature acquisition technology solution that combines the advantages of an aerial platform while effectively solving temperature measurement accuracy problems such as airflow interference.

[0062] This invention addresses the technical problems of existing ground-based mobile data acquisition robots, such as poor terrain mobility, low operational efficiency, poor spatiotemporal data consistency, and high costs, as well as the lack of a dedicated method for canopy temperature acquisition and the existence of rotor airflow interference leading to temperature measurement distortion in existing UAV aerial photography technology. It provides a UAV-based crop canopy temperature acquisition system and its control method.

[0063] This invention utilizes the aerial flight characteristics of multi-rotor drones to fundamentally avoid terrain obstacles encountered on the ground. The drones are not limited by ground conditions such as ridges, ditches, or densely planted crops, and can quickly reach the target collection point under any terrain conditions.

[0064] This invention utilizes a thermal infrared camera to acquire thermal infrared images (temperature distribution) of the canopy, while a depth camera simultaneously acquires depth images (three-dimensional spatial information) of the canopy. Because there are significant differences in depth between the crop canopy and soil (the canopy is located within a certain height range, while the soil is at ground level), the depth values ​​of each pixel in the depth image can accurately distinguish between the canopy and soil. By registering the thermal infrared image and the depth image, a canopy mask is generated by identifying the canopy region using the depth image. This mask is then superimposed onto the thermal infrared image, retaining only the temperature data of the canopy region pixels and removing soil background pixels, thereby obtaining pure canopy temperature data and solving the problem of soil background interference.

[0065] This invention involves installing wind speed sensors around the sensor pod to collect real-time wind speed data near the canopy surface (as a characterization parameter of the rotor downwash airflow intensity). An offline calibration experiment establishes a mapping model (compensation model) between wind speed and temperature measurement deviation. During actual data acquisition, the temperature data is compensated and corrected based on the real-time wind speed to eliminate the influence of rotor airflow on temperature measurement and solve the rotor airflow interference problem.

[0066] In the flight path planning stage, this invention sets the main flight direction to a headwind based on wind direction data. When flying against the wind, the rotor downwash is carried away from the canopy surface by the natural wind, rather than accumulating above the canopy, thereby reducing the impact intensity of the downwash on the canopy and minimizing airflow interference at the source.

[0067] After the UAV of this invention arrives at the data collection point, it does not wait at a fixed time, but instead calculates the standard deviation of the wind speed data in real time as an indicator of airflow stability. Data collection is only triggered when the standard deviation is continuously lower than the threshold, ensuring that the airflow field is sufficiently stable at the time of data collection and guaranteeing the stability of the airflow field during data collection.

[0068] This invention continuously acquires multiple frames of images, performs statistical analysis on the temperature values ​​of the same pixel in the canopy region across multiple frames, removes outliers that deviate from the mean by more than 2 standard deviations (caused by instantaneous airflow disturbances or leaf obstruction), and takes the average of the remaining values ​​as the canopy temperature at that point, thereby improving the reliability of temperature data.

[0069] This invention calculates true solar time based on the longitude of the target field and constrains the data collection window to between 11:00 and 14:00 true solar time to ensure consistent solar radiation conditions at all collection points. It also adaptively adjusts the data based on meteorological conditions (sunny, cloudy, rainy), dynamically selecting the period with the highest radiation intensity during cloudy weather and canceling data collection during rainy or windy weather, thus ensuring spatiotemporal consistency of the data.

[0070] Existing ground-based mobile data collection robots are limited by terrain and cannot operate effectively in complex farmland. They also lack a system architecture design to correlate thermal infrared temperature data with spatial location, making it difficult to generate usable canopy temperature distribution maps. Furthermore, canopy temperature data collected by drones is easily affected by soil background interference, leading to inaccurate temperature data. This invention utilizes a thermal infrared camera + depth camera: depth images acquire three-dimensional spatial information of the canopy, while thermal infrared images acquire canopy temperature information; an onboard computing unit registers the thermal infrared and depth images, uses the depth image to identify canopy regions and generate a canopy mask, and overlays the mask onto the thermal infrared image, retaining canopy region pixels while removing soil background pixels; simultaneously, wind speed compensation correction is applied to the canopy temperature; positioning data is received and correlated with temperature data to generate canopy temperature data with geographic coordinates; a ground station system includes a waypoint planning module to generate collection waypoints and a time window calculation module to set the collection window; and a flight controller to control the drone's flight based on waypoints and the time window. It provides an end-to-end complete system architecture for canopy temperature acquisition, realizing the extraction of pure canopy temperature under soil background interference, accurate temperature correction under wind speed interference, and spatialization of temperature data with geographic coordinate binding. It can directly output corrected canopy temperature data with geographic coordinates for CWSI calculation.

[0071] In this embodiment of the invention, the multi-rotor UAV platform can be a heavy-duty hexacopter industrial-grade UAV. The multi-rotor UAV platform is equipped with a flight controller and an RTK high-precision positioning module (positioning accuracy at the centimeter level). A sensor pod is fixed to the underside of the multi-rotor UAV platform fuselage via mounting brackets. A thermal infrared camera and a depth camera are mounted on the sensor pod. An environmental perception module, including a wind speed sensor (such as the FS300A thermal film anemometer), is installed around the sensor pod. An onboard computing unit is installed on the UAV fuselage, connecting to the thermal infrared camera and depth camera via a USB interface, and to the wind speed sensor and flight controller via a UART serial port. The ground station system is a software system installed on the ground control computer, communicating with the onboard computing unit via a 4G or 5G data transmission module.

[0072] At the start of the data collection mission, the ground station system's waypoint planning module divides the target field (e.g., a 100m x 100m potato field) into a 10m x 10m grid, resulting in 100 data collection waypoints. The time window calculation module calculates true solar time based on the field's longitude (e.g., 104.06°E) and sets the data collection window to 11:30 to 14:30 Beijing time. The flight controller then directs the UAV to fly to each waypoint along the planned route. At each waypoint, the onboard computing unit receives thermal infrared images from the thermal infrared camera and depth images from the depth camera. After registering the two images, a canopy mask is generated based on the depth values ​​of each pixel in the depth image (canopy depth ranges from 0.3 to 0.8 m, and soil depth is 0 m). The canopy mask is then overlaid onto the thermal infrared image to extract canopy temperature data. Simultaneously, compensation and correction are performed based on wind speed data. Finally, the corrected temperature data is correlated with RTK positioning data to generate canopy temperature data with geographic coordinates, which is then sent to the ground station.

[0073] This embodiment can also be applied to corn canopy temperature acquisition. The corn plant height is approximately 2.5m, and the canopy depth value identified by the depth camera ranges from 2.0 to 3.0m. The onboard computing unit pre-stores the corresponding canopy height parameters for corn, which are automatically retrieved during acquisition. The remaining methods are the same as those for potatoes, and will not be described in detail in this embodiment.

[0074] In this embodiment of the invention, the pixel value of the canopy region in the canopy mask is 1, and the pixel value of the non-canopy region in the canopy mask is 0. After the airborne computing unit overlays the canopy mask onto the registered thermal infrared image, it sets the pixels with pixel values ​​of 0 in the thermal infrared image to invalid values, and retains the pixels with pixel values ​​of 1 as the temperature data of the canopy region. The airborne computing unit can complete the canopy region preservation and soil background removal through simple numerical judgment, which has high computational efficiency and is convenient for real-time processing on an embedded platform.

[0075] The airborne computing unit generates and applies the canopy mask in the following manner:

[0076] The depth values ​​of each pixel in the depth image acquired by the depth camera are stored as 16-bit grayscale images in millimeters. After reading the depth image, the onboard computing unit sets the canopy height threshold range from hmin to hmax (for potatoes, hmin = 0.25m, hmax = 0.80m; this threshold range is pre-set and stored in the onboard computing unit according to the target crop type and growth stage, and can be adjusted according to actual conditions in practical applications). Each pixel in the depth image is traversed. If the depth value d of the pixel satisfies hmin ≤ d ≤ hmax, it is determined to be a canopy area, and the corresponding mask pixel value is set to 1; otherwise, it is determined to be a non-canopy area (soil background), and the corresponding mask pixel value is set to 0. After traversal, a binary mask image with the same resolution as the depth image is generated. Then, this binary mask image is superimposed pixel-by-pixel with the registered thermal infrared image: the temperature value of the position corresponding to the mask pixel value of 1 in the thermal infrared image is retained; the temperature value of the position corresponding to the mask pixel value of 0 is set to an invalid value. The onboard computing unit only performs subsequent processing on the retained temperature values.

[0077] In one embodiment of the invention, the sensor pod is mounted below the fuselage of the UAV platform via a retractable boom. The retractable boom comprises at least two interconnected tubular components and is equipped with a drive mechanism. This drive mechanism extends and retracts the retractable boom to adjust its extension length. The onboard computing unit controls the extension length of the retractable boom based on the current flight altitude of the multi-rotor UAV platform, ensuring that the distance between the thermal infrared camera and the depth camera and the top of the target crop canopy remains within a preset temperature measurement distance range. By using a retractable boom, the onboard computing unit automatically adjusts the boom's extension length based on the current flight altitude, maintaining the distance between the thermal infrared camera and the depth camera and the top of the canopy within a preset optimal temperature measurement distance range. When the flight altitude is high, extending the boom shortens the distance between the sensor and the canopy, ensuring spatial resolution; when the flight altitude is low, shortening the boom reduces the lever arm and mechanical vibration, avoiding increased airflow interference. This achieves adaptive optimization of the sensor-canopy distance at different flight altitudes, balancing the dual requirements of spatial resolution and temperature measurement accuracy.

[0078] The sensor pod is mounted on the underside of the multi-rotor UAV platform via a retractable boom. The retractable boom consists of two interconnected carbon fiber tubes. The drive mechanism is a miniature electric actuator, which extends and retracts via a lead screw. The onboard computing unit automatically calculates the required extension length, Larm, based on the current flight altitude, H. Specifically, the target distance between the sensor and the top of the canopy, Dtarget, is set to 10m. Given the current flight altitude H (provided by the flight controller), the target crop canopy height, hcrop (pre-inputted, e.g., hcrop = 0.4m for potatoes), and the fixed height of the sensor pod, hsensor = 0.1m, the required extension length, Larm, is calculated as: Dtarget + hcrop + hsensor - H. If the calculated result is less than 0.3m, 0.3m is used; if it is greater than 1.2m, 1.2m is used. The onboard computing unit sends the target extension length command to the electric actuator, which extends or retracts the inner tube to the target position. A potentiometer built into the electric actuator provides real-time feedback on the current position, forming a closed-loop control system.

[0079] In one embodiment of the invention, a three-axis stabilization gimbal is mounted on the sensor pod, and the thermal infrared camera and the depth camera are mounted on the three-axis stabilization gimbal. The three-axis stabilization gimbal has a built-in inertial measurement unit (IMU) for recording the attitude data of the three-axis stabilization gimbal in real time when the thermal infrared camera acquires multiple frames of thermal infrared images. Based on the attitude data, the airborne computing unit uses the attitude of the first frame of the multiple frames of thermal infrared images as a reference attitude, calculates the attitude deviation of each subsequent frame of thermal infrared images relative to the reference attitude, calculates a homography transformation matrix based on the attitude deviation, projects each subsequent frame of thermal infrared images onto the viewpoint corresponding to the reference attitude using the homography transformation matrix, and fuses the projected multiple frames of thermal infrared images to obtain a stabilized thermal infrared image. The airborne computing unit performs image registration based on the stabilized thermal infrared image and the depth image.

[0080] The sensor pod is equipped with a three-axis stabilized gimbal (controllable pitch, roll, and yaw axes). The thermal infrared camera and depth camera are mounted back-to-back on the fixed panel of the gimbal, with parallel optical axes and similar field of view. They move synchronously with the gimbal, maintaining a fixed relative posture.

[0081] The three-axis stabilized gimbal has a built-in high-precision IMU that records the gimbal's attitude data (including pitch angle θp, roll angle θr, and yaw angle θy) in real time at a sampling rate of 1000Hz for each frame of the multi-frame thermal infrared images captured by the thermal infrared camera.

[0082] At a specific acquisition waypoint, the thermal infrared camera continuously acquires five frames of thermal infrared images (denoted as F1 to F5, with a frame interval of 0.5 seconds), and the IMU simultaneously records the attitude angles corresponding to each frame. The onboard computing unit performs the following stabilization and fusion operations:

[0083] The attitude of the first frame image F1 is used as the reference attitude (θp0, θr0, θy0). For the i-th frame image Fi (i=2,3,4,5), its attitude deviation is calculated as follows: Δθp(i)=θp(ti)-θp0; Δθr(i)=θr(ti)-θr0; Δθy(i)=θy(ti)-θy0; where θp0 is the pitch angle at the time of F1 acquisition, θr0 is the roll angle at the time of F1 acquisition, and θy0 is the yaw angle at the time of F1 acquisition. ti is the acquisition time of the i-th frame image. Δθp(i), Δθr(i), and Δθy(i) represent the angular changes of the gimbal in the pitch, roll, and yaw directions at the time of acquisition of the i-th frame image relative to the time of acquisition of the first frame image, respectively. A 3×3 homography transformation matrix Hi is calculated based on the attitude deviation (the attitude deviation is converted into a rotation matrix using the Rodrigues formula, and the homography matrix is ​​calculated in combination with the camera intrinsic parameters).

[0084] The Rodriguez formula is R = I + sin(θ)·K + (1 - cos(θ))·K², where θ is the rotation angle (obtained by synthesizing Euler angles from attitude deviations), K is the antisymmetric matrix corresponding to the rotation axis, and I is the identity matrix. Combining the camera intrinsic parameter matrix Kcam (a 3×3 matrix composed of internal parameters such as the focal length and principal point coordinates of the thermal infrared camera) and the rotation matrix Ri, the homography transformation matrix is ​​calculated. The homography transformation matrix Hi is a 3×3 matrix, representing the projection transformation relationship from the i-th frame image plane to the reference pose image plane.

[0085] Image Fi is projected onto the reference pose viewpoint through homography transformation: Fi' = Hi·Fi, where Fi' is the i-th frame image after projection transformation, Hi is the homography transformation matrix corresponding to the i-th frame image, and Fi is the original i-th frame image. This operation means that each pixel coordinate in Fi is projected through Hi to obtain the pixel position under the new viewpoint.

[0086] The average value of each pixel in the five projected images F1', F2', F3', F4', and F5' (where F1' = F1, because the transformation matrix H1 of the first frame is the identity matrix) is taken to obtain a stabilized thermal infrared image Ffused = (F1' + F2' + F3' + F4' + F5') / 5, where Ffused is the fused output thermal infrared image, and each pixel value is the arithmetic mean of the corresponding pixel temperature values ​​in the five images. This stabilized thermal infrared image Ffused serves as the base image for the onboard computing unit to subsequently register with the depth image and extract the canopy temperature.

[0087] During hovering and data acquisition, drones inevitably experience swaying due to wind and their propulsion systems, leading to spatial misalignment between multiple frames of thermal infrared images. This invention synchronizes the thermal infrared camera and depth camera with the gimbal, maintaining a fixed relative pose (providing a prerequisite for constant extrinsic parameters for registration). The inertial measurement unit records the gimbal attitude data at the time of each image acquisition at a sampling rate of ≥1000Hz. Using the attitude of the first frame as a reference, the attitude deviation of subsequent frames is calculated. Each frame is projected onto the reference attitude viewpoint using a homography transformation matrix and fused to obtain a stabilized thermal infrared image. This stabilized thermal infrared image is then used by the onboard computing unit for registration with the depth image. This stabilization and fusion eliminates the inter-frame spatial misalignment caused by drone swaying, ensuring that temperature statistics for the same pixel in the canopy region across multiple frames are based on the correct spatial correspondence. This guarantees the physical meaning and accuracy of multi-frame mean calculation and anomaly removal, ultimately ensuring the reliability of temperature extraction.

[0088] In one embodiment of the present invention, the number of wind speed sensors is 2 to 4, and each wind speed sensor is evenly arranged along the circumference of the sensor pod; the airborne computing unit receives the wind speed data collected by each wind speed sensor, calculates the average value of each wind speed data as the current wind speed value, and determines the current wind direction based on the spatial distribution gradient of each wind speed data.

[0089] Among these methods, multi-point measurement reduces the impact of local turbulence on wind speed measurement, making the wind speed data more representative; at the same time, the use of spatial distribution gradients to determine wind direction provides a data basis for optimizing headwind flight and applying direction compensation coefficients.

[0090] Four miniature hot-film anemometers are evenly distributed around the circumference of the sensor pod, located at the front (0°), back (180°), left (90°), and right (270°) positions, respectively, with each sensor 5 cm horizontally from the center of the pod. The onboard computing unit 4 synchronously reads the wind speed data from the four sensors at a sampling rate of 50 Hz, denoted as vfront, vback, vleft, and vright. The current wind speed value v = (vfront + vback + vleft + vright) / 4.

[0091] The wind direction is determined as follows: Calculate the horizontal wind speed gradient: Gx = vright - vleft (right minus left, a positive value indicates the wind is blowing from left to right), Gy = vfront - vback (front minus back, a positive value indicates the wind is blowing from back to front). The current wind angle θwind = atan2(Gy, Gx) (with 0° in front of the pod as 0°, and counterclockwise as positive). For example, a positive Gx indicates the wind is blowing from left to right, and a positive Gy indicates the wind is blowing from back to front.

[0092] Example 2;

[0093] Please refer to Figure 2 Figure 2 is a flowchart illustrating a crop canopy temperature acquisition and control method. Based on Embodiment 1, Embodiment 2 of the present invention provides a crop canopy temperature acquisition and control method, the method comprising:

[0094] Step S1 - Time Window Calculation and Waypoint Planning: The ground station system calculates the true solar time based on the latitude and longitude coordinates of the target field to obtain the data collection time window (e.g., using true solar time 11:00 to 14:00 as the data collection time window (corresponding to Beijing time 11:30 to 14:30)); the ground station system divides the target field into multiple grids, generates data collection waypoints at the center of each grid, and sets the connection direction of each data collection waypoint to the upwind direction based on the wind direction data of the day of data collection, thus generating a flight path;

[0095] Step S2 - Fly to waypoint and hover: The multi-rotor UAV platform flies to the target data collection waypoint according to the flight path, determines the hovering height according to the crop canopy height, and keeps the distance between the thermal infrared camera and the depth camera on the sensor pod and the top of the canopy within the preset temperature measurement distance range, and hovers at the hovering height;

[0096] Step S3 - Multi-frame synchronous acquisition: After hovering and stabilizing, the airborne computing unit controls the thermal infrared camera and the depth camera to trigger synchronously and continuously acquire multiple frames of thermal infrared images and multiple frames of depth images corresponding to the thermal infrared images.

[0097] Step S4 - Image Registration, Canopy Recognition, and Temperature Extraction: The airborne computing unit performs image registration on each frame of the thermal infrared image and the depth image acquired at the same time, ensuring a one-to-one correspondence between their pixel coordinates; it reads the depth values ​​of each pixel in the registered depth image, identifies the canopy region based on the depth values ​​of each pixel, and generates a canopy mask. The pixel values ​​of the canopy region in the canopy mask are first values, and the pixel values ​​of the non-canopy region in the canopy mask are second values; the canopy mask is superimposed on the registered thermal infrared image, retaining the pixels in the thermal infrared image whose pixel values ​​correspond to the first values, and using the retained pixels as the temperature values ​​of the canopy region; it removes the soil background pixels in the thermal infrared image whose pixel values ​​correspond to the second values; it calculates the mean and standard deviation of the temperature values ​​of the same pixel in the canopy region across multiple frames, removes temperature values ​​that deviate from the mean by more than twice the standard deviation, and then averages the remaining temperature values ​​to obtain the initial canopy temperature Traw of the target acquisition waypoint;

[0098] Step S5 - Wind speed compensation: The airborne computing unit obtains the wind speed data v recorded by the wind speed sensor at the time of acquisition, and calls the pre-stored wind speed-temperature deviation compensation model ΔT=f(v) to compensate and correct the initial canopy temperature Traw, so as to obtain the corrected canopy temperature Tcorrected=Traw+ΔT.

[0099] Step S6 - Temperature Spatialization and Temperature Field Reconstruction: The airborne computing unit acquires the positioning data obtained by the positioning module at the target acquisition point, associates the corrected canopy temperature (Tcorrected) with the positioning data, generates canopy temperature data with geographic coordinates, and sends it to the ground station system; the ground station system fuses the canopy temperature data with geographic coordinates and spatial coordinates at each acquisition point to generate a canopy temperature distribution map of the target field.

[0100] In this embodiment of the invention, in step S2, the multi-rotor UAV platform rises from a position below the hovering height to the hovering height, and begins to perform hovering stability determination after reaching the hovering height. The hovering stability determination includes: the onboard computing unit receives wind speed data collected by the wind speed sensor in real time, calculates the standard deviation σv of the wind speed data within a preset sliding window, and when the standard deviation σv is continuously lower than a preset threshold for a preset stabilization time, it is determined that the hovering has stabilized and the acquisition action in step S3 is triggered; when the standard deviation σv is continuously higher than the preset threshold for a preset timeout time, it is determined that there is turbulence at the current acquisition waypoint, the onboard computing unit records the acquisition waypoint and controls the UAV platform to skip the acquisition waypoint.

[0101] In existing technologies, drones hover for a fixed period (e.g., 10-15 seconds) before data collection. However, the time it takes for airflow to stabilize varies under different weather conditions; it may stabilize in 3-5 seconds in a light breeze, but remain unstable for up to 20 seconds in gusts. This fixed-waiting method either wastes time or results in data collection being interfered with before the airflow has stabilized. This invention addresses this by having the drone ascend from below its hovering altitude (ascent phase data collection strategy). The standard deviation σv of wind speed data within a sliding window is calculated in real-time as an airflow stability indicator. Data collection is triggered only when σv remains below a preset threshold for a preset stabilization time; if σv remains above the threshold for a preset timeout, turbulence is detected, and the waypoint is skipped. This transforms passive, timed waiting into proactive sensing and dynamic decision-making. It allows for rapid triggering in light breezes, saving time, and accurate identification and skipping in gusts, avoiding the collection of invalid data interfered with by turbulence. The ascent phase data collection strategy enables the drone to complete data collection during ascent, avoiding descent through its own downwash airflow zone, thus reducing airflow interference at the source of the process.

[0102] In this embodiment of the invention, the wind speed-temperature deviation compensation model ΔT=f(v)=a·v²+b·v+c, where a, b, and c are compensation coefficients determined through offline calibration experiments. The method for determining the compensation coefficients a, b, and c is as follows: in a controlled environment, artificial airflows at different wind speeds blow across the crop canopy, and the canopy temperature is simultaneously measured using the thermal infrared camera and a ground reference thermometer. The temperature difference between the thermal infrared camera measurement and the ground reference thermometer measurement at each wind speed is recorded, and the compensation coefficients a, b, and c are obtained through polynomial fitting.

[0103] The specific implementation method is as follows:

[0104] Offline calibration phase: A calibration platform is set up inside the greenhouse. Potted potato plants are placed on the ground. A drone thermal infrared camera is fixed directly above the potted plants. Multiple thermocouple thermometers are deployed at the height of the canopy surface as ground references.

[0105] An axial flow fan was set up to blow air onto the canopy from a fixed direction at different wind speeds. Under each wind speed condition, after the wind speed stabilized, multiple frames of thermal infrared images were simultaneously acquired using a thermal infrared camera, and the canopy surface temperature was recorded using a thermocouple thermometer for several seconds.

[0106] For each wind speed condition, multiple sets of data are obtained by calculating the difference between the average canopy temperature measured by the thermal infrared camera and the average temperature measured by the thermocouple.

[0107] The least squares method was used to perform quadratic polynomial fitting to obtain the fitting results.

[0108] In the online application phase: During actual data collection, the airborne computing unit acquires the wind speed sensor readings, substitutes them into the model calculation, and adds ΔT to the original canopy temperature Traw to obtain the corrected temperature Tcorrected.

[0109] In this embodiment of the invention, the airborne computing unit compensates and corrects the initial canopy temperature Traw in the following manner: when the flight direction of the multi-rotor UAV platform at the data collection point is consistent with the wind direction indicated by the wind direction data, Tcorrected = Traw + (a·v² + b·v + c) + kdirection, where kdirection is the direction compensation coefficient; when the flight direction of the multi-rotor UAV platform at the data collection point is opposite to the wind direction indicated by the wind direction data, Tcorrected = Traw + (a·v² + b·v + c), and the direction compensation coefficient is not applied.

[0110] The applicant's research found that during tailwind flight, the rotor downwash is pushed towards the canopy surface by the natural wind, resulting in stronger convective heat transfer and a larger temperature deviation compared to headwind flight. Models that rely solely on wind speed for compensation do not consider the influence of wind direction relative to the flight direction, leading to a greater residual error in tailwind segments compared to headwind segments. This invention adds a direction compensation coefficient as an independent correction term to the wind speed deviation compensation result. The value of kdirection was determined through tailwind / headwind comparative calibration experiments, and the value range of kdirection is 0.3℃ to 0.8℃. By additionally adding the direction compensation coefficient under tailwind conditions, the residual temperature measurement error in tailwind segments is reduced from approximately 0.35℃ to approximately 0.15℃ (comparable to headwind segments), eliminating the systematic measurement bias caused by wind direction and ensuring consistency of temperature data for the entire field across different flight directions.

[0111] In this embodiment of the invention, the method for calculating true solar time by the ground station system is as follows: The longitude difference between the local true solar time and Beijing time is calculated based on the longitude of the target field; a mean time difference correction value is calculated based on the ordinal number of the collection date; based on the longitude difference and the mean time difference correction value, true solar time 11:00 and true solar time 14:00 are converted to their corresponding Beijing times, and the converted Beijing time interval is used as the collection time window; the ground station system obtains the weather forecast for the collection day; when the forecast is sunny, the Beijing time interval is used as the collection time window; when the forecast is cloudy, the ground station system obtains the radiation intensity data monitored in real time by the solar radiation sensor, determines the continuous period with the highest radiation intensity within the Beijing time interval, and uses this continuous period as the collection time window; when the forecast is cloudy, rainy, or the wind speed is greater than a preset wind speed threshold, the collection task for that day is cancelled.

[0112] The ground station system incorporates built-in astronomical algorithms. Operators only need to input the longitude of the target field and the data collection date into the system; the system automatically performs longitude and mean time difference corrections before outputting the Beijing time corresponding to the data collection window. The ground station system obtains the weather conditions for the data collection day through a weather forecast interface.

[0113] The existing technologies lack scientific basis for determining the data acquisition time window, allowing operators to arbitrarily select flight periods, resulting in incomparable canopy temperatures at different waypoints and on different dates due to varying solar radiation conditions. Furthermore, existing methods do not consider the impact of weather conditions on data acquisition feasibility; flights on cloudy, rainy, or windy days not only lead to inaccurate measurements but also pose safety risks. This invention calculates the longitude time difference based on longitude and the average time difference correction value based on the date ordinal number, converting true solar time 11:00 and 14:00 to their corresponding Beijing time to obtain an executable time window. On clear days, a fixed window is used; on cloudy days, the system dynamically selects the continuous period with the highest radiation intensity within the window based on real-time data from the solar radiation sensor; data acquisition is cancelled on cloudy days, during rainfall, or when wind speeds exceed 5 m / s. Based on the needs of astronomical calculations, meteorological sensing, and engineering execution, this invention converts the theoretical true solar time window into Beijing time, which can be executed by flight control, through time difference and mean time difference correction. This ensures that the solar radiation benchmark of data collected at different longitudes and on different dates is consistent. The classification and handling of three types of weather—sunny, cloudy, and rainy with strong winds—ensures that data collection is always carried out under optimal lighting conditions, while avoiding safety risks under severe weather conditions.

[0114] In this embodiment of the invention, in step S1, the ground station system calculates the estimated total time required to complete all data collection waypoints based on the area of ​​the target field and the endurance of the multi-rotor UAV platform, and compares the estimated total time with the duration of the data collection time window. When the estimated total time is less than or equal to the duration of the data collection time window, all data collection waypoints are executed at once according to the flight path. When the estimated total time is greater than the duration of the data collection time window, the ground station system calculates the number of waypoints that can be completed based on the available working time of the UAV platform within the data collection time window, sorts the data collection waypoints according to their azimuth and solar altitude angle, prioritizes the collection of waypoints whose azimuth is perpendicular to the direction of solar incidence, and selects waypoints from the sorted waypoints as executable waypoints for the current flight based on the number of achievable waypoints. The remaining waypoints that are not selected are reserved for supplementary data collection in the next flight.

[0115] When the target field area is large and the number of waypoints is numerous, the total time required for the task may exceed the duration of the data collection window. If all data is required to be collected at once, some waypoints will inevitably be collected outside the window, resulting in incomparable data; simply canceling some waypoints will lead to incomplete data coverage. This method optimizes the collection order based on lighting conditions, ensuring that the data collected within the limited window is all high-quality and usable data, with the remaining waypoints reserved for supplementary collection, thus balancing data quality and task completeness.

[0116] The estimated total time is calculated as follows: Total number of waypoints collected × Time per waypoint. The total number of waypoints is obtained by dividing the target field area by the area of ​​a single grid cell. The flight time is then estimated based on the path distance between waypoints and the drone's cruising speed. The sum of these two estimates represents the complete operation cycle for each waypoint, which is then multiplied by the total number of waypoints to obtain the estimated total time. Time per waypoint = Time to reach the waypoint + Hovering and stabilization time + Multi-frame acquisition time + Climb and departure time.

[0117] The azimuth and solar altitude angles are ordered as follows: the azimuth is the core of the ranking. If the scoring directions of the two angles are opposite, the azimuth score is given primary and the solar altitude angle score is given secondary.

[0118] In step S6, the method for determining the spatial coordinates is as follows: the airborne computing unit obtains the current extension length of the telescopic boom, the current attitude angle of the three-axis stabilization gimbal, and the positioning data provided by the positioning module of the UAV platform. Based on the current extension length and the current attitude angle, the coordinates of the optical center of the thermal infrared camera in the UAV body coordinate system are calculated by forward kinematics. Then, the coordinates are transformed to the geodetic coordinate system by combining the positioning data to obtain the geographic coordinates of each temperature pixel at each acquisition waypoint.

[0119] A three-axis stabilized gimbal is connected to the underside of the drone via a retractable boom, on which a thermal infrared camera and a depth camera are mounted. Therefore, the path from the drone fuselage to the physical point in the canopy corresponding to the temperature pixel involves the following physical steps: drone fuselage → boom (adjustable length) → gimbal (adjustable attitude: pitch, roll, yaw) → camera optical center → canopy point in the camera's field of view. The corresponding coordinate transformation path is: drone fuselage coordinate system → gimbal base coordinate system → camera coordinate system → canopy point geodetic coordinates. The onboard computing unit performs calculations step-by-step along this path, ultimately mapping each temperature pixel to geographic coordinates. That is, starting from the drone fuselage GPS coordinates, the displacement of the boom length, the rotational displacement of the gimbal attitude, and the field of view distance corresponding to the depth value are successively added to calculate the actual geographical location corresponding to each temperature pixel. Using the known drone position, boom length, gimbal angle, and depth camera ranging value, the latitude and longitude of each temperature measurement point in the canopy are calculated. The specific calculation method is not detailed in this embodiment.

[0120] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0121] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A crop canopy temperature acquisition system based on unmanned aerial vehicles (UAVs), characterized in that, The system includes: A multi-rotor unmanned aerial vehicle (UAV) platform, which has autonomous flight and hovering capabilities, is equipped with a flight controller and a positioning module. The positioning module is used to acquire positioning data of the multi-rotor UAV platform at various data collection points. A sensor pod is installed under the fuselage of the UAV in the multi-rotor UAV platform. The sensor pod is equipped with a thermal infrared camera and a depth camera. The thermal infrared camera is used to acquire thermal infrared images of the target crop canopy, and the depth camera is used to acquire depth images of the target crop canopy. An environmental perception module is installed on the fuselage of the UAV or on the sensor pod. The environmental perception module includes at least a wind speed sensor, which is used to collect wind speed data near the surface of the target crop canopy. An airborne computing unit is mounted on the fuselage of the UAV, and the airborne computing unit is communicatively connected to the thermal infrared camera, the depth camera, the wind speed sensor and the flight controller. The ground station system is communicatively connected to the airborne computing unit; The airborne computing unit is configured to: receive the thermal infrared image and the depth image; perform image registration between the thermal infrared image and the depth image acquired at the same time, so that their pixel coordinates correspond one-to-one; identify the canopy region based on the depth value of each pixel in the registered depth image, generate a canopy mask, wherein the pixel value of the canopy region in the canopy mask is a first value, and the pixel value of the non-canopy region in the canopy mask is a second value; superimpose the canopy mask onto the registered thermal infrared image, retain the pixels in the thermal infrared image whose pixel value corresponds to the first value as the temperature data of the canopy region, and remove the soil background pixels in the thermal infrared image whose pixel value corresponds to the second value; compensate and correct the temperature data of the canopy region based on the wind speed data to obtain corrected canopy temperature data; associate the corrected canopy temperature data with the positioning data to generate canopy temperature data with geographic coordinates; and send the canopy temperature data with geographic coordinates to the ground station system. The ground station system includes a waypoint planning module and a time window calculation module. The waypoint planning module is used to divide the target field into multiple grids and generate a data collection waypoint at the center of each grid. The time window calculation module is used to calculate the true solar time based on the latitude and longitude coordinates of the target field and set the data collection time window. The flight controller controls the multi-rotor UAV platform to fly to each of the data acquisition waypoints and the data acquisition time window planned by the ground station system to perform the canopy temperature data acquisition task.

2. The crop canopy temperature acquisition system based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The pixel value of the canopy region in the canopy mask is 1, and the pixel value of the non-canopy region in the canopy mask is 0. After the airborne computing unit superimposes the canopy mask onto the registered thermal infrared image, it sets the pixels with pixel values ​​of 0 in the thermal infrared image to invalid values ​​and retains the pixels with pixel values ​​of 1 as the temperature data of the canopy region.

3. The crop canopy temperature acquisition system based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The sensor pod is mounted under the fuselage of the multi-rotor UAV platform via a telescopic boom. The telescopic boom includes at least two interconnected tubular components and is equipped with a drive mechanism. The drive mechanism drives the telescopic boom to extend and retract to adjust its extension length. The onboard computing unit controls the extension length of the telescopic boom based on the current flight altitude of the multi-rotor UAV platform, so that the distance between the thermal infrared camera and the depth camera and the top of the target crop canopy is kept within a preset temperature measurement distance range.

4. A crop canopy temperature acquisition system based on an unmanned aerial vehicle (UAV) according to claim 1, characterized in that, The sensor pod is equipped with a three-axis stabilized gimbal, on which the thermal infrared camera and the depth camera are mounted. The three-axis stabilized gimbal has a built-in inertial measurement unit (IMU) for real-time recording of its attitude data as the thermal infrared camera acquires multiple frames of thermal infrared images. Based on this attitude data, the airborne computing unit uses the attitude of the first frame of the multiple thermal infrared images as a reference attitude to calculate the attitude deviation of each subsequent frame relative to the reference attitude. It then calculates a homography transformation matrix based on the attitude deviation, projects each subsequent frame of thermal infrared images onto the viewpoint corresponding to the reference attitude using the homography transformation matrix, and fuses the projected multiple frames of thermal infrared images to obtain a stabilized thermal infrared image. The airborne computing unit then performs image registration between the stabilized thermal infrared image and the depth image.

5. A crop canopy temperature acquisition system based on an unmanned aerial vehicle (UAV) according to claim 1, characterized in that, The number of wind speed sensors is 2 to 4, and each wind speed sensor is evenly arranged around the circumference of the sensor pod; the airborne computing unit receives the wind speed data collected by each wind speed sensor, calculates the average value of each wind speed data as the current wind speed value, and determines the current wind direction based on the spatial distribution gradient of each wind speed data.

6. A method for controlling crop canopy temperature acquisition based on a UAV-based crop canopy temperature acquisition system according to any one of claims 1-5, characterized in that, The method includes: Step S1: The ground station system calculates the true solar time based on the latitude and longitude coordinates of the target field to obtain the collection time window; the ground station system divides the target field into multiple grids, generates collection waypoints at the center of each grid, and sets the connection direction of each collection waypoint to the headwind direction based on the wind direction data of the collection day to generate a flight path; Step S2: The multi-rotor UAV platform flies to the target data collection point according to the flight route, determines the hovering height according to the crop canopy height, and keeps the distance between the thermal infrared camera and the depth camera on the sensor pod and the top of the canopy within the preset temperature measurement distance range, and hovers at the hovering height; Step S3: After hovering and stabilizing, the onboard computing unit controls the thermal infrared camera and the depth camera to be triggered synchronously, continuously acquiring multiple frames of thermal infrared images and multiple frames of depth images corresponding to the thermal infrared images. Step S4: The airborne computing unit performs image registration on each frame of the thermal infrared image and the depth image acquired at the same time, so that the pixel coordinates of the two correspond one-to-one; reads the depth value of each pixel in the registered depth image, identifies the canopy region based on the depth value of each pixel, and generates a canopy mask. The pixel value of the canopy region in the canopy mask is a first value, and the pixel value of the non-canopy region in the canopy mask is a second value; the canopy mask is superimposed on the registered thermal infrared image, retains the pixels in the thermal infrared image whose pixel value corresponds to the first value, and uses the retained pixels as the temperature value of the canopy region; removes the soil background pixels in the thermal infrared image whose pixel value corresponds to the second value; calculates the mean and standard deviation of the temperature value of the same pixel in the canopy region across multiple frames, removes temperature values ​​that deviate from the mean by more than 2 times the standard deviation, and takes the mean of the remaining temperature values ​​to obtain the initial canopy temperature Traw of the target acquisition waypoint; Step S5: The airborne computing unit acquires the wind speed data v recorded by the wind speed sensor at the acquisition time, calls the pre-stored wind speed-temperature deviation compensation model ΔT=f(v) to compensate and correct the initial canopy temperature Traw, and obtains the corrected canopy temperature Tcorrected=Traw+ΔT. Step S6: The airborne computing unit acquires the positioning data obtained by the positioning module at the target acquisition waypoint, associates the corrected canopy temperature Tcorrected with the positioning data, generates canopy temperature data with geographic coordinates, and sends it to the ground station system; the ground station system fuses the canopy temperature data with geographic coordinates at each acquisition waypoint with spatial coordinates to generate a canopy temperature distribution map of the target field.

7. The crop canopy temperature acquisition and control method according to claim 6, characterized in that, In step S2, the multi-rotor UAV platform ascends from a position below the hovering height to the hovering height, and begins to perform hovering stability judgment after reaching the hovering height. The hovering stability judgment includes: the onboard computing unit receives wind speed data collected by the wind speed sensor in real time, calculates the standard deviation σv of the wind speed data within a preset sliding window, and when the standard deviation σv is continuously lower than a preset threshold for a preset stabilization time, it is determined that the hovering has stabilized and the acquisition action in step S3 is triggered; when the standard deviation σv is continuously higher than the preset threshold for a preset timeout time, it is determined that there is turbulence at the current acquisition waypoint, the onboard computing unit records the acquisition waypoint and controls the UAV platform to skip the acquisition waypoint.

8. The crop canopy temperature acquisition and control method according to claim 6, characterized in that, The wind speed-temperature deviation compensation model is ΔT=f(v)=a·v²+b·v+c, where a, b, and c are compensation coefficients determined through offline calibration experiments. The compensation coefficients a, b, and c are determined as follows: in a controlled environment, artificial airflows at different wind speeds blow across the crop canopy, and the canopy temperature is simultaneously measured using the thermal infrared camera and a ground reference thermometer. The temperature difference between the thermal infrared camera measurement and the ground reference thermometer measurement at each wind speed is recorded, and the compensation coefficients a, b, and c are obtained through polynomial fitting.

9. The crop canopy temperature acquisition and control method according to claim 8, characterized in that, The airborne computing unit compensates and corrects the initial canopy temperature Traw in the following manner: when the flight direction of the multi-rotor UAV platform at the data acquisition point is consistent with the wind direction indicated by the wind direction data, Tcorrected = Traw + (a·v² + b·v + c) + kdirection, where kdirection is the direction compensation coefficient; when the flight direction of the multi-rotor UAV platform at the data acquisition point is opposite to the wind direction indicated by the wind direction data, Tcorrected = Traw + (a·v² + b·v + c), and the direction compensation coefficient is not applied.

10. The crop canopy temperature acquisition and control method according to claim 6, characterized in that, The method for calculating true solar time by the ground station system is as follows: The longitude difference between the local true solar time and Beijing time is calculated based on the longitude of the target field; the average time difference correction value is calculated based on the ordinal number of the collection date; based on the longitude difference and the average time difference correction value, true solar time 11:00 and true solar time 14:00 are converted to their corresponding Beijing times, and the Beijing time interval obtained after conversion is used as the collection time window; the ground station system obtains the weather forecast for the day of collection, and when the forecast is sunny, the Beijing time interval is used as the collection time window; When the forecast is cloudy, the ground station system acquires the radiation intensity data monitored in real time by the solar radiation sensor, determines the continuous period of highest radiation intensity within the Beijing time interval, and uses this continuous period as the acquisition time window. If the forecast indicates cloudy weather, rain, or wind speed exceeding the preset wind speed threshold, the data collection task for that day will be cancelled.

Images