A farmland water dynamic monitoring method and system supporting space-air-ground sensor fusion

Abstract

The application discloses a kind of support space-time-surface sensing fusion's farmland water dynamic monitoring method and system, it is related to water farmland state monitoring technical field, comprising: obtaining the farmland water in-situ data collected by ground anchor point monitoring layer;Based on farmland water in-situ data, trigger low-altitude patrol layer to execute dynamic patrol task, to obtain low-altitude hyperspectral image;Based on low-altitude hyperspectral image and farmland water in-situ data, generate low-altitude water distribution data, and with low-altitude water distribution data as constraint boundary, to the satellite remote sensing image obtained executes adaptive partition interpretation, to generate high-altitude water distribution data;Low-altitude water distribution data and high-altitude water distribution data are executed data merging, generate the farmland water dynamic distribution field with space-time-surface consistency expression.The spatial position adjustment is carried out to high-altitude water distribution data again and then weighted fusion is carried out, so that the farmland water dynamic distribution field finally generated realizes seamless connection of multilayer observation data.

Description

Technical Field

[0001] This invention relates to the field of farmland water status monitoring technology, specifically to a method and system for dynamic monitoring of farmland water that supports the fusion of air-space-ground sensors. Background Technology

[0002] Dynamic monitoring of farmland moisture plays a crucial role in farmland moisture management, drought assessment, and water-saving irrigation decisions. Existing farmland moisture monitoring methods mainly include three approaches: in-situ measurements at ground stations, low-altitude remote sensing by unmanned aerial vehicles (UAVs), and high-altitude remote sensing by satellites.

[0003] Ground-based in-situ measurements acquire soil moisture and crop stem flow data through sensors buried in the field. While accurate, these measurements only reflect moisture conditions at a point scale and cannot provide information on regional moisture distribution. Low-altitude remote sensing by drones, equipped with hyperspectral imaging devices, acquires spectral images of farmland areas, allowing for the inversion of moisture distribution within a specific area. However, low-altitude remote sensing operations require manual route planning and scheduled execution, resulting in a lag in monitoring activity compared to the onset of crop water stress. Furthermore, the coverage of low-altitude remote sensing is limited by the drone's endurance, hindering large-scale continuous monitoring. Satellite high-altitude remote sensing offers advantages such as wide coverage and fixed revisit periods, providing large-scale surface moisture information. However, the relatively low spatial resolution of satellite imagery and the difficulty of adapting single interpretation models to variations in spectral response across different moisture states lead to deficiencies in spatial detail and local accuracy. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for dynamic monitoring of farmland moisture that supports the fusion of air, space, and ground sensors.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic monitoring of farmland moisture supporting the fusion of air-space-ground sensors, specifically comprising the following steps: S1. Obtain in-situ farmland moisture data collected by the ground anchor monitoring layer, wherein the in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data. S2. Based on the in-situ data of farmland moisture, trigger the low-altitude patrol layer to perform a dynamic patrol task to obtain low-altitude hyperspectral images covering the area surrounding the ground anchor monitoring layer. S3. Based on the low-altitude hyperspectral image and the in-situ farmland moisture data, generate low-altitude moisture distribution data, and use the low-altitude moisture distribution data as a constraint boundary to drive the upper-altitude remote sensing layer to perform adaptive partitioning interpretation on the acquired satellite remote sensing image to generate upper-altitude moisture distribution data. S4. Merge the low-altitude moisture distribution data and the high-altitude moisture distribution data to generate a dynamic distribution field of farmland moisture with consistent expression between air, space, and ground.

[0006] Preferably, in step S1, in-situ farmland moisture data collected by the ground anchor monitoring layer is obtained. The in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data. The continuous time-series soil moisture data is used as the first type of in-situ data, and the continuous time-series crop stem flow data is used as the second type of in-situ data. The first trend extraction process is performed on the first type of in-situ data. The first trend extraction process includes: calculating the difference between soil moisture values ​​at adjacent sampling times; determining the baseline fluctuation reference value by the quantile value of the absolute value of the adjacent difference in the first type of in-situ data sequence recorded on the same day; determining that the soil moisture is in a stable stage when the absolute value of the adjacent difference is continuously less than the baseline fluctuation reference value and the number of consecutive occurrences reaches the first number determined by the number of stable continuous sampling points on the same day; determining that the soil moisture is in an increasing stage when the number of consecutive positive values ​​of the adjacent difference and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the baseline fluctuation reference value and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the baseline fluctuation reference value and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the third number determined by the number of consecutive decreasing sampling points on the same day; The second trend extraction process is performed on the second type of in-situ data. The second trend extraction process includes: smoothing the stem flow rate value sequence to obtain smooth stem flow rate values, calculating the smoothing difference between smooth stem flow rate values ​​at adjacent time points, determining the stem flow background fluctuation reference value based on the quantile value of the absolute value of the smoothing difference recorded on the same day, determining that the stem flow rate is in a stable stage when the absolute value of the smoothing difference is continuously less than the stem flow background fluctuation reference value and the number of consecutive times reaches the fourth number determined by the number of continuous sampling points for stable stem flow on the same day, determining that the stem flow rate is in an enhancing stage when the number of consecutive times the smoothing difference is continuously positive and the absolute value is continuously greater than the stem flow background fluctuation reference value reaches the fifth number determined by the number of continuous sampling points for enhanced stem flow, and determining that the stem flow rate is in a weakening stage when the number of consecutive times the smoothing difference is continuously negative and the absolute value is continuously greater than the stem flow background fluctuation reference value reaches the sixth number determined by the number of continuous sampling points for weakened stem flow. The soil moisture stage and the stem flow rate stage are compared time intervals. When the stem flow rate enters the enhancement stage while the soil moisture is in the stable stage, the corresponding time interval is determined as the data difference period. The patrol initiation time and patrol spatial range of the low-altitude patrol layer are determined based on the data difference period. The patrol initiation time is obtained by adding the start time of the data difference period and the response delay duration. The patrol spatial range is formed by expanding the radiation radius in all directions based on the ground anchor monitoring point where the data difference period occurs. The radiation radius is dynamically selected based on the difference in the slope of change between the first type of in-situ data and the second type of in-situ data during the data difference period, using either the first radius value or the second radius value. The first radius value is greater than the second radius value.

[0007] Preferably, in S2, the low-altitude patrol vehicle is controlled to fly according to the patrol space range and arrive at the first hovering position within the patrol space range. The first hovering position is selected at a spatial point that allows the hyperspectral imaging payload to cover the maximum effective area within the patrol space range in a single acquisition. At the first hovering position, the first deviation of the soil moisture value of the first type of in-situ data during the data difference period is calculated from the median value of the first type of in-situ data in the same period of the historical record. The second deviation of the stem flow rate value of the second type of in-situ data during the data difference period is calculated from the median value of the second type of in-situ data in the same period of the historical record. The first deviation and the second deviation are weighted and summed to obtain the comprehensive deviation value. The adjusted duration is obtained by multiplying the comprehensive deviation amplitude by the base duration. The base duration is the shortest integration time required for the hyperspectral imaging payload to obtain spectral data that meets the signal-to-noise requirements under standard conditions. The hyperspectral imaging payload is controlled to perform spectral data acquisition according to the adjusted duration, so as to obtain low-altitude hyperspectral image data at the first hovering position after duration adjustment.

[0008] Preferably, in step S3, the spectral values ​​of the water response band are extracted pixel by pixel from the low-altitude hyperspectral image, the extracted spectral values ​​are corrected based on the in-situ data of farmland moisture to determine the pixel-level moisture content value, and the pixel-level moisture content value is filled into the surrounding area to form low-altitude moisture distribution data. The boundary where the difference in moisture content between adjacent pixels in the low-altitude moisture distribution data is greater than the boundary judgment reference value is extracted as the boundary of moisture content change. The boundary judgment reference value is determined by the quantile value of the absolute value of the difference in moisture content of all pixels in the low-altitude moisture distribution data. The boundary of moisture content change is mapped onto satellite remote sensing imagery, and the satellite remote sensing imagery is segmented to generate multiple non-overlapping interpretation blocks; For each interpretation block, the arithmetic mean of the moisture content values ​​of the low-altitude moisture distribution data within the corresponding range of the block is calculated. The arithmetic mean is compared with the first and second content thresholds. If the arithmetic mean is less than the first content threshold, the first interpretation processing method is used to interpret the moisture content of the satellite remote sensing image pixels within the block. If the arithmetic mean is greater than the second content threshold, the second interpretation processing method is used to interpret the moisture content. If the arithmetic mean is between the first and second content thresholds, the third interpretation processing method is used to interpret the moisture content. The moisture content values ​​obtained from the interpretation of each block are spliced ​​together according to their spatial location to form upper-air moisture distribution data.

[0009] Preferably, in step S4, the low-altitude moisture distribution data is adjusted to the same spatial cell size as the high-altitude moisture distribution data to obtain low-altitude moisture distribution resampled data. Within the spatial overlap range between low-altitude moisture distribution resampling data and high-altitude moisture distribution data, the spatial distribution of pixel moisture content values ​​in the low-altitude moisture distribution resampling data is taken as the first spatial distribution form, and the spatial distribution of pixel moisture content values ​​in the high-altitude moisture distribution data is taken as the second spatial distribution form. Cross-correlation calculations are performed on the first spatial distribution form and the second spatial distribution form along the row and column directions, respectively. The translation step corresponding to the maximum cross-correlation value in the row direction is taken as the east-west position offset component, and the translation step corresponding to the maximum cross-correlation value in the column direction is taken as the north-south position offset component. The east-west position offset component and the north-south position offset component together constitute the position offset. The row and column coordinates of all pixels in the upper-level moisture distribution data are translated according to the position offset, so that the second spatial distribution pattern of the upper-level moisture distribution data within the overlapping range is aligned with the first spatial distribution pattern after translation. Within the overlapping range, a weighted fusion is performed on the moisture content values ​​of the translated upper-level moisture distribution data pixels and the moisture content values ​​of the low-level moisture distribution resampled data pixels. Outside the overlapping range, the moisture content values ​​of their respective data sources are used. The fusion result is then stitched together with the result outside the overlapping range to generate a dynamic distribution field of farmland moisture.

[0010] This invention also provides a dynamic monitoring system for farmland moisture that supports the fusion of air-space-ground sensors, comprising: Anchor point information acquisition module: acquires in-situ farmland moisture data collected by the ground anchor point monitoring layer, wherein the in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data; Low-altitude information acquisition module: Based on the in-situ data of farmland moisture, trigger the low-altitude patrol layer to perform dynamic patrol tasks in order to obtain low-altitude hyperspectral images covering the area surrounding the ground anchor monitoring layer; Remote sensing image acquisition module: Based on the low-altitude hyperspectral image and the in-situ farmland moisture data, it generates low-altitude moisture distribution data, and uses the low-altitude moisture distribution data as a constraint boundary to drive the high-altitude remote sensing layer to perform adaptive partitioning interpretation on the acquired satellite remote sensing image to generate high-altitude moisture distribution data. Data merging module: Merges the low-altitude moisture distribution data with the high-altitude moisture distribution data to generate a dynamic distribution field of farmland moisture with consistent expression between air, space, and ground.

[0011] This invention provides a method and system for dynamic monitoring of farmland moisture that supports the fusion of air-space-ground sensors, and has the following beneficial effects: This invention segments satellite remote sensing images by extracting the water content change boundaries from low-altitude moisture distribution data. It then applies matching interpretation methods to interpretation blocks with different average moisture content, ensuring that upper-altitude moisture distribution data achieves appropriate interpretation accuracy across different moisture states. During data merging, the low-altitude moisture distribution data is adjusted to the same spatial unit size as the upper-altitude moisture distribution data, and cross-correlation calculations are used to determine the positional offset between their spatial distribution patterns. Based on this offset, the upper-altitude moisture distribution data undergoes spatial position adjustment before weighted fusion. This eliminates spatial misalignment caused by differences in observation geometry and geographic registration errors, resulting in a seamless connection of multi-layered observation data in the final generated dynamic distribution field of farmland moisture. Attached Figure Description

[0012] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Please see Figure 1 This invention provides a method for dynamic monitoring of farmland moisture that supports the fusion of air-space-ground sensors, comprising the following steps: S1. Obtain in-situ farmland moisture data collected by the ground anchor monitoring layer, wherein the in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data. Furthermore, in S1, in-situ farmland moisture data collected by the ground anchor monitoring layer is obtained. The in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data. The continuous time-series soil moisture data is used as the first type of in-situ data, and the continuous time-series crop stem flow data is used as the second type of in-situ data. The first trend extraction process is performed on the first type of in-situ data. The first trend extraction process includes: calculating the difference between soil moisture values ​​at adjacent sampling times; determining the baseline fluctuation reference value by the quantile value of the absolute value of the adjacent difference in the first type of in-situ data sequence recorded on the same day; determining that the soil moisture is in a stable stage when the absolute value of the adjacent difference is continuously less than the baseline fluctuation reference value and the number of consecutive occurrences reaches the first number determined by the number of stable continuous sampling points on the same day; determining that the soil moisture is in an increasing stage when the number of consecutive positive values ​​of the adjacent difference and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the baseline fluctuation reference value and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the baseline fluctuation reference value and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the third number determined by the number of consecutive decreasing sampling points on the same day; The second trend extraction process is performed on the second type of in-situ data. The second trend extraction process includes: smoothing the stem flow rate value sequence to obtain smooth stem flow rate values, calculating the smoothing difference between smooth stem flow rate values ​​at adjacent time points, determining the stem flow background fluctuation reference value based on the quantile value of the absolute value of the smoothing difference recorded on the same day, determining that the stem flow rate is in a stable stage when the absolute value of the smoothing difference is continuously less than the stem flow background fluctuation reference value and the number of consecutive times reaches the fourth number determined by the number of continuous sampling points for stable stem flow on the same day, determining that the stem flow rate is in an enhancing stage when the number of consecutive times the smoothing difference is continuously positive and the absolute value is continuously greater than the stem flow background fluctuation reference value reaches the fifth number determined by the number of continuous sampling points for enhanced stem flow, and determining that the stem flow rate is in a weakening stage when the number of consecutive times the smoothing difference is continuously negative and the absolute value is continuously greater than the stem flow background fluctuation reference value reaches the sixth number determined by the number of continuous sampling points for weakened stem flow. The soil moisture stage and the stem flow rate stage are compared time intervals. When the stem flow rate enters the enhancement stage while the soil moisture is in the stable stage, the corresponding time interval is determined as the data difference period. The patrol initiation time and patrol spatial range of the low-altitude patrol layer are determined based on the data difference period. The patrol initiation time is obtained by adding the start time of the data difference period and the response delay duration. The patrol spatial range is formed by expanding the radiation radius in all directions based on the ground anchor monitoring point where the data difference period occurs. The radiation radius is dynamically selected based on the difference in the slope of change between the first type of in-situ data and the second type of in-situ data during the data difference period, using either the first radius value or the second radius value. The first radius value is greater than the second radius value.

[0015] It should be noted that continuous soil moisture time series data refers to a one-dimensional data sequence that reflects the change of soil pore moisture content over time, obtained by a dielectric constant type sensor buried in the soil layer of the crop root zone according to the first sampling interval. Each data point in the data sequence corresponds to the soil volume water content value at a sampling time. Crop stem flow continuous time series data refers to a one-dimensional data sequence that reflects the change of the upward flux of sap flow inside the crop stem over time, obtained by continuously recording data at a second sampling interval using a differential thermal probe wrapped around the outer periphery of the crop stem. Each data point in the data sequence corresponds to the stem flow rate value at a sampling time. The first and second sampling intervals can be set to the same sampling frequency, or they can be set to different sampling frequencies according to the differences in the lag characteristics of soil moisture changes and the sensitivity characteristics of stem flow changes. Since soil moisture changes usually lag behind changes in crop transpiration demand, there is a response time difference between continuous soil moisture time series data and continuous crop stem flow time series data, which is determined by both crop physiological processes and soil physical processes. Therefore, continuous soil moisture time series data is defined as the first type of in-situ data, and continuous crop stem flow time series data is defined as the second type of in-situ data. The time-series comparison operation between the first type of in-situ data and the second type of in-situ data includes a first trend extraction process for the first type of in-situ data and a second trend extraction process for the second type of in-situ data, as well as a comparison and judgment based on the trend extraction results. The first trend extraction process is as follows: obtain the soil moisture value at the current sampling time and the soil moisture value at the previous sampling time adjacent to the current sampling time in the first type of in-situ data sequence, and calculate the difference between the two. Obtain the daily sequence of all first-type in-situ data recorded at the monitoring point on that day, calculate the absolute value of the difference between adjacent sampling points in the daily sequence, sort all the obtained absolute values ​​in ascending order, and take the absolute value of the first percentile number in the sorted sequence as the reference value of the background fluctuation of the monitoring point on that day. The first percentile number is determined by the product of the total number of sampling points in the daily sequence and the first proportional coefficient and rounded down. When the absolute value of the difference between adjacent time points is less than the baseline fluctuation reference value, it is considered that no substantial change has occurred. When the absolute value of the difference between adjacent time points is continuously less than the baseline fluctuation reference value and the number of consecutive occurrences reaches the first consecutive number, it is determined that the soil moisture is in a stable stage. The first consecutive number is determined by rounding down the average number of consecutive sampling points in the stable state recorded at the monitoring point on that day. When the number of consecutive positive values ​​between adjacent time points and the number of consecutive times the absolute value of the difference between adjacent time points is greater than the baseline fluctuation reference value reaches the second consecutive number, it is determined that the soil moisture is in the rising stage. The second consecutive number is determined by rounding down the average number of consecutive positive sampling points during the historical rising process of this monitoring point. When the number of consecutive negative values ​​between adjacent time points and the number of consecutive times when the absolute value of the difference between adjacent time points is greater than the baseline fluctuation reference value reaches the third consecutive number, it is determined that the soil moisture is in a declining phase. The third consecutive number is determined by rounding down the average number of consecutive negative sampling points during the historical decline process of this monitoring point. The second trend extraction process is as follows: the arithmetic mean of the stem flow rate values ​​of several consecutive sampling points, including the current sampling time, is calculated as the smoothed stem flow rate value at the current time, and the smoothing difference between the smoothed stem flow rate value at the current time and the smoothed stem flow rate value at the previous time is calculated. Obtain the daily smoothed difference sequence consisting of all the smoothed differences recorded at the monitoring point on that day. Calculate the ascending order of the absolute values ​​of each smoothed difference in the daily smoothed difference sequence. Take the absolute value of the second percentile in the sequence as the reference value for the stem flow background fluctuation of the monitoring point on that day. The second percentile is determined by rounding down the product of the total number of data points in the daily smoothed difference sequence and the second proportional coefficient. When the absolute value of the smoothing difference is less than the reference value of the stem flow background fluctuation and the number of consecutive occurrences of this state reaches the fourth consecutive number, the stem flow rate is determined to be in a stable stage. The fourth consecutive number is determined by rounding down the average number of continuous sampling points of the stem flow stable state recorded at this monitoring point on the same day. When the number of consecutive positive smoothing differences and the absolute value of the smoothing differences being greater than the reference value of the stem flow background fluctuation reaches the fifth consecutive number, the stem flow rate is determined to be in the enhancement stage. The fifth consecutive number is determined by rounding down the average number of consecutive positive smoothing sampling points during the historical stem flow enhancement process at this monitoring point. When the number of consecutive negative smoothing differences and the absolute value of the smoothing difference being greater than the reference value of the stem flow background fluctuation reaches the sixth consecutive number, it is determined that the stem flow rate is in a weakening phase. The sixth consecutive number is determined by rounding down the average number of consecutive negative smoothing sampling points during the historical stem flow weakening process at this monitoring point. The soil moisture stage determined by the first trend extraction process is compared with the stem flow rate stage determined by the second trend extraction process on a time-by-time basis to identify the time intervals in which the two stages are inconsistent. When the stem flow rate has entered the enhancement stage while the soil moisture is still in the stable stage and has not shown a corresponding downward trend, it indicates that there has been a supply-demand deviation between the crop water transport demand and the soil water supply capacity. This time interval is determined as the data difference period. The output result of the determination of the data difference period includes the difference start time and the difference end time, or only the difference start time is output and continues until the next trend matching time. After determining the data difference period, the patrol initiation time and patrol space range of the low-altitude patrol layer were further determined based on the data difference period. The low-altitude patrol layer refers to the controllable flight vehicle equipped with hyperspectral imaging payload and its corresponding mission planning and execution link. It is normally in a standby state and only performs dynamic patrol missions when triggered. The patrol start time is determined by adding the start time of the data difference period to a response delay duration. The result is the patrol start time. The response delay duration refers to the time interval required from issuing the trigger command to the actual take-off and arrival of the low-altitude patrol vehicle in the target area. The time interval includes the command transmission time, the vehicle start-up self-check time, and the take-off and climb time. The patrol space range is determined as follows: taking the spatial coordinates of the specific monitoring point in the ground anchor monitoring layer during the period when data differences occur as the reference point, an expansion radius is extended in all directions to form a spatial envelope area centered on the reference point. The spatial envelope area is the patrol space range. When the trend deviation between the first type of in-situ data and the second type of in-situ data is large, it indicates that the degree of deviation between water supply and demand is high. In this case, a larger first radiation radius is used to cover a wider area. When the trend deviation is small, a smaller second radiation radius is used to focus on the area near the reference point; The specific method for quantifying the trend deviation is as follows: calculate the absolute value of the difference between the slope of change of the second type of in-situ data in the difference period and the slope of change of the first type of in-situ data in the same period; calculate the arithmetic mean of several sets of trend deviation values ​​recorded at the monitoring point on the same day; compare the absolute value of the difference obtained at the moment with the arithmetic mean; if it is less than the arithmetic mean, the second radiation radius is used; if it is greater than the arithmetic mean, the first radiation radius is used.

[0016] S2. Based on the in-situ data of farmland moisture, trigger the low-altitude patrol layer to perform a dynamic patrol task to obtain low-altitude hyperspectral images covering the area surrounding the ground anchor monitoring layer. Furthermore, in S2, the low-altitude patrol vehicle is controlled to fly according to the patrol space range and arrive at the first hovering position within the patrol space range. The first hovering position is selected at a spatial point that allows the hyperspectral imaging payload to cover the maximum effective area within the patrol space range in a single acquisition. At the first hovering position, the first deviation of the soil moisture value of the first type of in-situ data during the data difference period is calculated from the median value of the first type of in-situ data in the same period of the historical record. The second deviation of the stem flow rate value of the second type of in-situ data during the data difference period is calculated from the median value of the second type of in-situ data in the same period of the historical record. The first deviation and the second deviation are weighted and summed to obtain the comprehensive deviation value. The adjusted duration is obtained by multiplying the comprehensive deviation amplitude by the base duration. The base duration is the shortest integration time required for the hyperspectral imaging payload to obtain spectral data that meets the signal-to-noise requirements under standard conditions. The hyperspectral imaging payload is controlled to perform spectral data acquisition according to the adjusted duration, so as to obtain low-altitude hyperspectral image data at the first hovering position after duration adjustment.

[0017] It should be noted that the low-altitude patrol layer includes a controllable flight vehicle equipped with a hyperspectral imaging payload and a ground-based mission scheduling and data receiving unit. The controllable flight vehicle collects spectral data of ground objects by scanning or staring within the field of view of the hyperspectral imaging payload. Once the patrol space has been determined, the controllable flight vehicle takes off from the standby point and flies toward the target area. The flight control system of a controllable flight vehicle receives the boundary coordinate information of the spatial envelope region, generates a flight path from the current position to the interior of the spatial envelope region through the built-in route planning logic, and moves towards the target region along the flight path; Once the controllable flight vehicle enters the patrol space range, the flight control system selects a spatial point within the spatial envelope area as the first hovering position based on the swath width parameters of the hyperspectral imaging payload and the coverage requirements of the spatial envelope area. The principle for selecting the first hovering position is to ensure that the hyperspectral imaging payload can cover the maximum effective area within the spatial envelope area in one hovering acquisition operation. Typically, the first hovering position is set above the geometric center of the spatial envelope area. After the controllable flight vehicle reaches the first hovering position and completes hovering attitude stabilization, it enters the hyperspectral image acquisition stage. During the hyperspectral image acquisition process, the hyperspectral imaging payload performs spectral scanning of the ground target area line by line or frame by frame. Each line or frame of acquisition corresponds to a duration. The duration determines the integral of the spectral data in the time dimension, which directly affects the signal-to-noise level of the acquired spectral signal. In the conventional acquisition mode, the duration is a fixed value. However, in this scheme, the duration is dynamically adjusted according to the amplitude of the real-time in-situ data acquired during the data difference period. The real-time in-situ data amplitude refers to the degree to which the first type of in-situ data and the second type of in-situ data deviate from their respective reference levels during the data difference period. Obtain the soil moisture value of the first type of in-situ data during the data difference period, calculate the first deviation between the soil moisture value and the median value of the first type of in-situ data in the same period of the historical record, obtain the stem flow rate value of the second type of in-situ data during the data difference period, calculate the second deviation between the stem flow rate value and the median value of the second type of in-situ data in the same period of the historical record, and perform a weighted summation of the first deviation and the second deviation, and use the result as the comprehensive deviation value. The larger the overall deviation amplitude, the more significant the deviation between crop water supply and demand. In this case, a longer duration is needed to obtain higher quality spectral signals, thereby improving the accuracy of subsequent low-altitude water distribution data. The smaller the overall deviation amplitude, the less significant the deviation between crop water supply and demand. In this case, a shorter duration can meet the spectral signal quality requirements and shorten the single-point acquisition time to improve patrol efficiency. The specific calculation method for the duration is as follows: multiply the overall deviation amplitude by a base duration. The product is the adjusted duration. The base duration is the shortest integration time required for the hyperspectral imaging payload to obtain spectral data that meets the signal-to-noise requirements under standard illumination conditions and standard ground reflection conditions. After completing spectral data acquisition at the first hovering position, the controllable flight vehicle determines whether it needs to move to the second hovering position for supplementary acquisition based on the coverage requirements of the patrol space. If the patrol space exceeds the ground area that the hyperspectral imaging payload can cover in a single acquisition at the first hovering position, the flight control system calculates the spatial coordinates of the second hovering position based on the positional relationship between the boundary of the area covered by the first hovering position and the remaining uncovered area of ​​the patrol space. The controllable flight vehicle then moves from the first hovering position to the second hovering position and repeats the process of stabilizing the hovering attitude, adjusting the duration based on the real-time in-situ data amplitude, and acquiring spectral data until the entire area within the patrol space is covered by the effective field of view of the hyperspectral imaging payload. After acquisition at all hovering positions, the spectral data acquired at each hovering position are stitched together according to their corresponding spatial coordinates to form a low-altitude hyperspectral image with continuous spectral information covering the entire patrol space.

[0018] S3. Based on the low-altitude hyperspectral image and the in-situ farmland moisture data, generate low-altitude moisture distribution data, and use the low-altitude moisture distribution data as a constraint boundary to drive the upper-altitude remote sensing layer to perform adaptive partitioning interpretation on the acquired satellite remote sensing image to generate upper-altitude moisture distribution data. Furthermore, in S3, the spectral values ​​of the water response band are extracted pixel by pixel from the low-altitude hyperspectral image. The extracted spectral values ​​are corrected based on the in-situ data of farmland moisture to determine the pixel-level moisture content value. The pixel-level moisture content value is then filled into the surrounding area to form low-altitude moisture distribution data. The boundary where the difference in moisture content between adjacent pixels in the low-altitude moisture distribution data is greater than the boundary judgment reference value is extracted as the boundary of moisture content change. The boundary judgment reference value is determined by the quantile value of the absolute value of the difference in moisture content of all pixels in the low-altitude moisture distribution data. The boundary of moisture content change is mapped onto satellite remote sensing imagery, and the satellite remote sensing imagery is segmented to generate multiple non-overlapping interpretation blocks; For each interpretation block, the arithmetic mean of the moisture content values ​​of the low-altitude moisture distribution data within the corresponding range of the block is calculated. The arithmetic mean is compared with the first and second content thresholds. If the arithmetic mean is less than the first content threshold, the first interpretation processing method is used to interpret the moisture content of the satellite remote sensing image pixels within the block. If the arithmetic mean is greater than the second content threshold, the second interpretation processing method is used to interpret the moisture content. If the arithmetic mean is between the first and second content thresholds, the third interpretation processing method is used to interpret the moisture content. The moisture content values ​​obtained from the interpretation of each block are spliced ​​together according to their spatial location to form upper-air moisture distribution data.

[0019] It should be noted that the spectral values ​​of the moisture response band are extracted pixel by pixel from the low-altitude hyperspectral image. The low-altitude hyperspectral image is composed of multiple consecutive bands of spectral data stacked together. Each pixel records a spectral response value in each band. This spectral response value represents the reflection or radiation intensity of the corresponding ground object under electromagnetic wave illumination in that band. The moisture response band refers to a specific wavelength range in the electromagnetic spectrum where moisture exhibits characteristic absorption or reflection attenuation of incident energy. Within this wavelength range, there is a correspondence between the spectral response value of the ground object and its water content. By traversing each pixel in the low-altitude hyperspectral image, for... For each traversed pixel, the spectral response value located at the center wavelength of the moisture response band is read from its corresponding spectral curve. The spectral response value is the spectral value of the moisture response band of that pixel. During the traversal, the spectral value of a reference band outside the moisture response band can also be read at the same time. The reference band is located at the wavelength position where the moisture absorption characteristics are not significant but the brightness of ground objects is sensitive. After the traversal is completed, each pixel in the low-altitude hyperspectral image is assigned a moisture response band spectral value, forming a moisture response band spectral value array with the same spatial resolution as the low-altitude hyperspectral image. Based on in-situ farmland moisture data, spectral values ​​are corrected to determine pixel-level moisture content values. The first type of in-situ data in the farmland moisture data provides the true measured value of soil moisture content at ground anchor monitoring points, and the second type of in-situ data provides the true measured value of crop stem flow rate at ground anchor monitoring points. Together, these two types of in-situ data reflect the true moisture status at the point. In low-altitude hyperspectral images, locate anchor pixels corresponding to the spatial coordinates of ground anchor monitoring points, extract the spectral values ​​of the moisture response band of the anchor pixels, and then expand a neighboring pixel window around the anchor pixel, extract the spectral values ​​of the moisture response band of all pixels in the neighboring pixel window and calculate the spectral mean of the window. The first type of in-situ data and the second type of in-situ data at the anchor monitoring point are converted into a comprehensive measured value of water according to their respective conversion factors. The conversion factors are determined according to the crop type and growth stage. They are used to unify the soil moisture content and stem flow rate to the same water state dimension. The ratio between the comprehensive measured value of water and the mean value of the window spectrum is calculated. The ratio is the spectral moisture conversion ratio at the anchor monitoring point. The spectral moisture conversion ratio is applied to pixels other than anchor pixels in low-altitude hyperspectral images. That is, for any non-anchor pixel in the image, its moisture response band spectral value is multiplied by the spectral moisture conversion ratio, and the resulting product is the pixel-level moisture content value of that pixel. If the ground anchor monitoring layer contains multiple monitoring points, the above calculation is performed for each monitoring point to obtain its own spectral moisture conversion ratio. Then, a spatial interpolation method is used to generate a spectral moisture conversion ratio distribution surface covering the entire image range. For any non-anchor pixel, the corresponding interpolation conversion ratio is obtained from the spectral moisture conversion ratio distribution surface based on its spatial coordinates, and then multiplied by its moisture response band spectral value to obtain the pixel-level moisture content value. After determining the moisture content values ​​of all pixels, these pixel-level moisture content values ​​are arranged according to the original row and column positions of each pixel in the low-altitude hyperspectral image to form a two-dimensional numerical array with the same spatial range and spatial resolution as the low-altitude hyperspectral image. Each numerical element in the two-dimensional numerical array corresponds to the estimated moisture content value of a spatial location. The two-dimensional numerical array is the low-altitude moisture distribution data. After generating low-altitude moisture distribution data, the low-altitude moisture distribution data is used as a constraint boundary to drive the upper-altitude remote sensing layer to perform adaptive partitioning interpretation on the acquired satellite remote sensing images to generate upper-altitude moisture distribution data. The moisture content change boundary in the low-altitude moisture distribution data is extracted. The moisture content change boundary refers to the location where the moisture content value in the low-altitude moisture distribution data changes significantly in space. These locations usually correspond to the dividing line between fields with different moisture states, the boundary of crop planting rows, or the edge of the irrigation wetting front. The extraction process is as follows: the two-dimensional numerical array composed of low-altitude moisture distribution data is scanned row by row and column by column. The absolute value of the difference between the moisture content values ​​of adjacent pixels is calculated. The absolute value of the difference is compared with a boundary judgment reference value, which is determined by the quantile value of the absolute value of the difference between the moisture content values ​​of all pixels in the low-altitude moisture distribution data. When the absolute value of the difference between adjacent pixels is greater than the boundary judgment reference value, the boundary between adjacent pixels is marked as a candidate boundary segment. All candidate boundary segments are connected according to spatial adjacency to form a continuous boundary line. The boundary lines are closed by extending any incompletely closed boundary lines along their endpoint tangents to the image edge or intersecting with other boundary lines, forming several closed moisture content change boundary lines. These closed moisture content change boundary lines divide the area covered by the low-altitude moisture distribution data into multiple sub-regions with relatively uniform internal moisture content. The satellite remote sensing image is segmented using the boundary of moisture content change to generate multiple non-overlapping interpretation blocks. The satellite remote sensing image is a wide-area raster image acquired by the upper-air remote sensing layer through a spaceborne multi-channel radiometer. Its coverage area is larger than that of the low-altitude moisture distribution data. Its pixel spectral information reflects the electromagnetic radiation characteristics of ground objects in multiple bands. The segmentation process is as follows: spatial registration is performed between low-altitude moisture distribution data and satellite remote sensing images. The closed moisture content change boundary line extracted from the low-altitude moisture distribution data is mapped to the pixel coordinate system of the satellite remote sensing images. During the mapping process, the spatial shape and relative position of the boundary line remain unchanged. In satellite remote sensing images, the mapped closed boundary lines are used as dividing lines to assign pixels located inside different closed boundary lines to different interpretation blocks, forming multiple non-overlapping interpretation blocks. Each interpretation block corresponds to a sub-region with relatively uniform internal moisture content in the low-altitude moisture distribution data. The moisture status of ground features within the interpretation block has a high degree of consistency, while the moisture status between different interpretation blocks has significant differences. For each interpretation block, an interpretation processing method corresponding to the average low-altitude moisture content within the block is adopted to perform independent interpretation of the moisture content within the block. For blocks with a low average low-altitude moisture content, the surface water content is low and the contribution of soil background in the spectral response is relatively high. In this case, the first interpretation processing method that focuses on soil moisture signal extraction is adopted. This method prioritizes the use of the first band combination that is sensitive to changes in soil moisture for moisture estimation. For blocks with high average low-altitude moisture content, sufficient surface water content, high vegetation cover, and a high contribution of crop canopy in the spectral response, a second interpretation processing method focusing on vegetation moisture signal extraction is adopted. This method prioritizes the use of second band combinations that are sensitive to canopy moisture content for moisture estimation. For blocks where the average low-altitude moisture content falls between the two mentioned above, a third interpretation processing method that takes into account both soil and vegetation signals is adopted. For each interpretation block, the arithmetic mean of the moisture content values ​​of all pixels within the corresponding range of the low-altitude moisture distribution data for that block is obtained. The arithmetic mean is compared with the first and second content thresholds. If it is less than the first content threshold, the first interpretation processing method is called; if it is greater than the second content threshold, the second interpretation processing method is called; if it is between the two, the third interpretation processing method is called. The first and second content thresholds are determined based on the quantile distribution of the low-altitude moisture distribution data across all pixels on the day. The selected interpretation processing method is used to calculate the moisture content of each pixel in the satellite remote sensing image within the interpretation block. After the calculation is completed, the moisture content values ​​of all pixels within the interpretation block are filled in according to their spatial location to form the upper-altitude moisture distribution sub-data corresponding to the interpretation block. By stitching together the upper-air moisture distribution sub-data of all the interpreted blocks according to their spatial location, we obtain upper-air moisture distribution data covering the entire range of satellite remote sensing imagery.

[0020] S4. Merge the low-altitude moisture distribution data and the high-altitude moisture distribution data to generate a dynamic distribution field of farmland moisture with consistent expression between air, space, and ground.

[0021] Furthermore, in step S4, the low-altitude moisture distribution data is adjusted to the same spatial cell size as the high-altitude moisture distribution data to obtain low-altitude moisture distribution resampled data. Within the spatial overlap range between low-altitude moisture distribution resampling data and high-altitude moisture distribution data, the spatial distribution of pixel moisture content values ​​in the low-altitude moisture distribution resampling data is taken as the first spatial distribution form, and the spatial distribution of pixel moisture content values ​​in the high-altitude moisture distribution data is taken as the second spatial distribution form. Cross-correlation calculations are performed on the first spatial distribution form and the second spatial distribution form along the row and column directions, respectively. The translation step corresponding to the maximum cross-correlation value in the row direction is taken as the east-west position offset component, and the translation step corresponding to the maximum cross-correlation value in the column direction is taken as the north-south position offset component. The east-west position offset component and the north-south position offset component together constitute the position offset. The row and column coordinates of all pixels in the upper-level moisture distribution data are translated according to the position offset, so that the second spatial distribution pattern of the upper-level moisture distribution data within the overlapping range is aligned with the first spatial distribution pattern after translation. Within the overlapping range, a weighted fusion is performed on the moisture content values ​​of the translated upper-level moisture distribution data pixels and the moisture content values ​​of the low-level moisture distribution resampled data pixels. Outside the overlapping range, the moisture content values ​​of their respective data sources are used. The fusion result is then stitched together with the result outside the overlapping range to generate a dynamic distribution field of farmland moisture.

[0022] It should be noted that the low-altitude moisture distribution data is adjusted to the same spatial cell size as the upper-altitude moisture distribution data. Spatial cell size refers to the scale and grid arrangement of the basic constituent units of the data in the spatial dimension. The spatial cell size parameters of the upper-altitude moisture distribution data are obtained, including the ground span size of the pixel in the east-west direction, the ground span size of the pixel in the north-south direction, and the starting coordinate offset of the pixel grid. Using the spatial cell size of the upper-altitude moisture distribution data as the target size, a resampling operation is performed on the low-altitude moisture distribution data. The resampling operation includes: within the spatial range of the low-altitude moisture distribution data, dividing the area into resampling grids corresponding one-to-one with the pixels of the high-altitude moisture distribution data according to the ground span size of the pixels in the target specification; for each resampling grid, acquiring all the original pixels covered by the grid in the low-altitude moisture distribution data, and arithmetically averaging the moisture content values ​​of these original pixels, using the arithmetic mean as the adjusted pixel moisture content value corresponding to the resampling grid; for edge pixels of the resampling grid that only partially cover the boundary of the low-altitude moisture distribution data, weighting the original pixel values ​​participating in the averaging according to the coverage area ratio; after completing the calculation of all resampling grids, low-altitude moisture distribution resampling data with the same pixel size, the same number of pixel rows and columns, and the same grid starting coordinates as the high-altitude moisture distribution data is formed. At this time, the low-altitude moisture distribution data and the high-altitude moisture distribution data have the same spatial unit specification; After completing the spatial unit specification adjustment, the positional offset between the first spatial distribution pattern of low-altitude moisture distribution data and the second spatial distribution pattern of high-altitude moisture distribution data is determined within the spatial overlap range. The spatial overlap range refers to the part of the geographical space where the low-altitude moisture distribution data coverage area and the high-altitude moisture distribution data coverage area overlap. The first spatial distribution pattern refers to the texture structure characteristics of the moisture content values ​​of low-altitude moisture distribution data within the overlapping range as a function of spatial location, specifically manifested in the spatial location, shape outline, and boundary orientation of high- and low-moisture content areas; the second spatial distribution pattern refers to the texture structure characteristics of the moisture content values ​​of high-altitude moisture distribution data within the same overlapping range as a function of spatial location. Due to differences in observation geometry, residual atmospheric correction errors, and slight deviations in geographic registration during the acquisition process of low-altitude moisture distribution data and high-altitude moisture distribution data, the spatial locations of the high-value and low-value areas of moisture content expressed by the two may be shifted or misaligned. This misalignment manifests as a positional shift between the first spatial distribution pattern and the second spatial distribution pattern. Determining the position offset includes: within the overlapping range, extracting pixel sequences with the same spatial location from the low-altitude moisture distribution resampled data and the high-altitude moisture distribution data respectively. Taking the first row of pixels as an example, the moisture content values ​​of all pixels in that row in the low-altitude moisture distribution resampled data are obtained to form the first row sequence, and the moisture content values ​​of all pixels in that row in the high-altitude moisture distribution data are obtained to form the second row sequence. Calculate the cross-correlation function between the first row sequence and the second row sequence. The cross-correlation function is obtained by shifting the second row sequence relative to the first row sequence by different step lengths and calculating the sum of the products of the corresponding positions of the two sequences after each shift. The shift step length corresponding to the maximum value of the cross-correlation function is the row offset in that row direction. Perform cross-correlation calculations on each row within the overlapping range to obtain the row offset corresponding to each row, and use the arithmetic mean of all row offsets as the position offset component in the east-west direction. The same method is used to perform cross-correlation calculations on each column within the overlapping range, and the arithmetic mean of the offsets of all columns is taken as the north-south position offset component. The east-west position offset component and the north-south position offset component together constitute the position offset between the first spatial distribution pattern and the second spatial distribution pattern. If there are cases where the upper-air moisture distribution data is partially invalid due to cloud cover or terrain shadows within the overlapping area, then only the cell column segments or row segments that are effectively covered by both sets of data will be used when performing cross-correlation calculations. After determining the position offset, spatial position adjustment is performed on the upper-air moisture distribution data based on the position offset to make the first spatial distribution pattern and the second spatial distribution pattern continuously connected. Spatial position adjustment refers to translating all pixels in the upper-air moisture distribution data according to the calculated position offset. After translation, the spatial distribution pattern of moisture content in the upper-air moisture distribution data within the overlapping range is aligned with the spatial distribution pattern of the lower-air moisture distribution data. For each pixel in the upper-air moisture distribution data, the east-west position offset component and the north-south position offset component are subtracted from its original row and column coordinates to obtain the adjusted row and column coordinates. The moisture content value of the pixel is assigned to the pixel position corresponding to the adjusted row and column coordinates. During the translation process, pixel values ​​that exceed the boundaries of the original data are discarded, and the newly generated vacant pixel positions are filled with the nearest valid pixel values. After spatial adjustment, the spatial positions of the high and low moisture content areas in the overlapping range of the upper-air moisture distribution data are aligned with the spatial positions of the corresponding feature areas in the lower-air moisture distribution data. After spatial location adjustment, the adjusted upper-air moisture distribution data and the low-air moisture distribution resampled data are merged to generate a dynamic distribution field of farmland moisture. Within the overlapping range, the pixel moisture content values ​​of the adjusted upper-air moisture distribution data and the pixel moisture content values ​​of the low-air moisture distribution resampled data are weighted and fused. The weight coefficient of the weighted fusion is determined by the quality evaluation factor of the two sets of data in their respective acquisition processes. The quality evaluation factor of the low-air moisture distribution data is the normalized signal-to-noise ratio of the low-air hyperspectral image at the corresponding pixel, and the quality evaluation factor of the upper-air moisture distribution data is the normalized atmospheric correction confidence value of the satellite remote sensing image at the corresponding pixel. In areas outside the overlapping range, if the area is only covered by low-altitude moisture distribution data and not by high-altitude moisture distribution data, the moisture content value of the low-altitude moisture distribution resampled data is directly used. If the area is only covered by upper-level moisture distribution data and not by lower-level moisture distribution data, the moisture content value of the adjusted upper-level moisture distribution data is directly adopted. The results of the fusion within the overlapping area and the results of the single data source outside the overlapping area are spliced ​​together according to spatial location to form a dynamic distribution field of farmland moisture that fully covers the farmland monitoring area and has a continuous and consistent expression in spatial location.

[0023] like Figure 2 In this embodiment of the invention, a dynamic monitoring system for farmland moisture that supports the fusion of air-space-ground sensors is also provided, comprising: Anchor point information acquisition module: acquires in-situ farmland moisture data collected by the ground anchor point monitoring layer, wherein the in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data; Low-altitude information acquisition module: Based on the in-situ data of farmland moisture, trigger the low-altitude patrol layer to perform dynamic patrol tasks in order to obtain low-altitude hyperspectral images covering the area surrounding the ground anchor monitoring layer; Remote sensing image acquisition module: Based on the low-altitude hyperspectral image and the in-situ farmland moisture data, it generates low-altitude moisture distribution data, and uses the low-altitude moisture distribution data as a constraint boundary to drive the high-altitude remote sensing layer to perform adaptive partitioning interpretation on the acquired satellite remote sensing image to generate high-altitude moisture distribution data. Data merging module: Merges the low-altitude moisture distribution data with the high-altitude moisture distribution data to generate a dynamic distribution field of farmland moisture with consistent expression between air, space, and ground.

[0024] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0025] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0026] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0027] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0028] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the described technical solution.

Claims

1. A method for dynamic monitoring of farmland moisture supporting the fusion of air-space-ground sensors, characterized in that, Includes the following steps: S1. Obtain in-situ farmland moisture data collected by the ground anchor monitoring layer, wherein the in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data. S2. Based on the in-situ data of farmland moisture, trigger the low-altitude patrol layer to perform a dynamic patrol task to obtain low-altitude hyperspectral images covering the area surrounding the ground anchor monitoring layer. S3. Based on the low-altitude hyperspectral image and the in-situ farmland moisture data, generate low-altitude moisture distribution data, and use the low-altitude moisture distribution data as a constraint boundary to drive the upper-altitude remote sensing layer to perform adaptive partitioning interpretation on the acquired satellite remote sensing image to generate upper-altitude moisture distribution data. S4. Merge the low-altitude moisture distribution data and the high-altitude moisture distribution data to generate a dynamic distribution field of farmland moisture with consistent expression between air, space, and ground.

2. The method for dynamic monitoring of farmland moisture supporting the fusion of air-space-ground sensors according to claim 1, characterized in that, In step S1, in-situ farmland moisture data collected by the ground anchor monitoring layer is obtained. The in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data. The continuous time-series soil moisture data is used as the first type of in-situ data, and the continuous time-series crop stem flow data is used as the second type of in-situ data. The first trend extraction process is performed on the first type of in-situ data. The first trend extraction process includes: calculating the difference between soil moisture values ​​at adjacent sampling times; determining the baseline fluctuation reference value by the quantile value of the absolute value of the adjacent difference in the first type of in-situ data sequence recorded on the same day; determining that the soil moisture is in a stable stage when the absolute value of the adjacent difference is continuously less than the baseline fluctuation reference value and the number of consecutive occurrences reaches the first number determined by the number of stable continuous sampling points on the same day; determining that the soil moisture is in an increasing stage when the number of consecutive positive values ​​of the adjacent difference and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the baseline fluctuation reference value and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the baseline fluctuation reference value and the number of consecutive occurrences of the absolute value of the adjacent difference are greater than the third number determined by the number of consecutive decreasing sampling points on the same day; The second trend extraction process is performed on the second type of in-situ data. The second trend extraction process includes: smoothing the stem flow rate value sequence to obtain smooth stem flow rate values, calculating the smoothing difference between smooth stem flow rate values ​​at adjacent time points, determining the stem flow background fluctuation reference value based on the quantile value of the absolute value of the smoothing difference recorded on the current day, determining that the stem flow rate is in a stable stage when the absolute value of the smoothing difference is continuously less than the stem flow background fluctuation reference value for a continuous number of times, as determined by the number of continuous sampling points for stable stem flow on the current day; determining that the stem flow rate is in an enhancing stage when the smoothing difference is continuously positive and the absolute value is continuously greater than the stem flow background fluctuation reference value for a continuous number of times, as determined by the number of continuous sampling points for enhanced stem flow; and determining that the stem flow rate is in a weakening stage when the smoothing difference is continuously negative and the absolute value is continuously greater than the stem flow background fluctuation reference value for a continuous number of times, as determined by the number of continuous sampling points for weakened stem flow.

3. The method for dynamic monitoring of farmland moisture supporting the fusion of air-space-ground sensors according to claim 2, characterized in that, The soil moisture stage and the stem flow rate stage are compared time intervals. When the stem flow rate enters the enhancement stage while the soil moisture is in the stable stage, the corresponding time interval is determined as the data difference period. The patrol initiation time and patrol spatial range of the low-altitude patrol layer are determined based on the data difference period. The patrol initiation time is obtained by adding the start time of the data difference period and the response delay duration. The patrol spatial range is formed by expanding the radiation radius in all directions based on the ground anchor monitoring point where the data difference period occurs. The radiation radius is dynamically selected based on the difference in the slope of change between the first type of in-situ data and the second type of in-situ data during the data difference period, using either the first radius value or the second radius value. The first radius value is greater than the second radius value.

4. The method for dynamic monitoring of farmland moisture supporting the fusion of air-space-ground sensors according to claim 1, characterized in that, In S2, the low-altitude patrol vehicle is controlled to fly according to the patrol space range and arrive at the first hovering position within the patrol space range. The first hovering position is selected at a spatial point that allows the hyperspectral imaging payload to cover the maximum effective area within the patrol space range in a single acquisition. At the first hovering position, the first deviation of the soil moisture value of the first type of in-situ data during the data difference period is calculated from the median value of the first type of in-situ data in the same historical period on the same day. The second deviation of the stem flow rate value of the second type of in-situ data during the data difference period is calculated from the median value of the second type of in-situ data in the same historical period on the same day. The first deviation and the second deviation are weighted and summed to obtain the comprehensive deviation value.

5. A method for dynamic monitoring of farmland moisture supporting air-space-ground sensor fusion according to claim 4, characterized in that, The adjusted duration is obtained by multiplying the comprehensive deviation amplitude by the base duration. The base duration is the shortest integration time required for the hyperspectral imaging payload to obtain spectral data that meets the signal-to-noise requirements under standard conditions. The hyperspectral imaging payload is controlled to perform spectral data acquisition according to the adjusted duration, so as to obtain low-altitude hyperspectral image data at the first hovering position after duration adjustment.

6. The method for dynamic monitoring of farmland moisture supporting the fusion of air-space-ground sensors according to claim 1, characterized in that, In step S3, the spectral values ​​of the water response band are extracted pixel by pixel from the low-altitude hyperspectral image. The extracted spectral values ​​are corrected based on the in-situ data of farmland water to determine the pixel-level water content value. The pixel-level water content value is then filled into the surrounding area to form low-altitude water distribution data. The boundary where the difference in moisture content between adjacent pixels in the low-altitude moisture distribution data is greater than the boundary judgment reference value is extracted as the boundary of moisture content change. The boundary judgment reference value is determined by the quantile value of the absolute value of the difference in moisture content of all pixels in the low-altitude moisture distribution data. The boundary of moisture content change is mapped onto satellite remote sensing imagery, and the satellite remote sensing imagery is segmented to generate multiple non-overlapping interpretation blocks.

7. A method for dynamic monitoring of farmland moisture supporting air-space-ground sensor fusion according to claim 6, characterized in that, For each interpretation block, the arithmetic mean of the moisture content values ​​of the low-altitude moisture distribution data within the corresponding range of the block is calculated. The arithmetic mean is compared with the first and second content thresholds. If the arithmetic mean is less than the first content threshold, the first interpretation processing method is used to interpret the moisture content of the satellite remote sensing image pixels within the block. If the arithmetic mean is greater than the second content threshold, the second interpretation processing method is used to interpret the moisture content. If the arithmetic mean is between the first and second content thresholds, the third interpretation processing method is used to interpret the moisture content. The moisture content values ​​obtained from the interpretation of each block are spliced ​​together according to their spatial location to form upper-air moisture distribution data.

8. The method for dynamic monitoring of farmland moisture supporting the fusion of air-space-ground sensors according to claim 1, characterized in that, In step S4, the low-altitude moisture distribution data is adjusted to the same spatial cell size as the high-altitude moisture distribution data to obtain low-altitude moisture distribution resampling data. Within the spatial overlap range of low-altitude moisture distribution resampling data and high-altitude moisture distribution data, the spatial distribution of pixel moisture content values ​​in the low-altitude moisture distribution resampling data is taken as the first spatial distribution form, and the spatial distribution of pixel moisture content values ​​in the high-altitude moisture distribution data is taken as the second spatial distribution form. Cross-correlation calculations are performed on the first spatial distribution form and the second spatial distribution form along the row and column directions, respectively. The translation step corresponding to the maximum cross-correlation value in the row direction is taken as the east-west position offset component, and the translation step corresponding to the maximum cross-correlation value in the column direction is taken as the north-south position offset component. The east-west position offset component and the north-south position offset component together constitute the position offset.

9. A method for dynamic monitoring of farmland moisture supporting air-space-ground sensor fusion according to claim 8, characterized in that, The row and column coordinates of all pixels in the upper-level moisture distribution data are translated according to the position offset, so that the second spatial distribution pattern of the upper-level moisture distribution data within the overlapping range is aligned with the first spatial distribution pattern after translation. Within the overlapping range, a weighted fusion is performed on the moisture content values ​​of the translated upper-level moisture distribution data pixels and the moisture content values ​​of the low-level moisture distribution resampled data pixels. Outside the overlapping range, the moisture content values ​​of their respective data sources are used. The fusion result is then stitched together with the result outside the overlapping range to generate a dynamic distribution field of farmland moisture.

10. A farmland moisture dynamic monitoring system supporting air-space-ground sensor fusion for executing the farmland moisture dynamic monitoring method supporting air-space-ground sensor fusion as described in any one of claims 1-9, characterized in that, include: Anchor point information acquisition module: acquires in-situ farmland moisture data collected by the ground anchor point monitoring layer, wherein the in-situ farmland moisture data includes continuous time-series soil moisture data and continuous time-series crop stem flow data; Low-altitude information acquisition module: Based on the in-situ data of farmland moisture, trigger the low-altitude patrol layer to perform dynamic patrol tasks in order to obtain low-altitude hyperspectral images covering the area surrounding the ground anchor monitoring layer; Remote sensing image acquisition module: Based on the low-altitude hyperspectral image and the in-situ farmland moisture data, it generates low-altitude moisture distribution data, and uses the low-altitude moisture distribution data as a constraint boundary to drive the high-altitude remote sensing layer to perform adaptive partitioning interpretation on the acquired satellite remote sensing image to generate high-altitude moisture distribution data. Data merging module: Merges the low-altitude moisture distribution data with the high-altitude moisture distribution data to generate a dynamic distribution field of farmland moisture with consistent expression between air, space, and ground.

Images