Oil sunflower field precise fertilization control system based on unmanned aerial vehicle hyperspectral technology

Abstract

The invention discloses an oil sunflower field precise fertilization control system based on an unmanned aerial vehicle hyperspectral technology, and relates to the technical field of fertilization control. The system carries out precise region division on an oil sunflower field by utilizing a GIS technology through fusion of multi-source data such as hyperspectral data, soil humidity and temperature; and accurate fertilization can be carried out in different management areas according to soil characteristics and crop requirements. The differentiated fertilization strategy can reduce excessive fertilization and fertilizer waste and optimize the utilization efficiency of the fertilizer. The crop growth model is established through the deep learning technology, and the system can accurately calculate the required fertilization amount according to the specific growth conditions and historical data of each region, so that the growth conditions of the oil sunflowers are optimized, and the overall yield of crops is improved. The problem that a traditional fertilization system is difficult to effectively deal with soil and environment differences in different areas is solved. Through a hyperspectral technology and multi-source data acquired by a sensor, the system can realize accurate fertilization according to environmental differences in a field.

Description

Technical Field

[0001] The present invention relates to the technical field of fertilization control, and in particular to a precise fertilization control system for oil sunflower fields based on unmanned aerial vehicle (UAV) hyperspectral technology. Background Art

[0002] Precision agriculture is a technical means to achieve refined management of the crop production process by using modern information technology, especially sensors, satellites and data analysis tools. Specifically speaking, precision agriculture involves collecting and analyzing soil and crop growth data from different regions to develop personalized farmland management strategies.

[0003] In the traditional field management and fertilization process, farmers usually use a unified fertilization strategy, that is, fertilizer is applied according to the average conditions of the entire farmland, ignoring the differences in soil conditions and crop requirements between different areas of the field. This method not only fails to meet the specific nutritional needs of oil sunflower crops at different growth stages, but also leads to fertilizer waste and environmental pollution.

[0004] Existing field management systems make it difficult to accurately monitor and personalize fertilization in different areas, resulting in unbalanced crop nutrition, which in turn affects overall yield. A significant disadvantage of this traditional fertilization model is that it cannot be dynamically adjusted based on the combined influence of multiple factors such as soil conditions, humidity, and temperature. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a sunflower field precision fertilization control system based on unmanned aerial vehicle hyperspectral technology, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a sunflower field precision fertilization control system based on drone hyperspectral technology, including a data acquisition module, a data preprocessing module, a region division module, a regional feature extraction module, a fertilization strategy formulation module and a feedback and optimization module;

[0007] The data acquisition module collects the original soil moisture data S(x, y) and the original temperature data T(x, y) through sensors, and obtains the historical yield data Y from the historical yield. His (x, y), and use a drone equipped with a hyperspectral imaging device to perform a flight scan of the sunflower field to obtain the original hyperspectral reflectance R(λ, x, y) of different bands;

[0008] The data preprocessing module performs denoising, normalization and illumination correction on the collected hyperspectral reflectance R(λ,x,y), obtains the processed corrected hyperspectral reflectance R′(λ,x,y), and standardizes the original soil moisture data S(x,y) and the original temperature data T(x,y), and obtains standardized soil moisture data S′(x,y) and standardized temperature data T′(x,y);

[0009] The region division module uses the corrected hyperspectral reflectance, standardized soil moisture data S′(x,y) and standardized temperature data T′(x,y) to obtain a high-dimensional feature vector F(x,y) and divide the field into multiple management areas through GIS;

[0010] The regional feature extraction module obtains a set Z of management areas according to the high-dimensional feature vector F(x, y) for the multiple management areas divided by the regional division module, and extracts the regional feature vector F i Including correction for the average value of hyperspectral reflectance Average soil moisture and the average temperature

[0011] The fertilization strategy formulation module is based on the regional feature vector F i , by using deep learning technology to establish a crop growth model, and use the crop growth model to calculate the amount of fertilizer E for each area R , and generate the fertilization strategy E i ;

[0012] After fertilization is completed, the feedback and optimization module collects feedback data by monitoring the growth of oil sunflowers and generates a new fertilization strategy E i，next .

[0013] Preferably, the data acquisition module includes a sensor data acquisition unit and a UAV hyperspectral imaging unit;

[0014] The sensor data acquisition unit collects the original soil moisture data S(x, y) and original temperature data T(x, y) at different positions of the field through the embedded soil moisture sensor and the embedded temperature sensor, and obtains the historical yield data Y of each plot from the historical database His (x,y);

[0015] The UAV hyperspectral imaging unit scans the entire sunflower field by carrying a hyperspectral imaging device on the UAV to obtain the original hyperspectral reflectance R(λ, x, y) of different bands;

[0016] Where R(λ,x,y) represents the original hyperspectral reflectance with wavelength λ at position (x,y).

[0017] Preferably, the data preprocessing module includes a hyperspectral data processing unit and a soil data standardization unit;

[0018] The hyperspectral data processing unit uses a filter to perform denoising on the original hyperspectral reflectance R(λ, x, y) to remove random noise and obtain the denoised hyperspectral reflectance R denoise (λ,x,y);

[0019] Then the denoised hyperspectral reflectance R denoise (λ, x, y) is normalized to obtain the normalized hyperspectral reflectance R norm (λ,x,y), for the normalized hyperspectral reflectance R norm (λ, x, y) performs illumination condition correction to obtain the corrected hyperspectral reflectance R′(λ, x, y);

[0020] The corrected hyperspectral reflectance R′(λ,x,y) is obtained by the following formula:

[0021]

[0022] Where L(λ,x,y) represents the light intensity correction factor;

[0023] The soil data standardization unit performs standardization processing on the original soil moisture data S(x, y) and the original temperature data T(x, y) to obtain standardized soil moisture data S′(x, y) and standardized temperature data T′(x, y);

[0024] The standardized soil moisture data S′(x, y) is obtained by the following formula:

[0025]

[0026] In the formula, μS represents the mean of the original soil moisture data, and σS represents the standard deviation of the original soil moisture data;

[0027] The standardized temperature data T′(x, y) is obtained by the following formula:

[0028]

[0029] Wherein, μT represents the mean of the original temperature data, and σT represents the standard deviation of the original temperature data.

[0030] Preferably, the region division module includes a high-dimensional feature extraction unit and a GIS region division unit;

[0031] The high-dimensional feature extraction unit extracts features from each position (x, y) according to the corrected hyperspectral reflectance R′(λ, x, y), the standardized soil moisture data S′(x, y) and the standardized temperature data T′(x, y) to obtain a high-dimensional feature vector F(x, y);

[0032] The high-dimensional feature vector F(x, y) is obtained by the following formula:

[0033] F(x,y)=[R′(λ1,x,y),R′(λ2,x,y),,,R′(λ n ,x,y),...,S′(x,y),T′(x,y)];

[0034] The high-dimensional feature vector F(x,y) includes corrected hyperspectral reflectance at multiple wavelengths;

[0035] The GIS area division unit divides the field into multiple management areas through GIS, applies a clustering algorithm based on the high-dimensional feature vector F(x,y), and divides locations (x,y) with similar features into the same management area; obtains a set Z of management areas:

[0036] The set Z of management areas is obtained by the following formula:

[0037] Z={(x,y)F(x,y)∈Cluster i};

[0038] In the formula, Cluster i represents the i-th management area obtained by the clustering algorithm;

[0039] Map the set of management areas Z to GIS, generate a regional division layer, and intuitively display each management area Z i location and range.

[0040] Preferably, the regional feature extraction module includes a regional data aggregation unit and a feature calculation unit;

[0041] The regional data aggregation unit aggregates data for all positions (x, y) of the divided area, and aggregates the corrected hyperspectral reflectance R′(λ, x, y), the standardized soil moisture data S′(x, y), and the standardized temperature data T′(x, y);

[0042] The classified data of the location points (x, y)∈Z within the set Z of the management area are summarized, including the set of corrected hyperspectral reflectance R′(λ, x, y) in the area, the set of standardized soil moisture data S′(x, y) and the set of standardized temperature data T′(x, y):

[0043] Among them, the set of corrected hyperspectral reflectance R′(λ,x,y) in the region is:

[0044] {R′(λ,x,y)|(x,y)∈Z};

[0045] The set of standardized soil moisture data S′(x,y) in the region is:

[0046] {S′(x,y)|(x,y)∈Z};

[0047] The set of standardized temperature data T′(x,y) in the region is:

[0048] {T′(x,y)(x,y)∈Z}.

[0049] Preferably, the feature calculation unit calculates the average value of the corrected hyperspectral reflectance in the management area set Z

[0050] The average value of the corrected hyperspectral reflectance Obtained by the following formula:

[0051]

[0052] Where M represents the number of location points in the set Z of management areas;

[0053] Calculate the average soil moisture in a set of management areas Z and the average temperature

[0054] The average soil moisture Obtained by the following formula:

[0055]

[0056] The average value of the temperature Obtained by the following formula:

[0057]

[0058] The average value of the hyperspectral reflectance will be corrected Average soil moisture and the average temperature Combine to obtain the regional feature vector F i :

[0059] The regional feature vector F i Obtained by the following formula:

[0060]

[0061] Preferably, the fertilization strategy formulation module includes a crop growth model unit and a fertilization amount calculation unit;

[0062] The crop growth model unit establishes a crop growth model by using deep learning technology, and according to the regional feature vector F i The crop growth model is used to predict the crop growth status and demand in each region and output the crop growth potential G P and nutritional requirements req ;

[0063] The growth potential of the crop G P Obtained by the following formula:

[0064]

[0065] In the formula, α1 represents light parameters, α2 represents soil parameters, α3 represents temperature parameters, α4 represents management measures parameters, and N represents nitrogen content in the soil;

[0066] According to the growth potential of crops P Calculate the nutrient requirements of crops at different growth stages req ;

[0067] The nutritional requirement N req Obtained by the following formula:

[0068]

[0069] Where, b1, b2 and b3 represent the growth potential of crops, G P , the average soil moisture and the average temperature The weight coefficient of .

[0070] Preferably, the fertilizer amount calculation unit calculates the nutrient requirement N provided by the growth model unit. req , calculate each management area Z i The amount of fertilizer E R , and generate the fertilization strategy E i ;

[0071] For a set of management areas Z, according to the nutrient demand N req Calculate the required amount of fertilizer E based on the nitrogen content N in the soil R ;

[0072] The amount of fertilizer E R Obtained by the following formula:

[0073]

[0074] Where N req (Fi ) represents the regional feature vector F i Nutritional requirements req , represents the average soil nitrogen content in the set Z of management areas, ω1 and ω2 represent the regional characteristic vectors F i Nutritional requirements req and the weighted value of the mean soil nitrogen content within the set Z of management areas;

[0075] According to the obtained fertilizer amount E R , calculate the fertilization strategy E i ;

[0076] The fertilization strategy E i Obtained by the following formula:

[0077]

[0078] In the formula, and represents the empirical parameters derived from field experiments and historical data, which are used to adjust the amount of fertilizer to adapt to different crop growth potentials and time points. θ1 and θ2 represent the empirical weight coefficients, reflecting the impact of growth stage and climate conditions on fertilization time. GS represents the current crop growth stage, WC represents the sequence value of the current climate condition type, Md(x) represents the selection coefficient of the fertilization method, AF represents the adjustment factor, and E f Indicates historical fertilization efficiency;

[0079] The selection coefficient Md(x) of the fertilization method is obtained by matching the following formula:

[0080]

[0081] System Utilization Fertilization Strategy E i , control the amount and method of fertilizer use, and regulate the growth of crops.

[0082] Preferably, the feedback and optimization module includes a growth monitoring unit and a strategy optimization unit;

[0083] The growth monitoring unit collects the growth data of the oil sunflower, including the growth height H and the chlorophyll content C, and compares it with the expected growth target to calculate the growth deviation ΔG(t):

[0084] The growth deviation ΔG(t) is obtained by the following formula:

[0085] ΔG(t)=Gx(t)-Ga(t);

[0086] In the formula, Gx(t) represents the expected growth target predicted by the growth model at time t, and Ga(t) represents the actual growth situation monitored at time t;

[0087] When ΔG(t)≤0, it means that there is no need to adjust the fertilization strategy E i ;

[0088] When ΔG(t)>0, it means that the fertilization strategy E needs to be adjusted. i ; According to the growth deviation ΔG(t) and the fertilizer amount E R , get the new fertilizer amount E next ;

[0089] The new fertilizer amount E next Obtained by the following formula:

[0090] E next =E R +k*ΔG(t)*E f ;

[0091] Wherein, k represents the feedback gain parameter.

[0092] A precise fertilization control system for oil sunflower fields based on UAV hyperspectral technology, wherein the strategy optimization unit uses the new fertilizer amount E next and the actual growth situation Ga(t) at time t, and obtain the adjusted fertilization strategy E i，next , used to regulate crop growth.

[0093] The adjusted fertilization strategy E i，next Obtained by the following formula:

[0094]

[0095] The present invention provides a sunflower field precision fertilization control system based on UAV hyperspectral technology, which has the following beneficial effects:

[0096] (1) The system uses GIS technology to accurately divide the sunflower field by fusing hyperspectral data, soil moisture, temperature and other multi-source data, ensuring that different management areas are accurately fertilized according to their soil characteristics and crop needs. This differentiated fertilization strategy can reduce over-fertilization and fertilizer waste and optimize fertilizer utilization efficiency. By establishing a crop growth model through deep learning technology, the system can accurately calculate the required amount of fertilizer based on the specific growth conditions and historical data of each area, thereby optimizing the growth conditions of sunflowers and increasing the overall yield of crops. This data-driven strategy can tailor fertilization plans for each area to avoid resource mismatch. The feedback and optimization module of the system can monitor the growth of sunflowers after fertilization is completed, collect feedback data, and dynamically adjust subsequent fertilization strategies based on this data. This can respond to environmental changes and fluctuations in crop demand in a timely manner, avoid over-fertilization or nutrient deficiency, and further improve the accuracy of fertilization.

[0097] (2) Through the sensor data acquisition unit, the embedded soil moisture and temperature sensors can collect the original soil moisture data S(x, y) and original temperature data T(x, y) at different locations in the sunflower field, which can fully consider the differences in soil in different regions. At the same time, the historical yield data Y in the historical database is used His (x, y) provides a basis for decision-making and ensures that the fertilization strategy is more targeted. The UAV hyperspectral imaging unit can obtain hyperspectral reflectance in different bands by scanning the field. Through filter denoising, normalization processing and light condition correction, the system can eliminate the influence of environmental noise and light changes, and obtain accurate corrected hyperspectral reflectance R′(λ, x, y), providing high-quality data support for subsequent crop growth models.

[0098] (3) Through the combination of high-dimensional feature extraction unit and GIS regional division unit, this system can use clustering algorithm to divide locations with similar features into the same management area based on multi-source data including corrected hyperspectral reflectance R′(λ,x,y), standardized soil moisture data S′(x,y) and standardized temperature data T′(x,y), and generate intuitive regional division layers. This method avoids the limitations of traditional single-factor regional division, ensures that the internal characteristics of each management area are more consistent, and thus provides a basis for refined fertilization strategies.

[0099] (4) The crop growth model unit uses deep learning technology to establish a crop growth model. The system can calculate the crop growth model based on the feature vectors F of different regions. i Including correction for the average value of hyperspectral reflectance Average soil moisture and the average temperature Accurately predict the growth status and nutritional needs of crops in each area. This prediction method greatly improves the insight into the growth potential of crops, combines multi-source data for refined analysis, and ensures that crops in different areas receive the most appropriate nutritional support. The fertilizer calculation unit dynamically adjusts the fertilizer amount in each management area based on the nutritional needs of the crops and the nitrogen content in the soil, combined with the weight coefficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] Figure 1 The present invention is a schematic diagram of the flowchart of a sunflower field precision fertilization control system based on unmanned aerial vehicle hyperspectral technology. DETAILED DESCRIPTION

[0101] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0102] Example 1

[0103] The present invention provides a sunflower field precision fertilization control system based on UAV hyperspectral technology. Figure 1 , including data acquisition module, data preprocessing module, regional division module, regional feature extraction module, fertilization strategy formulation module and feedback and optimization module;

[0104] The data acquisition module collects the original soil moisture data S(x, y) and the original temperature data T(x, y) through sensors, and obtains the historical yield data Y from the historical yield. His (x, y), and use a drone equipped with a hyperspectral imaging device to perform a flight scan of the sunflower field to obtain the original hyperspectral reflectance R(λ, x, y) of different bands;

[0105] The data preprocessing module performs denoising, normalization and illumination correction on the collected hyperspectral reflectance R(λ,x,y), obtains the processed corrected hyperspectral reflectance R′(λ,x,y), and standardizes the original soil moisture data S(x,y) and the original temperature data T(x,y), and obtains standardized soil moisture data S′(x,y) and standardized temperature data T′(x,y);

[0106] The region division module uses the corrected hyperspectral reflectance, standardized soil moisture data S′(x,y) and standardized temperature data T′(x,y) to obtain a high-dimensional feature vector F(x,y) and divide the field into multiple management areas through GIS;

[0107] The regional feature extraction module obtains a set Z of management areas according to the high-dimensional feature vector F(x, y) for the multiple management areas divided by the regional division module, and extracts the regional feature vector F i Including correction for the average value of hyperspectral reflectance Average soil moisture and the average temperature

[0108] The fertilization strategy formulation module is based on the regional feature vector F i , by using deep learning technology to establish a crop growth model, and use the crop growth model to calculate the amount of fertilizer E for each area R , and generate the fertilization strategy E i ;

[0109] After fertilization is completed, the feedback and optimization module collects feedback data by monitoring the growth of oil sunflowers and generates a new fertilization strategy E i，next .

[0110] In this embodiment, the system uses GIS technology to accurately divide the oil sunflower field by fusing multi-source data such as hyperspectral data, soil moisture and temperature, ensuring that different management areas are accurately fertilized according to their soil characteristics and crop needs. This differentiated fertilization strategy can reduce over-fertilization and fertilizer waste and optimize fertilizer utilization efficiency. By establishing a crop growth model through deep learning technology, the system can accurately calculate the required amount of fertilizer according to the specific growth conditions and historical data of each region, thereby optimizing the growth conditions of oil sunflowers and increasing the overall yield of crops. This data-driven strategy can tailor fertilization plans for each region to avoid resource mismatches. The feedback and optimization module of the system can monitor the growth of oil sunflowers after fertilization is completed, collect feedback data, and dynamically adjust subsequent fertilization strategies based on these data. This can respond to environmental changes and fluctuations in crop demand in a timely manner, avoid over-fertilization or malnutrition, and further improve the accuracy of fertilization.

[0111] Example 2

[0112] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the data acquisition module includes a sensor data acquisition unit and a UAV hyperspectral imaging unit;

[0113] The sensor data acquisition unit collects the original soil moisture data S(x, y) and original temperature data T(x, y) at different positions of the field through the embedded soil moisture sensor and the embedded temperature sensor, and obtains the historical yield data Y of each plot from the historical database His (x,y);

[0114] The UAV hyperspectral imaging unit scans the entire sunflower field by carrying a hyperspectral imaging device on the UAV to obtain the original hyperspectral reflectance R(λ, x, y) of different bands;

[0115] Where R(λ,x,y) represents the original hyperspectral reflectance with wavelength λ at position (x,y).

[0116] The data preprocessing module includes a hyperspectral data processing unit and a soil data standardization unit;

[0117] The hyperspectral data processing unit uses a filter to perform denoising on the original hyperspectral reflectance R(λ, x, y) to remove random noise and obtain the denoised hyperspectral reflectance R denoise (λ,x,y);

[0118] Then the denoised hyperspectral reflectance R denoise (λ, x, y) is normalized to obtain the normalized hyperspectral reflectance R norm (λ,x,y), for the normalized hyperspectral reflectance R norm (λ, x, y) performs illumination condition correction to obtain the corrected hyperspectral reflectance R′(λ, x, y);

[0119] The corrected hyperspectral reflectance R′(λ,x,y) is obtained by the following formula:

[0120]

[0121] Where L(λ,x,y) represents the light intensity correction factor;

[0122] The soil data standardization unit performs standardization processing on the original soil moisture data S(x, y) and the original temperature data T(x, y) to obtain standardized soil moisture data S′(x, y) and standardized temperature data T′(x, y);

[0123] The standardized soil moisture data S′(x, y) is obtained by the following formula:

[0124]

[0125] In the formula, μS represents the mean of the original soil moisture data, and σS represents the standard deviation of the original soil moisture data;

[0126] The standardized temperature data T′(x, y) is obtained by the following formula:

[0127]

[0128] Wherein, μT represents the mean of the original temperature data, and σT represents the standard deviation of the original temperature data.

[0129] In this embodiment, the embedded soil moisture and temperature sensor can collect the original soil moisture data S(x, y) and original temperature data T(x, y) at different locations of the sunflower field through the sensor data acquisition unit, which can fully consider the differences in soil in different areas. At the same time, the historical yield data Y in the historical database is used His (x, y) provides a basis for decision-making and ensures that the fertilization strategy is more targeted. The UAV hyperspectral imaging unit can obtain hyperspectral reflectance in different bands by scanning the field. Through filter denoising, normalization processing and light condition correction, the system can eliminate the influence of environmental noise and light changes, and obtain accurate corrected hyperspectral reflectance R′(λ, x, y), providing high-quality data support for subsequent crop growth models.

[0130] This embodiment combines soil moisture, temperature and hyperspectral reflectance through multi-source fusion of sensor data and drone hyperspectral data, accurately describing the crop growth environment from multiple dimensions. Compared with the traditional fertilization system that relies on only a single data source, this embodiment can capture the environmental status more comprehensively and accurately. The present invention solves the problem that traditional fertilization systems are difficult to effectively deal with soil and environmental differences in different regions. Through hyperspectral technology and multi-source data collected by sensors, the system can achieve precise fertilization based on environmental differences in the field.

[0131] Example 3

[0132] This embodiment is explained in Example 2. Please refer to Figure 1 ,Specifically: the region division module includes a high-dimensional feature extraction unit and a GIS region division unit;

[0133] The high-dimensional feature extraction unit extracts features from each position (x, y) according to the corrected hyperspectral reflectance R′(λ, x, y), the standardized soil moisture data S′(x, y) and the standardized temperature data T′(x, y) to obtain a high-dimensional feature vector F(x, y);

[0134] The high-dimensional feature vector F(x, y) is obtained by the following formula:

[0135] F(x,y)=[R′(λ1,x,y),R′(λ2,x,y),,,R′(λ n ,x,y),...,S′(x,y),T′(x,y)];

[0136] The high-dimensional feature vector F(x,y) includes corrected hyperspectral reflectance at multiple wavelengths;

[0137] The GIS area division unit divides the field into multiple management areas through GIS, applies a clustering algorithm based on the high-dimensional feature vector F(x,y), and divides locations (x,y) with similar features into the same management area; obtains a set Z of management areas:

[0138] The set Z of management areas is obtained by the following formula:

[0139] Z={(x,y)|F(x,y)∈Cluster i};

[0140] In the formula, Cluster i represents the i-th management area obtained by the clustering algorithm;

[0141] Map the set of management areas Z to GIS, generate a regional division layer, and intuitively display each management area Z i The regional feature extraction module includes a regional data aggregation unit and a feature calculation unit;

[0142] The regional data aggregation unit aggregates data for all positions (x, y) of the divided area, and aggregates the corrected hyperspectral reflectance R′(λ, x, y), the standardized soil moisture data S′(x, y), and the standardized temperature data T′(x, y);

[0143] The classified data of the location points (x, y)∈Z within the set Z of the management area are summarized, including the set of corrected hyperspectral reflectance R′(λ, x, y) in the area, the set of standardized soil moisture data S′(x, y) and the set of standardized temperature data T′(x, y):

[0144] Among them, the set of corrected hyperspectral reflectance R′(λ,x,y) in the region is:

[0145] {R′(λ,x,y)|(x,y)∈Z};

[0146] The set of standardized soil moisture data S′(x,y) in the region is:

[0147] {S′(x,y)(x,y)∈Z};

[0148] The set of standardized temperature data T′(x,y) in the region is:

[0149] {T′(x,y)|(x,y)∈Z}.

[0150] The feature calculation unit calculates the average value of the corrected hyperspectral reflectance in the management area set Z

[0151] The average value of the corrected hyperspectral reflectance Obtained by the following formula:

[0152]

[0153] Where M represents the number of location points in the set Z of management areas;

[0154] Calculate the average soil moisture in a set of management areas Z and the average temperature

[0155] The average soil moisture Obtained by the following formula:

[0156]

[0157] The average value of the temperature Obtained by the following formula:

[0158]

[0159] The average value of the hyperspectral reflectance will be corrected Average soil moisture and the average temperature Combine to obtain the regional feature vector F i :

[0160] The regional feature vector F i Obtained by the following formula:

[0161]

[0162] In this embodiment, through the combination of the high-dimensional feature extraction unit and the GIS regional division unit, the system can use a clustering algorithm to divide locations with similar features into the same management area based on multi-source data including corrected hyperspectral reflectance R′(λ,x,y), standardized soil moisture data S′(x,y) and T′(x,y) standardized temperature data, and generate an intuitive regional division layer. This method avoids the limitations of traditional single-factor regional division, ensures that the internal characteristics of each management area are more consistent, and thus provides a refined basis for fertilization strategies. The regional feature extraction module summarizes the multi-dimensional data such as hyperspectral reflectance, soil moisture, and temperature within each management area through the regional data summary unit, and calculates the average value of the area through the feature calculation unit to extract the regional feature vector. Compared with traditional systems, the present invention combines more dimensional environmental information to ensure that the fertilization strategy can comprehensively consider the complex conditions of different regions and adapt to the diverse needs of different regions.

[0163] Example 4

[0164] This embodiment is explained in Example 3, please refer to Figure 1 ,Specifically: the fertilization strategy formulation module includes a crop growth model unit and a fertilizer amount calculation unit;

[0165] The crop growth model unit establishes a crop growth model by using deep learning technology, and according to the regional feature vector F i The crop growth model is used to predict the crop growth status and demand in each region, and the crop growth potential G is output. P and nutritional requirements req ;

[0166] The growth potential of the crop G P Obtained by the following formula:

[0167]

[0168] In the formula, α1 specifically represents the light parameter, α2 represents the soil parameter, α3 represents the temperature parameter, α4 represents the management measure parameter, and N represents the nitrogen content in the soil;

[0169] According to the growth potential of crops P Calculate the nutrient requirements of crops at different growth stages req ;

[0170] The nutritional requirement N req Obtained by the following formula:

[0171]

[0172] Where, b1, b2 and b3 represent the growth potential of crops, G P , the average soil moisture and the average temperature The fertilizer amount calculation unit calculates the nutrient requirement N provided by the growth model unit. req , calculate each management area Z i The amount of fertilizer E R , and generate the fertilization strategy E i ;

[0173] For a set of management areas Z, according to the nutrient demand N req Calculate the required amount of fertilizer E based on the nitrogen content N in the soil R ;

[0174] The amount of fertilizer E R Obtained by the following formula:

[0175]

[0176] Where N req (F i ) represents the regional feature vector F i Nutritional requirements req , represents the average soil nitrogen content in the set Z of management areas, ω1 and ω2 represent the regional characteristic vectors F i Nutritional requirements req and the weighted value of the mean soil nitrogen content within the set Z of management areas;

[0177] According to the obtained fertilizer amount E R , calculate the fertilization strategy E i ;

[0178] The fertilization strategy E i Obtained by the following formula:

[0179]

[0180] In the formula, and represents the empirical parameters obtained from field experiments and historical data, θ1 and θ2 represent the empirical weight coefficients (reflecting the impact of growth stage and climate conditions on fertilization time), GS represents the current crop growth stage, WC represents the sequence value of the current climate condition type, Md(x) represents the selection coefficient of the fertilization method, AF represents the adjustment factor, E f Indicates historical fertilization efficiency;

[0181] The selection coefficient Md(x) of the fertilization method is obtained by matching the following formula:

[0182]

[0183] System Utilization Fertilization Strategy E i , control the amount and method of fertilizer use, and regulate the growth of crops

[0184] In this embodiment, the crop growth model unit uses the deep learning technology to establish a crop growth model, and the system can generate a crop growth model based on the feature vectors F of different regions. i Including correction for the average value of hyperspectral reflectance Average soil moisture and the average temperature Accurately predict the growth status and nutritional needs of crops in each area. This prediction method greatly improves the insight into the growth potential of crops. It combines multi-source data for refined analysis to ensure that crops in different areas receive the most appropriate nutritional support. The fertilizer calculation unit dynamically adjusts the fertilizer amount in each management area based on the nutritional needs of crops and the nitrogen content in the soil, combined with the weight coefficient. This process fully considers the differences in soil conditions and crop growth stages, avoids excessive or insufficient use of fertilizers, further reduces resource waste and improves fertilizer utilization, and maximizes crop yields and growth potential.

[0185] The fertilization strategy not only takes into account the growth potential and nutritional needs of crops, but also comprehensively calculates the fertilization strategy through historical yield data, climatic conditions, crop growth stages and other factors. The fertilization time, fertilization method and fertilization amount are dynamically adjusted through weight coefficients and adjustment factors to ensure that the fertilization strategy is more flexible and accurate. This optimization scheme reduces the limitations of traditional fertilization modes, making fertilization strategies more intelligent and more adaptable, especially in complex agricultural environments, it can effectively respond to changes in crop demand. This embodiment comprehensively considers regional nutritional needs and nitrogen content in the soil through the fertilizer amount calculation formula, combines historical data and empirical weight coefficients, and implements differentiated fertilization strategies for different management areas. This differentiated fertilization amount calculation avoids the unified fertilization mode in traditional methods, and can provide personalized fertilizer use plans for each region, ensuring the optimal allocation of resources and reducing the risk of environmental pollution.

[0186] Example 5

[0187] A precise fertilization control system for oil sunflower fields based on UAV hyperspectral technology, please refer to Figure 1 ,Specifically: the feedback and optimization module includes a growth monitoring unit and a strategy optimization unit;

[0188] The growth monitoring unit collects the growth data of the oil sunflower, including the growth height H and the chlorophyll content C, and compares it with the expected growth target to calculate the growth deviation ΔG(t):

[0189] The growth deviation ΔG(t) is obtained by the following formula:

[0190] ΔG(t)=Gx(t)-Ga(t);

[0191] In the formula, Gx(t) represents the expected growth target predicted by the growth model at time t, and Ga(t) represents the actual growth situation monitored at time t;

[0192] When ΔG(t)≤0, it means that there is no need to adjust the fertilization strategy E i ;

[0193] When ΔG(t)>0, it means that the fertilization strategy E needs to be adjusted. i ; According to the growth deviation ΔG(t) and the fertilizer amount E R , get the new fertilizer amount E next ;

[0194] The new fertilizer amount E next Obtained by the following formula:

[0195] E next =E R +k*ΔG(t)*E f ;

[0196] Wherein, k represents the feedback gain parameter.

[0197] The strategy optimization unit uses the new fertilizer amount E next and the actual growth situation Ga(t) at time t, and obtain the adjusted fertilization strategy E i，next , used to regulate crop growth.

[0198] The adjusted fertilization strategy E i，next Obtained by the following formula:

[0199]

[0200] In this embodiment, the key growth data of oil sunflower, including growth height H and chlorophyll content C, are collected by the growth monitoring unit. The system can compare the monitoring data with the expected growth target and calculate the growth deviation. The judgment condition of growth deviation is defined in the formula. When the growth deviation ΔG(t) is greater than 0, the adjustment of the fertilization strategy will be triggered in time. Such real-time monitoring and feedback mechanism ensures that crop growth always remains in an ideal state, which helps to quickly respond to environmental changes and changes in crop demand. Through the feedback gain formula of growth deviation and fertilizer application, the new fertilizer application amount can be accurately regulated according to the actual growth of the crop. The introduction of feedback gain parameter k enhances the adaptability of the fertilization strategy, and can dynamically adjust the fertilizer application amount according to the specific growth conditions of the crop, avoid excessive or insufficient fertilizer application, thereby improving the utilization efficiency of fertilizers, reducing resource waste, and reducing the negative impact on the environment.

[0201] The strategy optimization unit dynamically generates an adjusted fertilization strategy based on the new fertilizer amount and actual growth conditions. This automated and intelligent optimization process makes the fertilization strategy more precise and flexible, and can be adjusted according to the real-time needs of the crop to ensure the rationality of the crop's nutrient supply. This mechanism not only ensures the healthy growth of sunflowers, but also significantly improves the final yield and quality of the crop.

[0202] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A sunflower field precision fertilization control system based on drone hyperspectral technology, characterized by: It includes data acquisition module, data preprocessing module, regional division module, regional feature extraction module, fertilization strategy formulation module and feedback and optimization module; The data acquisition module collects the original soil moisture data S(x, y) and the original temperature data T(x, y) through sensors, and obtains the historical yield data Y from the historical yield. His (x, y), and use a drone equipped with a hyperspectral imaging device to perform a flight scan of the sunflower field to obtain the original hyperspectral reflectance R(λ, x, y) of different bands; The data preprocessing module performs denoising, normalization and illumination correction on the collected hyperspectral reflectance R(λ,x,y), obtains the processed corrected hyperspectral reflectance R′(λ,x,y), and standardizes the original soil moisture data S(x,y) and the original temperature data T(x,y), and obtains standardized soil moisture data S′(x,y) and standardized temperature data T′(x,y); The region division module uses the corrected hyperspectral reflectance, standardized soil moisture data S′(x,y) and standardized temperature data T′(x,y) to obtain a high-dimensional feature vector F(x,y) and divide the field into multiple management areas through GIS; The regional feature extraction module obtains a set Z of management areas according to the high-dimensional feature vector F(x, y) for the multiple management areas divided by the regional division module, and extracts the regional feature vector F i Including the average value R of the corrected hyperspectral reflectance i ′(λ), the average soil moisture S i ′ and the average temperature T i ′; The fertilization strategy formulation module is based on the regional feature vector F i , by using deep learning technology to establish a crop growth model, and use the crop growth model to calculate the amount of fertilizer E for each area R , and generate the fertilization strategy E i ; After fertilization is completed, the feedback and optimization module collects feedback data by monitoring the growth of oil sunflowers and generates a new fertilization strategy E i，next .

2. According to claim 1, a sunflower field precision fertilization control system based on unmanned aerial vehicle hyperspectral technology is characterized by: The data acquisition module includes a sensor data acquisition unit and a UAV hyperspectral imaging unit; The sensor data acquisition unit collects the original soil moisture data S(x, y) and original temperature data T(x, y) at different positions of the field through the embedded soil moisture sensor and the embedded temperature sensor, and obtains the historical yield data Y of each plot from the historical database His (x,y); The UAV hyperspectral imaging unit scans the entire sunflower field by carrying a hyperspectral imaging device on the UAV to obtain the original hyperspectral reflectance R(λ, x, y) of different bands; Where R(λ,x,y) represents the original hyperspectral reflectance with wavelength λ at position (x,y).

3. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 2 is characterized by: The data preprocessing module includes a hyperspectral data processing unit and a soil data standardization unit; The hyperspectral data processing unit uses a filter to perform denoising on the original hyperspectral reflectance R(λ, x, y) to remove random noise and obtain the denoised hyperspectral reflectance R denoise (λ,x,y); Then the denoised hyperspectral reflectance R denoise (λ, x, y) is normalized to obtain the normalized hyperspectral reflectance R norm (λ,x,y), for the normalized hyperspectral reflectance R norm (λ, x, y) performs illumination condition correction to obtain the corrected hyperspectral reflectance R′(λ, x, y); The corrected hyperspectral reflectance R′(λ,x,y) is obtained by the following formula: Where L(λ,x,y) represents the light intensity correction factor; The soil data standardization unit performs standardization processing on the original soil moisture data S(x, y) and the original temperature data T(x, y) to obtain standardized soil moisture data S′(x, y) and standardized temperature data T′(x, y); The standardized soil moisture data S′(x, y) is obtained by the following formula: In the formula, μS represents the mean of the original soil moisture data, and σS represents the standard deviation of the original soil moisture data; The standardized temperature data T′(x, y) is obtained by the following formula: Wherein, μT represents the mean of the original temperature data, and σT represents the standard deviation of the original temperature data.

4. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 1 is characterized by: The region division module includes a high-dimensional feature extraction unit and a GIS region division unit; The high-dimensional feature extraction unit extracts features from each position (x, y) according to the corrected hyperspectral reflectance R′(λ, x, y), the standardized soil moisture data S′(x, y) and the standardized temperature data T′(x, y) to obtain a high-dimensional feature vector F(x, y); The high-dimensional feature vector F(x, y) is obtained by the following formula: F(x,y)=[R′(λ1,x,y),R′(λ2,x,y),,,R′(λ n ,x,y),...,S′(x,y),T′(x,y)]; The high-dimensional feature vector F(x,y) includes multiple wavelengths λ n Corrected hyperspectral reflectance under ; The GIS area division unit divides the field into multiple management areas through GIS, applies a clustering algorithm based on the high-dimensional feature vector F(x,y), and divides locations (x,y) with similar features into the same management area; obtains a set Z of management areas: The set Z of management areas is obtained by the following formula: Z={(x,y)|F(x,y)∈Cluster i }; In the formula, Cluster i represents the i-th management area obtained by the clustering algorithm; Map the set of management areas Z to GIS, generate a regional division layer, and intuitively display each management area Z i location and range.

5. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 4 is characterized by: The regional feature extraction module includes a regional data aggregation unit and a feature calculation unit; The regional data aggregation unit aggregates data for all positions (x, y) of the divided area, and aggregates the corrected hyperspectral reflectance R′(λ, x, y), the standardized soil moisture data S′(x, y), and the standardized temperature data T′(x, y); The classified data of the location points (x, y)∈Z within the set Z of the management area are summarized, including the set of corrected hyperspectral reflectance R′(λ, x, y) in the area, the set of standardized soil moisture data S′(x, y) and the set of standardized temperature data T′(x, y): Among them, the set of corrected hyperspectral reflectance R′(λ,x,y) in the region is: {R′(λ,x,y)|(x,y)∈Z}; The set of standardized soil moisture data S′(x,y) in the region is: {S′(x,y)|(x,y)∈Z}; The set of standardized temperature data T′(x,y) in the region is: {T′(x,y)|(x,y)∈Z}.

6. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 5 is characterized by: The feature calculation unit calculates the average value of the corrected hyperspectral reflectance in the management area set Z The average value of the corrected hyperspectral reflectance Obtained by the following formula: Where M represents the number of location points in the set Z of management areas; Calculate the average soil moisture in a set of management areas Z and the average temperature The average soil moisture Obtained by the following formula: The average value of the temperature Obtained by the following formula: The average value of the hyperspectral reflectance will be corrected Average soil moisture and the average temperature Combine to obtain the regional feature vector F i : The regional feature vector F i Obtained by the following formula:

7. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 1 is characterized by: The fertilization strategy formulation module includes a crop growth model unit and a fertilization amount calculation unit; The crop growth model unit establishes a crop growth model by using deep learning technology, and according to the regional feature vector F i The crop growth model is used to predict the crop growth status and demand in each region, and the crop growth potential G is output. P and nutritional requirements req ; The growth potential of the crop G P Obtained by the following formula: In the formula, α1 represents light parameters, α2 represents soil parameters, α3 represents temperature parameters, α4 represents management measures parameters, and N represents the nitrogen content in the soil; According to the growth potential of crops P Calculate the nutrient requirements of crops at different growth stages req ; The nutritional requirement N req Obtained by the following formula: Where, b1, b2 and b3 represent the growth potential of crops, G P , the average soil moisture and the average temperature The weight coefficient of .

8. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 7 is characterized by: The fertilizer amount calculation unit calculates the nutrient requirement N provided by the growth model unit. req , calculate each management area Z i The amount of fertilizer E R , and generate the fertilization strategy E i ; For a set of management areas Z, according to the nutrient demand N req Calculate the required amount of fertilizer E based on the nitrogen content N in the soil R ; The amount of fertilizer E R Obtained by the following formula: Where N req (F i ) represents the regional feature vector F i Nutritional requirements req , represents the average soil nitrogen content in the set Z of management areas, ω1 and ω2 represent the regional characteristic vectors F i Nutritional requirements req and the weighted value of the mean soil nitrogen content within the set Z of management areas; According to the obtained fertilizer amount E R , calculate the fertilization strategy E i ; The fertilization strategy E i Obtained by the following formula: In the formula, and represents the empirical parameters obtained from field experiments and historical data, θ1 and θ2 represent the empirical weight coefficients, GS represents the current crop growth stage, WC represents the sequence value of the current climate condition type, Md(x) represents the selection coefficient of the fertilization method, AF represents the adjustment factor, E f Indicates historical fertilization efficiency; The selection coefficient Md(x) of the fertilization method is obtained by matching the following formula: System Utilization Fertilization Strategy E i , control the amount and method of fertilizer use, and regulate the growth of crops.

9. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 8 is characterized by: The feedback and optimization module includes a growth monitoring unit and a strategy optimization unit; The growth monitoring unit collects the growth data of the oil sunflower, including the growth height H and the chlorophyll content C, and compares it with the expected growth target to calculate the growth deviation ΔG(t): The growth deviation ΔG(t) is obtained by the following formula: ΔG(t)=Gx(t)-Ga(t); In the formula, Gx(t) represents the expected growth target predicted by the growth model at time t, and Ga(t) represents the actual growth situation monitored at time t; When ΔG(t)≤0, it means that there is no need to adjust the fertilization strategy E i ; When ΔG(t)>0, it means that the fertilization strategy E needs to be adjusted. i ; According to the growth deviation ΔG(t) and the fertilizer amount E R , get the new fertilizer amount E next ; The new fertilizer amount E next Obtained by the following formula: AND next =And R +k*ΔG(t)*E f ; Wherein, k represents the feedback gain parameter.

10. The oil sunflower field precision fertilization control system based on UAV hyperspectral technology according to claim 9, characterized in that: The strategy optimization unit uses the new fertilizer amount E next and the actual growth situation Ga(t) at time t, and obtain the adjusted fertilization strategy E i，next , used for crop growth regulation; The adjusted fertilization strategy E i，next Obtained by the following formula:

Images