Southern irrigated area fruit tree irrigation method based on space-air-ground integrated monitoring

Abstract

The invention provides a southern irrigation area fruit tree irrigation method based on space-air-ground integrated monitoring. The southern irrigation area fruit tree irrigation method comprises the steps that space-air-ground facility equipment is used for collecting data of the surface temperature, humidity, fruit tree transpiration intensity and soil moisture content of an irrigation area; constructing a time sequence fusion data set for the space-air-ground multi-source data, and generating a high-precision soil moisture content distribution map; day-by-day weather and fruit tree growth data are obtained, and subarea division is carried out on the fruit tree park according to soil texture and fruit tree plant distribution; constructing a fruit tree water demand model, constructing an LSTM rainfall forecasting model, and constructing an irrigation amount forecasting model based on a water amount balance principle; a decision-making system of feedforward prediction and feedback decision-making is constructed, and an irrigation decision-making instruction is issued to a water taking valve attached to the high-level water tank for irrigation by means of an internet-of-things platform; analyzing the water stress condition of fruit tree canopies according to an unmanned aerial vehicle thermal infrared imaging technology, and optimizing a fruit tree water demand model; and establishing an irrigation expert knowledge base, and iteratively updating an irrigation strategy.

Description

Technical Field

[0001] The present invention relates to the field of precision irrigation technology, and in particular to a fruit tree irrigation method in southern irrigation areas based on integrated air-space-ground-space monitoring. Background Art

[0002] Fruit tree planting in the southern irrigation areas is affected by objective conditions such as uneven temporal and spatial distribution of water resources, uneven soil texture, and uneven water distribution from water conservancy projects. The traditional experience-based irrigation model has led to water resource waste and secondary soil salinization to a certain extent. In addition, soil moisture equipment can only reflect soil temperature and humidity information within a certain range, which is difficult to accurately match with the water demand dynamics of crops over a large area.

[0003] At present, air-space-ground sensors and related technologies are mostly used independently, with insufficient data fusion, making it difficult to achieve long-term, multi-scale, and multi-dimensional collaborative analysis. In addition, the uneven temporal and spatial distribution of rainfall in the south and the alternation of rainy and dry seasons can easily cause plant root corruption and water stress problems, and existing technical methods lack dynamic adaptability.

[0004] Although the existing patent document CN202410240500.6 discloses "a technical networking method for integrated air-space-ground field management of crops", this method uses integrated air-space-ground technology to collect crop growth environment, growth status, meteorological and soil data, and adopts drones and ground facilities for zoning management based on the above data analysis to regulate agriculture and take effect evaluation, but it does not explicitly mention the irrigation area water demand model and how to make irrigation decisions. The implementation examples are relatively macro, and there are difficulties in the specific implementation of the plan.

[0005] For another example, the existing patent document CN202211440918.9 discloses "a digital village construction system based on the fusion of multi-source data from air, space, and land", which builds a cloud database, a farmland assessment module, a field disaster warning module and a comprehensive processing platform to integrate, store, analyze, evaluate, warn and dispose of farmland health status information to realize the intelligence of digital villages. However, it focuses more on pest control and farmland health status assessment, and lacks in-depth exploration of how to deal with the uneven temporal and spatial distribution of rainfall and more efficient use of water resources. Summary of the Invention

[0006] In response to the above technical problems, the present invention proposes a fruit tree irrigation method in the southern irrigation area based on integrated air-space-ground monitoring. This method focuses on fruit tree irrigation work, forms differentiated results in the field of analyzing fruit tree water stress, and provides a certain reference value for fruit tree irrigation; this method proposes a reference implementation path for fruit tree water-saving irrigation, which is easy to implement and easy to form irrigation standardization and process-based irrigation guidance, which can be effectively implemented during the work process and improve work efficiency; this method can be well applied to agricultural irrigation scenarios in mountainous and hilly terrain in southern irrigation areas and uneven temporal and spatial distribution of rainfall.

[0007] In order to achieve the above technical features, the purpose of the present invention is to achieve the following: a fruit tree irrigation method in the southern irrigation area based on integrated air-space-ground monitoring, comprising: Step 1, data collection: Use air-ground and ground-based equipment to collect data on surface temperature, humidity, transpiration intensity of fruit trees, and soil moisture in the irrigation area; Step 2: Generate soil moisture distribution map: Construct a simultaneous fusion dataset of multi-source data from space, ground and space to generate a high-precision soil moisture distribution map; Step 3, sub-area division: Obtain daily meteorological and fruit tree growth data, and divide the orchard into sub-areas based on soil texture and fruit tree plant distribution; Step 4: Construction of irrigation decision model and irrigation: The Hargreaves-Samani formula was used to construct a fruit tree water demand model, and an LSTM rainfall forecast model was constructed using weather type and rainfall intensity. Based on the water balance principle, an irrigation volume forecast model was constructed. A "feedforward prediction + feedback decision" decision-making system was constructed, using the rainfall forecast model as the feedforward prediction layer. An irrigation decision model was constructed using random forests to form a feedback optimization layer. Irrigation decision instructions were then issued to the water intake valves assigned to the high-level water tank via the Internet of Things platform for irrigation. Step 5: Water demand model optimization: Analyze the water stress of fruit tree canopies using drone thermal infrared imaging technology and optimize the fruit tree water requirement model; Step 6: Establish an irrigation expert knowledge base and iteratively update the irrigation strategy.

[0008] Preferably, the data collection process in step 1 specifically includes: Use satellite remote sensing data to obtain the inversion results of irrigation area vegetation index, surface temperature and humidity; Using drones equipped with high-performance multispectral cameras and thermal infrared imaging cameras, the canopy temperature and plant growth conditions in the fruit tree planting area are obtained in sync with the satellite remote sensing cycle; The soil moisture station was used to collect root moisture content, temperature and humidity data.

[0009] Preferably, the satellite remote sensing data source uses Landsat-8 satellite data to perform vegetation index, surface temperature, and humidity inversion; Data preprocessing for Landsat-8 satellite data includes radiometric calibration, atmospheric correction, orthorectification, image fusion, image mosaicking, and image cropping; Normalized Difference Vegetation Index NDVI Calculate vegetation index, eliminate radiation error, and enhance the ability to respond to vegetation. NDVI The calculation formula is: ; Where: NIR is the reflectance value in the near-infrared band, R is the reflectance value of the infrared band; Get the surface emissivity, the calculation formula is: ; Get the blackbody radiation brightness, the calculation formula is: ; Where: X is the surface emissivity, b This is the radiation calibration result of Band 10; Get the surface temperature, which is calculated as: ; Constructed using multispectral bands WET Index, reflecting the moisture content of the surface humidity, is calculated as follows: ; In the formula, B, G, and R correspond to the B2, B3, and B4 bands of Landsat8 respectively. NIR It is the B5 band of Landsat8. SWIR 1 and SWIR 2 is the B6 and B7 bands of Landsat 8.

[0010] Preferably, a multispectral drone is used to obtain images of fruit trees in the same time sequence as Landsat 8 satellite images, which are interpreted by the drone platform, and the canopy temperature of the fruit trees is inverted based on a thermal infrared imaging camera. The time domain method was used to collect root moisture content, temperature and humidity data of fruit tree plants using soil moisture monitoring stations.

[0011] Preferably, the generation of the soil moisture distribution map in step 2 specifically includes: The Landsat8 satellite data with a certain spatial resolution is fused with the drone data with a certain spatial resolution to generate a high-precision soil moisture distribution map. Based on comprehensive consideration of topography, hydrology, and vegetation natural resource factors, geographically weighted regression was used to spatially downscale the Landsat8 satellite data with a pixel base value of 1 meter. Kriging interpolation was performed on the drone data to generate 1-meter spatial resolution image data and data fusion was performed to generate a high-temporal and spatial resolution and high-precision soil moisture distribution map. The geographically weighted regression calculation formula is as follows: ; Where, For the locationi The value of the dependent variable at ; For location i The value of the independent variable at k =1,2,..., m ;( u i , v i ) is the regression analysis point i coordinates; is the intercept term; is the regression analysis coefficient, k =1,2,..., m ; is a random error; The Kriging interpolation calculation formula is as follows: ; Where, is the value at some observation point x0, is the weight coefficient, The attribute value of the sampling point.

[0012] Preferably, the sub-area division in step 3 specifically includes: Collect the past 20 years of daily historical meteorological data from local stations from the local meteorological department or the China Meteorological Data Network, including minimum temperature, maximum temperature, average temperature, average wind speed, sunshine hours, and average relative humidity. Also obtain weather forecast data with a 7-day forecast period for the past 10 years from Caiyun Meteorological Service, including maximum temperature, minimum temperature, weather type, and wind speed. Obtain data on fruit tree water requirements, irrigation area soil parameters, and field irrigation water utilization coefficients from local representative experimental stations; wherein the irrigation area soil parameters include soil type, field water holding capacity, wilting coefficient, and deep percolation; Irrigation zoning is carried out according to 45-50 fruit trees with the same or similar field water holding capacity as the most basic irrigation unit.

[0013] Preferably, in step 4: Considering that the Hargreaves-Samani model has high calculation accuracy in all climate zones, the Hargreaves-Samani formula method is used to construct the fruit tree water requirement model, and the root mean square error is used to evaluate the model prediction accuracy. The Hargreaves-Samani expression is: ; Where: Calculated by the Hargreaves-Samani formula ET 0 value; and The recommended values ​​of the two parameters are 0.0023 and 0.5 respectively; is extraterrestrial radiation, MJ·m -2 / sky; is the latent heat of evaporation at the average temperature, which is 2.45 MJ·kg -1 ; and are the maximum and minimum temperatures, respectively, in degrees Celsius; Feature encoding is performed based on weather type and rainfall intensity. The interactive features of weather type and rainfall intensity are coupled using LSTM to build a rainfall forecast model, and residual analysis is used to optimize model performance. Based on the water balance formula, an irrigation amount forecast model is constructed. The water balance formula is as follows: ; Where, H i For the i The depth of water layer in the fruit field, H i-1 For the i -1 day water depth in the fruit field, D i For the i Daily drainage, S i For the i Leakage volume per day.

[0014] Preferably, in step 4: A "feedforward prediction + feedback decision-making" decision-making system is constructed, with the rainfall forecast model as the feedforward prediction layer. Based on the Markov property of the irrigation decision-making process, when considering future rainfall to make irrigation decisions, it is difficult to immediately know the final result and judge the impact of the current operation on the final result. Only the current water layer depth feedback can be obtained. The random forest algorithm is used to interact with the environment through a series of observations, actions, and rewards to find the optimal feedback. Lessons are learned from the wrong irrigation experiences of irrigation after rain, not irrigation, and waiting for rain but no rain, so as to find the optimal decision and construct an irrigation decision-making model. The following is the Gini impurity formula in the random forest algorithm: ; Where, p ( ui ) represents the probability that a random sample belongs to a category; Taking into account the mountainous and hilly terrain, the 5G wireless APN networking method is adopted, and the Internet of Things platform is used to issue irrigation decision instructions to the water intake valve assigned to the high-level water tank for irrigation.

[0015] Preferably, the step 5 specifically includes: The canopy temperature distribution map of fruit trees is obtained by using an unmanned aerial vehicle thermal infrared imager. After radiometric calibration, atmospheric correction, and emissivity correction, the plant water stress index is calculated. CWSI , where CWSI =0 indicates no stress, i.e., full irrigation, and CWSI = 1 indicates severe stress, i.e., complete drought. 0 < CWSI < 1 indicates different degrees of growth stress states, and then a stress spatial distribution map is generated. The water requirement model of fruit trees is optimized using feedback, and the plant water stress index CWSI The calculation formula is as follows: ; In the formula, T wet is the canopy-air temperature difference of fully irrigated plants, i.e., the lower limit; T dry is the canopy-air temperature difference of completely drought plants, i.e., the upper limit.

[0016] Preferably, the specific steps in step 6 include: Based on daily meteorological data, fruit tree growth data, and historical irrigation data, an irrigation expert knowledge base is established, and the irrigation strategy is iteratively updated.

[0017] Preferably, a large language model is used to construct an irrigation expert knowledge base, and daily meteorological data, fruit tree growth data, and historical irrigation data are used as training samples to iteratively optimize and update the strategy.

[0018] The present invention has the following beneficial effects: 1. This method focuses on fruit tree irrigation work, forms differentiated results in the field of analyzing fruit tree water stress, and provides certain reference value for fruit tree irrigation.

[0019] 2. This method proposes an implementable path for water-saving irrigation of fruit trees, which is easy to execute and can easily form standardized and process-based irrigation guidance. It can be effectively implemented during the work process, improving work efficiency.

[0020] 3. This method has a simple principle and a wide application range, and can be widely promoted in southern hilly irrigation areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be further described below with reference to the drawings and embodiments.

[0022] Figure 1 is a schematic diagram of the overall process of the present invention. <000018�> Figure 2 is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] In order to make the purpose, innovation, application field and operation method of the present invention clearer and more transparent, the present invention is further described in detail below with reference to the accompanying drawings.

[0025] See also Figure 1 The present invention provides a fruit tree irrigation method in southern irrigation areas based on integrated air-space-ground monitoring, which specifically includes: Step 1: Use air-space, ground-ground facilities and equipment to collect data on surface temperature, humidity, fruit tree transpiration intensity, soil moisture conditions, etc.

[0026] Satellite remote sensing data was used to obtain vegetation indices, surface temperature, and humidity in the irrigation area. Unmanned aerial vehicles (UAVs) equipped with high-performance multispectral and thermal infrared cameras were used to obtain canopy temperature and plant growth in the fruit tree planting areas in sync with satellite remote sensing cycles. Soil moisture stations were used to collect data on root moisture content, temperature, and humidity.

[0027] Furthermore, the satellite remote sensing data source uses Landsat-8 satellite data to perform vegetation index, surface temperature, and humidity inversion.

[0028] Furthermore, the Landsat-8 satellite data is subjected to data preprocessing such as radiation calibration, atmospheric correction, orthorectification, image fusion, image mosaicking, and image cropping.

[0029] Furthermore, the normalized difference vegetation index NDVI Calculate vegetation index, eliminate radiation error, and enhance the ability to respond to vegetation. NDVI The calculation formula is: ; Where: NIR is the reflectance value in the near-infrared band, R is the reflectance value of the infrared band; Get the surface emissivity, the calculation formula is: ; Get the blackbody radiation brightness, the calculation formula is: ; Where: X is the surface emissivity, b This is the radiation calibration result of Band 10; Furthermore, the surface temperature is obtained, and the calculation formula is: ; Furthermore, multispectral bands are used to construct WET Index, reflecting the moisture content of the surface humidity, is calculated as follows: ; Where,B (Blu-ray), G (Green Light), R (Red light) corresponds to the B2, B3, and B4 bands of Landsat8, and NIR (near infrared) corresponds to the B5 band of Landsat8. SWIR 1 and SWIR 2 is the B6 and B7 bands of Landsat 8.

[0030] Preferably, a DJI MAVIC 3 multispectral drone is used to obtain images of fruit trees in the same time sequence as Landsat 8 satellite images, which are interpreted through the DJI drone platform, and the canopy temperature of the fruit trees is inverted based on a thermal infrared imaging camera.

[0031] Furthermore, a time domain soil moisture monitoring station was used to collect data on the moisture content, temperature, and humidity of the fruit tree roots.

[0032] Step 2: Construct a simultaneous fusion dataset of multi-source data from the air, ground and sky to generate a high-precision soil moisture distribution map.

[0033] Furthermore, the Landsat8 satellite data with a spatial resolution of 15m and the UAV data with a spatial resolution of 0.05m were simultaneously downscaled and fused to generate a high-precision soil moisture distribution map.

[0034] Furthermore, based on comprehensive consideration of natural resource factors such as topography, hydrology, and vegetation, we used geographically weighted regression to spatially downscale the Landsat8 satellite data to a pixel base value of 1 meter. We then performed kriging interpolation on the drone data to generate 1-meter spatial resolution image data and performed data fusion to generate a high-temporal and high-precision soil moisture distribution map. The geographically weighted regression calculation formula is as follows: ; Where, For the location i The value of the dependent variable at ; For location i The value of the independent variable at k =1,2,..., m ;( u i , v i ) is the regression analysis point i coordinates; is the intercept term; is the regression analysis coefficient, k =1,2,..., m ; is a random error; The Kriging interpolation calculation formula is as follows: ; Where, is the value at some observation point x0, is the weight coefficient, is the attribute value of the sampling point.

[0035] Step 3: Obtain daily meteorological and fruit tree growth data, and divide the orchard into sub-areas based on soil texture and fruit tree plant distribution.

[0036] Furthermore, daily historical meteorological data of local stations for the past 20 years are collected from local meteorological departments or the China Meteorological Data Network, including the minimum temperature, maximum temperature, average temperature, average wind speed, sunshine hours, and average relative humidity. Weather forecast data with a forecast period of 7 days for the past 10 years are obtained through Caiyun Meteorological Service, including the maximum temperature, minimum temperature, weather type, wind force level, etc.

[0037] Furthermore, data on fruit tree water requirements from local representative experimental stations, soil parameters of the irrigation area (soil type, field water holding capacity, wilting coefficient, deep infiltration), and field irrigation water utilization coefficient were obtained.

[0038] Furthermore, irrigation zoning is carried out according to 45-50 fruit trees with the same or similar field water holding capacity as the most basic irrigation unit.

[0039] Step 4: Use the Hargreaves-Samani formula to build a fruit tree water demand model. Use weather type and rainfall intensity to construct an LSTM rainfall forecast model. Based on the water balance principle, build an irrigation quantity forecast model. Build a "feedforward prediction + feedback decision-making" system, using the rainfall forecast model as the feedforward prediction layer. Use a random forest algorithm to build an irrigation decision model, forming a feedback optimization layer. Irrigation decision instructions are issued to the water intake valves assigned to the high-level water tank via the IoT platform to implement irrigation.

[0040] Furthermore, considering that the Hargreaves-Samani model has high calculation accuracy in all climate zones, the Hargreaves-Samani formula method is used to construct the fruit tree water requirement model, and the root mean square error is used to evaluate the model prediction accuracy. The Hargreaves-Samani expression is: ; Where: Calculated by the Hargreaves-Samani formula ET 0 value; and The recommended values ​​of the two parameters are 0.0023 and 0.5 respectively; is extraterrestrial radiation, MJ·m-2 / sky; is the latent heat of evaporation at the average temperature, which is 2.45 MJ·kg -1 ; and are the maximum and minimum temperatures, respectively, in degrees Celsius; Furthermore, feature encoding is performed based on weather type and rainfall intensity, and the interactive features of weather type and rainfall intensity are coupled using LSTM to build a rainfall forecast model, and residual analysis is used to optimize the model performance; Furthermore, based on the water balance formula, an irrigation amount forecast model is constructed. The water balance formula is as follows: ; Where, H i For the i The depth of water layer in the fruit field, H i-1 For the i -1 day water depth in the fruit field, D i For the i Daily drainage, S i For the i Leakage volume per day.

[0041] Furthermore, a "feedforward prediction + feedback decision-making" decision-making system is constructed, with the rainfall forecast model as the feedforward prediction layer. Based on the Markov property of the irrigation decision-making process, when considering future rainfall to make irrigation decisions, it is difficult to immediately know the final result and judge the impact of the current operation on the final result. Only feedback on the current water layer depth can be obtained. The random forest algorithm is used to interact with the environment through a series of observations, actions, and rewards to find the optimal feedback. Lessons are learned from the wrong irrigation experiences of irrigation after rain, not irrigation, and waiting for rain but no rain, so as to find the optimal decision and construct an irrigation decision-making model. The following is the Gini impurity formula in the random forest algorithm: ; Where, p ( ui ) represents the probability that a random sample belongs to a category; Furthermore, considering the mountainous and hilly terrain, a 5G wireless APN networking method is adopted, and irrigation decision instructions are sent to the water intake valve assigned to the high-level water tank based on the Internet of Things platform for irrigation.

[0042] Step 5: Analyze the water stress of the fruit tree canopy based on UAV thermal infrared imaging technology and optimize the fruit tree water requirement model.

[0043] The canopy temperature distribution map of fruit trees is obtained by using an unmanned aerial vehicle thermal infrared imager. After radiometric calibration, atmospheric correction, and emissivity correction, the plant water stress index is calculated CWSI , where CWSI =0 indicates no stress, that is, full irrigation, CWSI = 1 indicates severe stress, that is, complete drought, and 0 < CWSI < 1 indicates different degrees of growth stress states. Furthermore, a stress spatial distribution map is generated, and the fruit tree water demand model is optimized using feedback. The plant water stress index CWSI The calculation formula is as follows: ; In the formula, T wet is the canopy-air temperature difference of fully irrigated plants, that is, the lower limit; T dry is the canopy-air temperature difference of completely drought plants, that is, the upper limit.

[0044] Step 6: Establish an irrigation expert knowledge base and iteratively update the irrigation strategy.

[0045] Based on daily meteorological data, fruit tree growth data, historical irrigation data, etc., an irrigation expert knowledge base is established, and the irrigation strategy is iteratively updated.

[0046] Furthermore, a large language model is used to construct an irrigation expert knowledge base, mainly using daily meteorological data, fruit tree growth data, historical irrigation data, etc. as training samples, and the update strategy is iteratively optimized.

[0047] So far, the present invention has described the technical method of the present invention in detail through the attached drawings and specific flowcharts. However, the protection scope of the present invention is not limited to the specific embodiments described. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and these technical solutions after the changes or substitutions still fall within the protection scope of the present invention.

Claims

1. A fruit tree irrigation method in southern irrigation areas based on integrated air-ground-space monitoring, characterized in that: include: Step 1, data collection: Use air-ground and ground-based equipment to collect data on surface temperature, humidity, transpiration intensity of fruit trees, and soil moisture in the irrigation area; Step 2: Generate soil moisture distribution map: Construct a simultaneous fusion dataset of multi-source data from space, ground and space to generate a high-precision soil moisture distribution map; Step 3, sub-area division: Obtain daily meteorological and fruit tree growth data, and divide the orchard into sub-areas based on soil texture and fruit tree plant distribution; Step 4: Construction of irrigation decision model and irrigation: The Hargreaves-Samani formula was used to construct a fruit tree water demand model. An LSTM rainfall forecast model was constructed using weather type and rainfall intensity. Based on the water balance principle, an irrigation volume forecast model was constructed. A "feedforward prediction + feedback decision-making" decision-making system was constructed, using the rainfall forecast model as the feedforward prediction layer. A random forest-based irrigation decision model was used to form a feedback optimization layer. Irrigation decision instructions were then issued to the water intake valves assigned to the high-level water tank via the Internet of Things platform. Step 5: Water demand model optimization: Analyze the water stress of fruit tree canopies using drone thermal infrared imaging technology and optimize the fruit tree water requirement model; Step 6: Establish an irrigation expert knowledge base and iteratively update the irrigation strategy.

2. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 1, characterized in that: The data collection process in step 1 specifically includes: Use satellite remote sensing data to obtain the inversion results of irrigation area vegetation index, surface temperature and humidity; Using drones equipped with high-performance multispectral cameras and thermal infrared imaging cameras, the canopy temperature and plant growth conditions in the fruit tree planting area are obtained in sync with the satellite remote sensing cycle; The soil moisture station was used to collect root moisture content, temperature and humidity data.

3. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 2, characterized in that: The satellite remote sensing data source uses Landsat-8 satellite data to perform vegetation index, surface temperature, and humidity inversion; Data preprocessing for Landsat-8 satellite data includes radiometric calibration, atmospheric correction, orthorectification, image fusion, image mosaicking, and image cropping; Normalized Difference Vegetation Index NDVI Calculate vegetation index, eliminate radiation error, and enhance the ability to respond to vegetation. NDVI The calculation formula is: ； Where: NIR is the reflectance value in the near-infrared band, R is the reflectance value of the infrared band; Get the surface emissivity, the calculation formula is: ； Get the blackbody radiation brightness, the calculation formula is: ； Where: X is the surface emissivity, b This is the radiation calibration result of Band 10; Get the surface temperature, which is calculated as: ； Constructed using multispectral bands WET Index, reflecting the moisture content of the surface humidity, is calculated as follows: ； In the formula, B, G, and R correspond to the B2, B3, and B4 bands of Landsat8 respectively. NIR It is the B5 band of Landsat8. SWIR 1 and SWIR 2 is the B6 and B7 bands of Landsat 8.

4. The fruit tree irrigation method in the southern irrigation area based on integrated air-space-ground monitoring according to claim 3 is characterized in that: Using a multispectral drone, we acquired fruit tree images in the same time series as Landsat 8 satellite images, interpreted them using the drone platform, and used a thermal infrared imaging camera to invert the canopy temperature of the fruit trees. The time domain method was used to collect root moisture content, temperature and humidity data of fruit tree plants using soil moisture monitoring stations.

5. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 1, characterized in that: The generation of the soil moisture distribution map in step 2 specifically includes: The Landsat8 satellite data with a certain spatial resolution is fused with the drone data with a certain spatial resolution to generate a high-precision soil moisture distribution map. Based on comprehensive consideration of topography, hydrology, and vegetation natural resource factors, geographically weighted regression was used to spatially downscale the Landsat8 satellite data with a pixel base value of 1 meter. Kriging interpolation was performed on the drone data to generate 1-meter spatial resolution image data and data fusion was performed to generate a high-temporal and spatial resolution and high-precision soil moisture distribution map. The geographically weighted regression calculation formula is as follows: ； Where, For the location i The value of the dependent variable at For location i The value of the independent variable at k =1,2,..., m ;( u i , v i ) is the regression analysis point i coordinates; is the intercept term; is the regression analysis coefficient, k =1,2,..., m ; is a random error; The Kriging interpolation calculation formula is as follows: ； Where, is the value at some observation point x0, is the weight coefficient, is the attribute value of the sampling point.

6. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 1, characterized in that: The sub-area division in step 3 specifically includes: Collect the past 20 years of daily historical meteorological data from local stations from the local meteorological department or the China Meteorological Data Network, including minimum temperature, maximum temperature, average temperature, average wind speed, sunshine hours, and average relative humidity. Also obtain weather forecast data with a 7-day forecast period for the past 10 years from Caiyun Meteorological Service, including maximum temperature, minimum temperature, weather type, and wind speed. Obtain data on fruit tree water requirements, irrigation area soil parameters, and field irrigation water utilization coefficients from local representative experimental stations; wherein the irrigation area soil parameters include soil type, field water holding capacity, wilting coefficient, and deep percolation; Irrigation zoning is carried out according to 45-50 fruit trees with the same or similar field water holding capacity as the most basic irrigation unit.

7. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 1, characterized in that: In step 4: Considering that the Hargreaves-Samani model has high calculation accuracy in all climate zones, the Hargreaves-Samani formula method is used to construct the fruit tree water requirement model, and the root mean square error is used to evaluate the model prediction accuracy. The Hargreaves-Samani expression is: ； Where: Calculated by the Hargreaves-Samani formula ET 0 value; and The recommended values ​​of the two parameters are 0.0023 and 0.5 respectively; is extraterrestrial radiation, MJ·m -2 / sky; is the latent heat of evaporation at the average temperature, which is 2.45 MJ·kg -1 ; and are the maximum and minimum temperatures, respectively, in degrees Celsius; Feature encoding is performed based on weather type and rainfall intensity. The interactive features of weather type and rainfall intensity are coupled using LSTM to build a rainfall forecast model, and residual analysis is used to optimize model performance. Based on the water balance formula, an irrigation amount forecast model is constructed. The water balance formula is as follows: ； Where, H i For the i The depth of water layer in the fruit field, H i-1 For the i -1 day water depth in the fruit field, D i For the i Daily drainage, S i For the i Leakage volume per day.

8. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 7, characterized in that: In step 4: A "feedforward prediction + feedback decision-making" decision-making system is constructed, with the rainfall forecast model as the feedforward prediction layer. Based on the Markov property of the irrigation decision-making process, when considering future rainfall to make irrigation decisions, it is difficult to immediately know the final result and judge the impact of the current operation on the final result. Only feedback on the current water layer depth can be obtained. The random forest algorithm is used to interact with the environment through a series of observations, actions, and rewards to find the optimal feedback. Lessons are learned from the wrong irrigation experiences of irrigation after rain, not irrigation, and waiting for rain but no rain, so as to find the optimal decision and construct an irrigation decision-making model. The following is the Gini impurity formula in the random forest algorithm: ； Where, p ( ui ) represents the probability that a random sample belongs to a category; Taking into account the mountainous and hilly terrain, the 5G wireless APN networking method is adopted, and the Internet of Things platform is used to issue irrigation decision instructions to the water intake valve assigned to the high-level water tank for irrigation.

9. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 1, characterized in that: The step 5 specifically includes: The canopy temperature distribution map of fruit trees is obtained by using an unmanned aerial vehicle thermal infrared imager. After radiometric calibration, atmospheric correction, and emissivity correction, the plant water stress index is calculated CWSI , where CWSI =0 indicates no stress, i.e., sufficient irrigation, CWSI = 1 indicates severe stress, i.e., complete drought, and 0 < CWSI < 1 indicates different degrees of growth stress states. Furthermore, a stress spatial distribution map is generated, and the water requirement model of fruit trees is optimized using feedback. The plant water stress index CWSI The calculation formula is as follows: ； Where, T wet The canopy-air temperature difference for adequate plant irrigation, i.e., the lower limit; T dry is the canopy-air temperature difference for completely arid plants, i.e., the upper limit.

10. The fruit tree irrigation method in southern irrigation areas based on integrated air-ground-space monitoring according to claim 1, characterized in that: The step 6 specifically includes: Based on daily meteorological data, fruit tree growth data, and historical irrigation data, an irrigation expert knowledge base is established to iteratively update irrigation strategies.

11. The fruit tree irrigation method in the southern irrigation area based on integrated air-ground-space monitoring according to claim 10, characterized in that: A large language model is used to build an irrigation expert knowledge base, using daily meteorological data, fruit tree growth data, and historical irrigation data as training samples to iteratively optimize and update strategies.

Images