Deep learning-based irrigation area scale crop nutrient diagnosis and early warning method and system
Through deep learning-based methods, satellite remote sensing data and meteorological data are used to quickly diagnose and warn crop nitrogen conditions, solving the problems of time-consuming, labor-intensive and lagging in traditional methods, and achieving accurate monitoring of the irrigation area scale.
Patent Information
- Application Number
- CN202411950951.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-27
- Publication Date
- 2025-02-18
AI Technical Summary
Traditional methods are time-consuming and labor-intensive and have lag in diagnosing crop nitrogen nutrient levels, making it difficult to apply to monitoring of the entire irrigation area scale.
Using a deep learning-based approach, by acquiring satellite remote sensing data and meteorological data, identifying crop types and phenological periods, calculating sunlight-induced chlorophyll fluorescence index and normalized vegetation index, and inputting a pre-trained deep learning model to output nitrogen nutrition index. Diagnosis and early warning of nitrogen conditions.
It realizes rapid and accurate diagnosis and early warning of crop nutrients at scale in irrigation areas, and solves the time-consuming, labor-intensive and lag problems of traditional methods.
Smart Images
Figure CN119478702A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of irrigation area crop nutrient diagnosis and early warning technology, and in particular to an irrigation area scale crop nutrient diagnosis and early warning method and system based on deep learning. Background Art
[0002] Nitrogen is a basic element for the synthesis of amino acids and proteins in plants, and is also a key component of chlorophyll, which is essential for crop growth and development. Insufficient nitrogen fertilizer supply will slow down crop photosynthesis and poor plant tissue development, which will lead to a simultaneous decrease in yield and quality. Excessive nitrogen fertilizer supply will lead to a decrease in crop nitrogen utilization efficiency, more nitrogen will be converted into NH3 and N2O, which will be emitted to pollute the atmosphere, or nitrogen leaching and lateral losses will cause regional non-point source pollution. As the center of grain production, irrigation areas have a wide planting area and a large demand for nitrogen fertilizer. Reasonable nitrogen application in irrigation areas is the core of ensuring food security and green agricultural development. However, traditional methods are time-consuming and labor-intensive, and there is a lag in diagnosing crop nitrogen nutrition levels, and this method is difficult to apply to monitoring at the scale of the entire irrigation area. Therefore, building a fast and accurate irrigation area crop nitrogen diagnosis and early warning system is of great significance to the sustainable development of agriculture. Summary of the invention
[0003] The present invention provides an irrigation district-scale crop nutrient diagnosis and early warning method and system based on deep learning, so as to solve the problem that traditional methods are time-consuming and labor-intensive and have a lag in diagnosing crop nitrogen nutrition levels, and are difficult to apply to monitoring at the scale of the entire irrigation district.
[0004] In a first aspect, the present invention provides an irrigation area scale crop nutrient diagnosis and early warning method based on deep learning, comprising: Obtain satellite remote sensing data and meteorological data of the irrigation area to be diagnosed; identifying crop types and phenological periods based on the satellite remote sensing data, and calculating the sunlight-induced chlorophyll fluorescence index and the normalized difference vegetation index; The crop type, phenological period, normalized difference vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index are input into a pre-trained deep learning model, and the nitrogen nutrition index is output through the deep learning model. The nitrogen status is diagnosed according to the nitrogen nutrition index and an early warning prompt is issued; wherein the training process of the deep learning model includes: obtaining the crop type, phenological period, normalized difference vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; constructing a data set for the first year of the field trial according to the crop type, phenological period, normalized difference vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; and using the data set for the first year of the field trial as input data for machine learning training to train the deep learning model.
[0005] Furthermore, during the training of the deep learning model, the nitrogen nutrition index is obtained by the following method: Design experiments on different nitrogen application rates for field crops; Destructive sampling was carried out at each key growth period of the crop to measure the dry matter mass and nitrogen content of the canopy leaves, and to obtain the dry matter mass of the aboveground leaves and the nitrogen concentration of the plants; According to the dry matter mass and nitrogen content of canopy leaves, the dry matter mass of aboveground leaves and the nitrogen concentration of plants, a critical nitrogen concentration dilution model for the critical growth period of crops was constructed; variance analysis was performed on the dry matter mass of aboveground leaves under different nitrogen application rates during the critical growth period, and the different nitrogen application rates during the critical growth period were divided into nitrogen-affected and nitrogen-unaffected tests; for the nitrogen application rate affected by nitrogen in crop growth, the dry matter mass of aboveground leaves was curve-fitted with the corresponding nitrogen concentration value; for the nitrogen application rate not affected by nitrogen in crop growth, the mean value of leaf dry matter mass was used to represent the maximum leaf dry matter mass; the critical nitrogen concentration of the aboveground part was calculated according to the critical nitrogen concentration dilution model:
[0006] In the formula, N c is the critical nitrogen concentration in the aboveground part; DM It is the maximum value of the dry matter of the aboveground leaves; a is the critical nitrogen concentration per unit leaf biomass of the aboveground part of the plant; b is the dilution factor that controls the slope of the curve;
[0007] In the formula, NNI is the nitrogen nutrition index; N a is the measured value of nitrogen concentration in the aboveground leaves of the plant; N c is the critical nitrogen concentration in the aboveground part.
[0008] Furthermore, during the training of the deep learning model, the canopy-scale sunlight-induced chlorophyll fluorescence index and meteorological data were obtained by the following method: Using the OCO-2 data from the carbon orbit observation satellite, combined with the non-missing OCO-2 SIF strip data and the corresponding MODIS surface reflectance data and MERRA-2 meteorological data, a machine learning algorithm was used to train and complete the missing data; finally, the canopy-scale sunlight-induced chlorophyll fluorescence index was obtained based on the OCO-2 SIF data and data-driven methods; Based on the existing meteorological station data, the spatial interpolation method is used to obtain the meteorological data of each point in the irrigation area.
[0009] Furthermore, during the training of the deep learning model, the phenological period of crops in the irrigation area is obtained by the following method: The irrigation area image data was obtained based on Sentinel-2 satellite images; The missing image data are filled by linear interpolation to achieve 8-day time series image data covering the entire time period; The Savitzjy-Golay filter was used to smooth the 8-day time series images, and finally a continuous 8-day time series of Sentinel-2 data was obtained; The normalized vegetation index of each pixel in the irrigation area was obtained using ENVI software, and the planting structure of the irrigation area was extracted based on the actual survey samples of the irrigation area and the random forest algorithm; The phenological period of crops is determined based on the normalized difference vegetation index time series curve.
[0010] Furthermore, the data set of the first year of the field trial is used as input data for machine learning training. After training the deep learning model, the method further includes: Obtain the dataset for the second year of the field experiment; The deep learning model is tested and verified based on the dataset from the second year.
[0011] Furthermore, the nitrogen nutrition index is used to diagnose the nitrogen status and issue early warning prompts, including: if NNI ∈[1-Δ1, 1+Δ2], diagnosed as the appropriate nitrogen range; if NNI <1-Δ1, diagnosed as plant nitrogen deficiency; if NNI >1+Δ2, diagnosed as excessive nitrogen nutrition of the plant; In the formula, NNI is the nitrogen nutrition index, Δ1 is the lower limit range threshold of nitrogen stress, Δ2 is the upper limit range threshold of nitrogen supersaturation, and the values of Δ1 and Δ2 vary with the type of crop.
[0012] In a second aspect, the present invention provides an irrigation area scale crop nutrient diagnosis and early warning system based on deep learning, comprising: The data acquisition module is used to obtain satellite remote sensing data and meteorological data of the irrigation area to be diagnosed.
[0013] A data calculation module, used to identify crop types and phenological periods according to the satellite remote sensing data, and calculate the sunlight-induced chlorophyll fluorescence index and the normalized vegetation index; A diagnosis and early warning module is used to input the crop type, phenological period, normalized vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index into a pre-trained deep learning model, output a nitrogen nutrition index through the deep learning model, diagnose the nitrogen status according to the nitrogen nutrition index and issue an early warning prompt; wherein the training process of the deep learning model includes: obtaining the crop type, phenological period, normalized vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; constructing a data set for the first year of the field trial based on the crop type, phenological period, normalized vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; using the data set for the first year of the field trial as input data for machine learning training to train the deep learning model.
[0014] The present invention has the following beneficial effects: a method and system for diagnosing and warning crop nutrients at the irrigation area scale based on deep learning of the present invention, the method obtains satellite remote sensing data and meteorological data of the irrigation area to be diagnosed; identifies crop types and phenological periods according to the satellite remote sensing data, and calculates the sunlight-induced chlorophyll fluorescence index and the normalized vegetation index; inputs the crop type, phenological period, normalized vegetation index, meteorological data and sunlight-induced chlorophyll fluorescence index into a pre-trained deep learning model, outputs the nitrogen nutrition index through the deep learning model, diagnoses the nitrogen status according to the nitrogen nutrition index and issues a warning prompt, which can realize rapid and accurate diagnosis and warning of crop nutrients at the irrigation area scale based on deep learning. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0016] Figure 1 A flow chart of the deep learning-based irrigation area-scale crop nutrient diagnosis and early warning method provided by the present invention. DETAILED DESCRIPTION
[0017] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.
[0018] See also Figure 1, a method for diagnosing and warning crop nutrients at the irrigation area scale based on deep learning provided by the present invention, comprising: Step 1: Obtain satellite remote sensing data and meteorological data of the irrigation area to be diagnosed.
[0019] Step 2: identifying crop types and phenological periods according to the satellite remote sensing data, and calculating the sunlight-induced chlorophyll fluorescence index and the normalized difference vegetation index; Step three, input the crop type, phenological period, normalized difference vegetation index, meteorological data and daylight induced chlorophyll fluorescence index into a pre-trained deep learning model, output the nitrogen nutrition index through the deep learning model, diagnose the nitrogen status according to the nitrogen nutrition index and issue an early warning prompt.
[0020] Among them, the training process of the deep learning model includes: obtaining the crop type, phenological period, normalized vegetation index, meteorological data and sunlight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; constructing a data set for the first year of the field trial based on the crop type, phenological period, normalized vegetation index, meteorological data and sunlight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; using the data set for the first year of the field trial as input data for machine learning training to train the deep learning model.
[0021] Specifically, during the training of the deep learning model, the nitrogen nutrition index is obtained by the following method: Field surveys were conducted on the main crop types in the irrigation area, and then traditional nitrogen nutrition index (NNI) acquisition tests were carried out for each crop. The traditional method of diagnosing the nitrogen status of crops requires first building a critical nitrogen concentration dilution model for the crops, and then building a nitrogen nutrition index. This requires destroying the samples to obtain the dry matter mass of the aboveground leaves of the crops, and then measuring the nitrogen concentration using a Kjeldahl nitrogen analyzer. The specific implementation steps are as follows: The experiment of different nitrogen application rates for field crops was designed, and the area of each field test plot was 10 m×10 m. Six nitrogen application rates were set, including no nitrogen application, ultra-low nitrogen application, lower nitrogen application, conventional nitrogen application, higher nitrogen application, and ultra-high nitrogen application. Destructive sampling was carried out at each key growth period of the crop to determine the dry matter mass and nitrogen content of the canopy leaves, and to obtain the dry matter mass of the aboveground leaves and the nitrogen concentration of the plant; for example, the key growth period of corn includes: seedling stage, jointing stage, trumpet stage, tassel stage, flowering stage, silking stage and grain filling stage. According to the dry matter mass and nitrogen content of canopy leaves, the dry matter mass of aboveground leaves and the nitrogen concentration of plants, a critical nitrogen concentration dilution model for the critical growth period of crops was constructed; variance analysis was performed on the dry matter mass of aboveground leaves under different nitrogen application rates during the critical growth period, and the different nitrogen application rates during the critical growth period were divided into nitrogen-affected and nitrogen-unaffected tests; for the nitrogen application rate affected by nitrogen in crop growth, the dry matter mass of aboveground leaves was curve-fitted with the corresponding nitrogen concentration value; for the nitrogen application rate not affected by nitrogen in crop growth, the mean value of leaf dry matter mass was used to represent the maximum leaf dry matter mass; the critical nitrogen concentration of the aboveground part was calculated according to the critical nitrogen concentration dilution model:
[0022] In the formula, N c is the critical nitrogen concentration in the aboveground part; DM It is the maximum value of the dry matter of the aboveground leaves; a is the critical nitrogen concentration per unit leaf biomass of the aboveground part of the plant; b is the dilution factor that controls the slope of the curve;
[0023] In the formula, NNI is the nitrogen nutrition index; N a is the measured value of nitrogen concentration in the aboveground leaves of the plant; N c is the critical nitrogen concentration in the aboveground part. NNI <1, indicating that the plant is nitrogen deficient. NNI = 1, indicating that the nitrogen nutrition in the plant is in the best state, NNI >1 indicates that the plant has excess nitrogen nutrition. In actual situations, NNI cannot be exactly equal to 1. According to actual conditions, the range of NNI∈[1-Δ1, 1+Δ2] can be set as the acceptable range of nitrogen. NNI <1-Δ1 indicates that the plant is nitrogen deficient. NNI >1+Δ2 indicates that the plant has excess nitrogen nutrition.
[0024] Specifically, during the training of the deep learning model, the canopy-scale solar-induced chlorophyll fluorescence index and meteorological data were obtained by the following method: Orbiting Carban Observatory-2 (OCO-2) data were used, combined with the non-missing OCO-2 solar-induced chlorophyll fluorescence (SIF) band data and the corresponding Moderate-Resolution Imaging Spectroradiometer (MODIS) surface reflectance data and Modern-Era Retrospective Analysis for Research and Applications-Version 2 (MERRA-2) meteorological data, and the missing data were completed by training a machine learning algorithm; finally, the canopy-scale solar-induced chlorophyll fluorescence index was obtained based on the OCO-2 SIF data and data-driven methods; according to the existing meteorological station data, the meteorological data of each point in the irrigation area were obtained by spatial interpolation.
[0025] Sunlight-induced chlorophyll fluorescence is a long-wave signal emitted in the form of fluorescence by the photosynthetic center of chlorophyll molecules under sunlight, with two peaks: red light and near infrared. The wavelength of the long-wave signal is 650nm-800nm, the wavelength of the red light is about 690nm, and the wavelength of the near infrared is about 740nm. SIF is considered to be an ideal probe for plant photosynthesis, which can directly reflect the photosynthetic intensity of plant leaves. 30%-40% of the nitrogen in the leaves participates in the carboxylation reaction, and the proportion of the total nitrogen content of the leaves allocated to the photosynthetic carboxylation system determines the final photosynthetic efficiency. In addition, the leaf nitrogen content (LNC) will affect the photosynthetic capacity of the plant, thereby affecting the non-photochemical quenching coefficient, that is, changing the emission of fluorescence. The amount of nitrogen applied during plant growth will affect the plant LNC, and the photosynthetic capacity, electron emission intensity and canopy structure parameters of the plant are extremely sensitive to LNC. Therefore, diagnosing the nitrogen status of crops through SIF has a theoretical basis and timeliness. Then, an all-weather vegetation fluorescence observation system is installed above the canopy. The system has two high-resolution, high-sensitivity, and high signal-to-noise ratio sensors. The fluorescence band spectrometer observes the downward and upward irradiance of the vegetation canopy and calculates the sunlight-induced chlorophyll fluorescence index of the vegetation in real time. The observation range of the fluorescence band spectrometer is 640 nm -800 nm. The full-band spectrometer can further calculate the normalized vegetation index by real-time monitoring of the full-band reflectance. The full-band spectrometer has an observation range of 200 nm -1100 nm. Record the specific number of days after sowing and the corresponding meteorological data, such as temperature, relative humidity, solar radiation, and wind speed.
[0026] SIF satellite remote sensing data can be obtained from the carbon orbit observation satellite OCO-2, which is synchronized with the solar orbit and has a revisit period of 16 days. The main observation bands of the satellite detector include: O2-A band, 757nm–775 nm). Weak CO2 absorption band, 1594nm–1627 nm. Strong CO2 absorption band, 2043nm–2087 nm. Although the SIF product inverted by OCO-2 has a high spatial resolution of 1.3 km×2.25 km, its original SIF data is discontinuous in space. It is necessary to combine the non-missing OCO-2 SIF strip data with the corresponding MODIS surface reflectance data and MERRA-2 meteorological data to train the machine learning algorithm to complete the missing data. Finally, based on the OCO-2 SIF data and data-driven methods, the global SIF dataset GOSIF with a higher spatial resolution of 0.05˚ and a temporal resolution of 8 days was developed. According to the existing meteorological station data, the meteorological data of each point in the irrigation area was obtained by spatial interpolation.
[0027] Specifically, during the training of the deep learning model, the phenological period of crops in the irrigation area is obtained through the following methods: the image data of the irrigation area is obtained based on the Sentinel-2 satellite images; the missing image data is filled by linear interpolation to achieve 8-day time series image data covering the entire time period; the 8-day time series images are smoothed by the Savitzky-Golay (SG) filter, and finally a continuous 8-day time series of Sentinel-2 data is obtained; the normalized vegetation index of each pixel in the irrigation area is obtained based on the complete remote sensing image processing platform (The Environment for Visualizing Images, ENVI) software, and the planting structure of the irrigation area is extracted based on the actual survey samples of the irrigation area and the random forest algorithm; the phenological period of crops is determined based on the normalized vegetation index time series curve.
[0028] Sentinel-2 satellite images come from the European Space Agency, with a revisit cycle of five days. Sentinel-2's 1C-level products are orthorectified, georeferenced, and radiometrically calibrated to form reflectance data at the top of the atmosphere. The atmospheric correction model Sen2Cor (Sentinel-2 Level-2A Atmospheric CorrectionProcessor) in the Sentinel application platform software is then used to process the reflectance of the 1C-level product to the surface. In order to eliminate the influence of the observation dates of different orbits of Sentinel-2 and cloud cover on the discontinuous time interval of Sentinel-2 data, the image gap data can be filled by linear interpolation to achieve 8-day time series image data covering the entire time period. The SG filter is used to smooth the 8-day time series image, and finally the continuous 8-day time series Sentinel-2 data is obtained. Then, the normalized vegetation index of each pixel in the irrigation area is obtained according to the ENVI software, and the planting structure of the irrigation area is extracted according to the actual survey samples of the irrigation area and the random forest algorithm. Then, the crop phenological period can be determined according to the normalized vegetation index time series curve.
[0029] Specifically, the data set of the first year of the field trial is used as input data for machine learning training. After training the deep learning model, the method also includes: obtaining the data set of the second year of the field trial; and testing and verifying the deep learning model based on the data set of the second year.
[0030] Specifically, the nitrogen nutrition index is used to diagnose the nitrogen status and issue early warning prompts, including: NNI ∈[1-Δ1, 1+Δ2], diagnosed as the appropriate nitrogen range; if NNI <1-Δ1, the plant is diagnosed as nitrogen deficiency; if NNI >1+Δ2, diagnosed as excessive nitrogen nutrition of the plant; In the formula,NNI is the nitrogen nutrition index, Δ1 is the lower limit range threshold of nitrogen stress, Δ2 is the upper limit range threshold of nitrogen supersaturation, and the values of Δ1 and Δ2 vary with the type of crop.
[0031] The embodiment of the present invention provides an irrigation area scale crop nutrient diagnosis and early warning system based on deep learning, comprising: The data acquisition module is used to obtain satellite remote sensing data and meteorological data of the irrigation area to be diagnosed.
[0032] A data calculation module, used to identify crop types and phenological periods according to the satellite remote sensing data, and calculate the sunlight-induced chlorophyll fluorescence index and the normalized vegetation index; The diagnosis and early warning module is used to input the crop type, phenological period, normalized vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index into a pre-trained deep learning model, output the nitrogen nutrition index through the deep learning model, diagnose the nitrogen status according to the nitrogen nutrition index and issue an early warning prompt.
[0033] An embodiment of the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the irrigation area scale crop nutrient diagnosis and early warning method based on deep learning of the present invention.
[0034] The embodiment of the present invention further provides a storage medium, in which a computer program is stored, and when the computer program is executed by a processor, some or all of the steps in each embodiment of the irrigation area scale crop nutrient diagnosis and early warning method based on deep learning provided by the present invention are implemented. The storage medium can be a disk, an optical disk, a read-only memory (English: Read-Only Memory, abbreviated: ROM) or a random access memory (English: Random Access Memory, abbreviated: RAM), etc.
[0035] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention or certain parts of the embodiments.
[0036] The same or similar parts between the various embodiments in this specification can be referred to each other. In particular, for the embodiment of the irrigation area scale crop nutrient diagnosis and early warning system based on deep learning, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.
[0037] The above-described embodiments of the present invention do not limit the protection scope of the present invention.
Claims
1. A deep learning-based irrigation area-scale crop nutrient diagnosis and early warning method, characterized in that: include: Obtain satellite remote sensing data and meteorological data of the irrigation area to be diagnosed; identifying crop types and phenological periods based on the satellite remote sensing data, and calculating the sunlight-induced chlorophyll fluorescence index and the normalized difference vegetation index; The crop type, phenological period, normalized difference vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index are input into a pre-trained deep learning model, and the nitrogen nutrition index is output through the deep learning model. The nitrogen status is diagnosed according to the nitrogen nutrition index and an early warning prompt is issued; wherein the training process of the deep learning model includes: obtaining the crop type, phenological period, normalized difference vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; constructing a data set for the first year of the field trial according to the crop type, phenological period, normalized difference vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; and using the data set for the first year of the field trial as input data for machine learning training to train the deep learning model.
2. The method for diagnosing and warning crop nutrients at the irrigation area scale based on deep learning according to claim 1, characterized in that: During the training of the deep learning model, the nitrogen nutrition index is obtained by the following method: Design experiments on different nitrogen application rates for field crops; Destructive sampling was carried out at each key growth period of the crop to measure the dry matter mass and nitrogen content of the canopy leaves, and to obtain the dry matter mass of the aboveground leaves and the nitrogen concentration of the plants; According to the dry matter mass and nitrogen content of canopy leaves, the dry matter mass of aboveground leaves and the nitrogen concentration of plants, a critical nitrogen concentration dilution model for the critical growth period of crops was constructed; variance analysis was performed on the dry matter mass of aboveground leaves under different nitrogen application rates during the critical growth period, and the different nitrogen application rates during the critical growth period were divided into nitrogen-affected and nitrogen-unaffected tests; for the nitrogen application rate affected by nitrogen in crop growth, the dry matter mass of aboveground leaves was curve-fitted with the corresponding nitrogen concentration value; for the nitrogen application rate not affected by nitrogen in crop growth, the mean value of leaf dry matter mass was used to represent the maximum leaf dry matter mass; the critical nitrogen concentration of the aboveground part was calculated according to the critical nitrogen concentration dilution model: In the formula, N c is the critical nitrogen concentration in the aboveground part; DM It is the maximum value of the dry matter of the aboveground leaves; a is the critical nitrogen concentration per unit leaf biomass of the aboveground part of the plant; b is the dilution factor that controls the slope of the curve; In the formula, NNI is the nitrogen nutrition index; N a is the measured value of nitrogen concentration in the aboveground leaves of the plant; N c is the critical nitrogen concentration in the aboveground part.
3. The method for diagnosing and warning crop nutrients at the irrigation area scale based on deep learning according to claim 1, characterized in that: During the training of the deep learning model, the canopy-scale sunlight-induced chlorophyll fluorescence index and meteorological data were obtained by the following methods: Using the OCO-2 data from the carbon orbit observation satellite, combined with the non-missing OCO-2 SIF strip data and the corresponding MODIS surface reflectance data and MERRA-2 meteorological data, a machine learning algorithm was used to train and complete the missing data; finally, the canopy-scale sunlight-induced chlorophyll fluorescence index was obtained based on the OCO-2 SIF data and data-driven methods; Based on the existing meteorological station data, the spatial interpolation method is used to obtain the meteorological data of each point in the irrigation area.
4. The method for diagnosing and warning crop nutrients at the irrigation area scale based on deep learning according to claim 1, characterized in that: During the training of the deep learning model, the phenological period of crops in the irrigation area is obtained through the following methods: The irrigation area image data was obtained based on Sentinel-2 satellite images; The missing image data were filled by linear interpolation to obtain the image data of 8 days covering the entire time period; The Savitzjy-Golay filter was used to smooth the 8-day time series images, and finally a continuous 8-day time series of Sentinel-2 data was obtained; The normalized vegetation index of each pixel in the irrigation area was obtained using ENVI software, and the planting structure of the irrigation area was extracted based on the actual survey samples of the irrigation area and the random forest algorithm; The phenological period of crops is determined based on the normalized difference vegetation index time series curve.
5. The method for diagnosing and warning crop nutrients at the irrigation area scale based on deep learning according to claim 1, characterized in that: The dataset of the first year of the field trial is used as the input data for machine learning training. After training the deep learning model, the following steps are also included: Obtain the dataset for the second year of the field experiment; The deep learning model is tested and verified based on the dataset from the second year.
6. The method for diagnosing and warning crop nutrients at the irrigation area scale based on deep learning according to claim 1, characterized in that: Diagnose nitrogen status and issue early warnings based on nitrogen nutrition index, including: if NNI ∈[1-Δ1, 1+Δ2], diagnosed as the appropriate nitrogen range; if NNI < 1-Δ1, diagnosed as plant nitrogen deficiency; if NNI > 1+Δ2, the plant was diagnosed as having excessive nitrogen nutrition; In the formula, NNI is the nitrogen nutrition index, Δ1 is the lower limit range threshold of nitrogen stress, Δ2 is the upper limit range threshold of nitrogen supersaturation, and the values of Δ1 and Δ2 vary with the type of crop.
7. A deep learning-based irrigation area-scale crop nutrient diagnosis and early warning system, characterized in that: include: The data acquisition module is used to obtain satellite remote sensing data and meteorological data of the irrigation area to be diagnosed. A data calculation module, used to identify crop types and phenological periods according to the satellite remote sensing data, and calculate the sunlight-induced chlorophyll fluorescence index and the normalized vegetation index; A diagnosis and early warning module is used to input the crop type, phenological period, normalized vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index into a pre-trained deep learning model, output a nitrogen nutrition index through the deep learning model, diagnose the nitrogen status according to the nitrogen nutrition index and issue an early warning prompt; wherein the training process of the deep learning model includes: obtaining the crop type, phenological period, normalized vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; constructing a data set for the first year of the field trial based on the crop type, phenological period, normalized vegetation index, meteorological data and daylight-induced chlorophyll fluorescence index of the crops in the irrigation area in the first year of the field trial; using the data set for the first year of the field trial as input data for machine learning training to train the deep learning model.