Agricultural fertilization method and system based on big data analysis

Abstract

The present application relates to the field of wisdom agricultural technology, specifically to an agricultural fertilization method and system based on big data analysis, comprising the following steps: capturing reflected spectrum to identify soil state, establishing multi-source agricultural condition representation set, evaluating vegetation situation and extracting missing degree to establish comprehensive nutrient deficiency characteristics, combining agricultural machinery moving speed to convert lag displacement to obtain fertilizer discharge touch front node, converting fertilizer discharge quality to infer motor rotation number to generate fertilization driving instruction. In the present application, the soil resistance state is identified by combining probe voltage, the vegetation growth situation is evaluated by calculating spectrum characteristics, the physiological and fertilizer missing degree is extracted, the lag displacement is converted by monitoring moving speed, the instruction is obtained by inferring motor rotation number, the nutrient difference and vegetation demand in farmland area are perceived, the limitation of traditional fixed rate fertilizer discharge is broken through, the fertilizer release amount is dynamically adjusted according to the plot state, and the local fertilizer accumulation or nutrient supply missing phenomenon is avoided.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture technology, and in particular to an agricultural fertilization method and system based on big data analysis. Background Technology

[0002] The field of smart agriculture technology is a comprehensive category that integrates the Internet of Things, cloud computing, artificial intelligence, and geographic information technology into the agricultural production and management process. The core aspects of this field include the sensing and collection of farmland physical environment information, dynamic monitoring of crop growth status, and automated operation control of agricultural machinery and equipment. It establishes a control and execution system covering the entire cycle from sowing, irrigation, and fertilization to pest and disease control. Traditional agricultural fertilization methods refer to the agricultural operations of supplementing and allocating nitrogen, phosphorus, and potassium nutrients to crops at different growth stages. This involves manually spreading fertilizer bags by hand or using a tractor-trailer-driven mechanical fertilizer applicator to spread the fertilizer evenly across the entire field. Operators mainly determine the timing of application and the type of fertilizer based on visual observation of crop leaf color and calendar solar terms. In the specific mechanical fertilization process, the tractor's power take-off shaft drives the fertilizer discharge wheel connected to the bottom of the fertilizer hopper to rotate physically, causing the granular fertilizer to fall onto the soil surface at a constant mechanical rate. Alternatively, the furrow opener at the front of the fertilizer applicator can create shallow furrows in the soil surface between crop rows, and the fertilizer can be guided downwards into the furrows through the fertilizer discharge plastic pipe connected to the hopper. Finally, a metal covering plate towed at the rear can level the turned-up soil and cover the fertilizer.

[0003] Traditional agricultural fertilization mainly relies on manual spreading or mechanical spreading across the entire field. Operators subjectively determine the timing of fertilization based on visual observation of crop leaf color and calendar solar terms. This extensive operation mode cannot accurately perceive the differences in nutrients in different areas of the farmland and the real-time growth status of vegetation. Mechanical fertilization can only maintain a fixed rate of fertilizer release into the soil and cannot dynamically adjust according to local soil resistance and fertility deficiency. The fixed fertilizer release mechanism leads to a disconnect between the amount of fertilizer applied and the actual needs of the crop, which can easily cause fertilizer accumulation or insufficient nutrient supply in some areas, thus affecting the overall crop yield. Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention provides an agricultural fertilization method based on big data analysis, comprising the following steps: S1: Based on cloud-based farmland nutrient distribution archives, near-infrared reflectance and red reflectance spectra are captured by multispectral devices to identify soil electrical impedance status and establish a multi-source agricultural condition characterization set; S2: Using the multi-source agricultural condition characterization set, the near-infrared reflectance spectrum is subtracted from the red reflectance spectrum and divided by the total sum of the spectra to assess the vegetation growth status, determine the relationship between the constant voltage applied by the probe and the soil electrical impedance state, and obtain the source plot characteristic quantities. S3: Based on the source plot characteristics, the regional vegetation growth benchmark value is set and the vegetation growth status is deducted to extract the degree of physiological deficiency. The electrical conductivity benchmark is deducted and the soil electrical conductivity characteristics are extracted to extract the degree of fertility deficiency. The degree of physiological deficiency and the degree of fertility deficiency are adjusted and weighted and summed according to the set weight coefficient ratio to establish a comprehensive nutrient deficiency characteristic. S4: Utilizing the comprehensive nutrient deficiency characteristics, monitor the real-time longitude and latitude coordinates and the agricultural machinery movement speed, multiply the movement speed by the mechanical delay time of the fertilizer applicator, calculate the lag displacement span, compare the lag displacement span with the Earth's meridian crossing length benchmark, derive the latitude offset, and obtain the fertilizer applicator touch control front node. S5: Using the aforementioned fertilizer discharge touch front node, calculate the comprehensive nutrient deficiency characteristics multiplied by the fertilizer conversion ratio, convert the total fertilizer discharge mass, adjust the relationship between the total fertilizer discharge mass and the constant fertilizer mass released per cycle, infer the number of motor rotations, generate alternating pulse signals, and obtain the fertilizer application drive command.

[0005] As a further aspect of the present invention, the multi-source agricultural characterization set includes a canopy optical image set, a substrate dielectric property library, and a cultivated layer habitat matrix; the source-end plot characteristics include a greenness index matrix, an ion ionization rate distribution map, and regional phenotypic parameters; the comprehensive nutrient deficiency characteristics include macro-element allocation extreme values, micro-element depletion indicators, and expected biochemical compensation amounts; the fertilizer application control pre-node includes a spatial advance trigger point, servo activation timing, and motion compensation margin; and the fertilizer application drive command includes a hardware control duty cycle, stepper transmission parameters, and valve opening and closing frequency.

[0006] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Based on cloud-based farmland nutrient distribution archives, near-infrared reflectance and red reflectance spectra are captured using multispectral equipment. The coordinate sequence embedded in the nutrient distribution archives is extracted, and the near-infrared reflectance and red reflectance spectra are registered and aligned according to the coordinate sequence. The pixel values ​​of the near-infrared reflectance spectra are subtracted from the pixel values ​​of the red reflectance spectra and divided by the total number of pixels. The spectral bands corresponding to the ratios greater than the set spectral limit values ​​are extracted to obtain the canopy optical map. S102: For the canopy optical spectrum, detect the soil electrical impedance state parameter associated with the sampling area, extract the impedance parameter value range fluctuation range, perform difference calculation between the impedance parameter value range fluctuation range and the preset impedance reference parameter, record the frequency of impedance difference exceeding the resistance change limit value, assign weights to the pixel coordinate points attached to the canopy optical spectrum according to the frequency of occurrence, remove the associated feature points whose weight values ​​do not reach the lower limit of the weight reference, and obtain the substrate dielectric property set; S103: Call the substrate dielectric property set, read the built-in resistance feature value and optical pixel vector, perform matrix concatenation of the resistance feature value and optical pixel vector according to the coordinate index order to establish a feature matrix, calculate the covariance value between each column vector in the feature matrix, remove the associated column vectors whose covariance value is less than the data correlation threshold, aggregate the remaining column vectors and perform dimensional format conversion to establish a multi-source agricultural condition characterization set.

[0007] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Based on the multi-source agricultural condition characterization set, extract near-infrared reflectance spectrum and red reflectance spectrum, subtract the difference between the near-infrared spectrum pixel value and the red spectrum pixel value, divide the difference by the sum of the two spectral pixel values, obtain the vegetation feature numerical sequence, calculate the arithmetic mean of the vegetation feature numerical sequence, extract the associated coordinate sites within the vegetation feature numerical sequence that are greater than the arithmetic mean, converge the coordinate sites and associated pixel values, and generate a greenness feature matrix. S202: For the greenness feature matrix, a constant voltage value is applied to the monitoring probe, and the soil electrical impedance state sequence at the corresponding coordinate is extracted simultaneously. The constant voltage value is divided by the impedance value in the soil electrical impedance state sequence to obtain the node conductivity parameter. The median of the conductivity parameter is taken as the conductivity reference. The difference between the conductivity parameter and the conductivity reference is calculated, and the coordinates of the positions where the difference exceeds the mean of all differences are selected. The conductivity parameters are vector-stitched according to the coordinate order to establish the impedance state distribution map. S203: Call the impedance state distribution map, extract the node conductivity parameters and location coordinates, match the vegetation feature values ​​under the corresponding coordinates in the greenness feature matrix, perform array splicing of the vegetation feature values ​​and node conductivity parameters according to coordinate index, construct a multi-dimensional feature vector group, calculate the Euclidean distance values ​​between each column vector in the multi-dimensional feature vector group, remove the associated discrete vector data whose Euclidean distance values ​​are greater than the median distance, aggregate the remaining vector sequence, and obtain the source plot feature quantity.

[0008] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Based on the source plot feature quantity, extract the vegetation growth status sequence, read the regional vegetation growth benchmark value, subtract the value of each element in the vegetation growth status sequence from the vegetation growth benchmark value to obtain the difference sequence, process the absolute value of each difference in the difference sequence, remove the coordinate points whose absolute value is less than the growth deviation limit value, extract the remaining coordinate points and the corresponding differences and perform vector splicing to obtain the physiological deficiency assessment matrix. S302: Call the physiological deficiency assessment matrix, extract coordinate sites, retrieve soil electrical conductivity characteristic parameters associated with coordinate sites, retrieve the pre-stored soil electrical conductivity benchmark threshold, calculate the difference between the soil electrical conductivity benchmark threshold and the soil electrical conductivity characteristic parameters, i.e., calculate the soil electrical conductivity compensation difference, compare the difference with the electrical conductivity fluctuation limit value, record the node index corresponding to the difference exceeding the electrical conductivity fluctuation limit value, extract the difference according to the node index and sort it in descending order to establish a fertility deficiency characterization vector; S303: For the fertility deficiency characterization vector, extract the differences of each item arranged in descending order, match the differences associated with the same coordinate positions in the physiological deficiency assessment matrix, perform a weighted summation operation on the two sets of differences according to the set weight coefficient ratio, obtain the node fusion feature value, compare the node fusion feature value with the nutrient compensation limit value, retain the coordinates associated with the node fusion feature value that has not reached the nutrient compensation limit value, and obtain the comprehensive nutrient deficiency feature.

[0009] As a further aspect of the present invention, the physiological deficiency weight coefficient and the fertility deficiency weight coefficient included in the set weight coefficient ratio are extracted, and the correlation difference of the same coordinate position in the physiological deficiency assessment matrix is ​​multiplied by the physiological deficiency weight coefficient to obtain the physiological feature weight product value. The differences of each item arranged in descending order are multiplied by the fertility deficiency weight coefficient to obtain the fertility feature weight product value. The physiological feature weight product value is added to the fertility feature weight product value to obtain the node fusion transition feature value.

[0010] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Extract the built-in regional locations of the comprehensive nutrient deficiency features, monitor the real-time longitude and latitude coordinates corresponding to the regional locations, collect the agricultural machinery movement speed parameters, retrieve the mechanical delay time value of the built-in fertilizer applicator, multiply the agricultural machinery movement speed parameters and the mechanical delay time value of the fertilizer applicator to obtain the product value sequence, extract the relevant elements in the product value sequence whose values ​​are greater than the displacement limit benchmark, aggregate the elements and their corresponding longitude and latitude coordinates, and establish the lag displacement span vector; S402: Read the value of the element and the corresponding latitude coordinate of the hysteresis displacement span vector, retrieve the reference value of the span length of the Earth's meridian, divide the element value by the reference value of the span length of the Earth's meridian to obtain the reference ratio parameter, perform array splicing on the reference ratio parameter according to the latitude coordinate sequence, calculate the deviation value of each parameter in the array from the standard latitude interval, remove the discrete coordinate items associated with deviation values ​​exceeding the offset judgment threshold, aggregate the coordinates of the remaining items and the corresponding deviation values ​​to obtain the latitude offset matrix; S403: Extract the coordinates and corresponding deviation values ​​of the latitude offset matrix, match the corresponding site feature parameters in the comprehensive nutrient deficiency features, sum the deviation values ​​and site feature parameters to obtain the node trigger weight parameter, compare the node trigger weight parameter with the preset trigger threshold, retain the spatial coordinates corresponding to the node trigger weight parameter being greater than the preset trigger threshold, arrange the spatial coordinates in ascending order along the agricultural machinery travel sequence, extract the front site coordinate data array, and obtain the fertilizer discharge touch front node.

[0011] As a further aspect of the present invention, a preset feature scaling factor and a preset offset weighting factor are extracted, and the site feature parameter is multiplied by the preset feature scaling factor to obtain the value of the mapped feature parameter. The deviation value is multiplied by the preset offset weighting coefficient to obtain the weighted offset feature value. The mapping feature parameter value is then summed with the weighted offset feature value to obtain the trigger evaluation benchmark parameter.

[0012] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Extract the coordinate sequence of the fertilizer discharge touch front node, retrieve the comprehensive nutrient deficiency feature parameter of the corresponding position, retrieve the fertilizer conversion ratio parameter built into the device, multiply the comprehensive nutrient deficiency feature parameter with the fertilizer conversion ratio parameter to obtain the total fertilizer discharge mass value, subtract the mass lower limit benchmark from the total fertilizer discharge mass value to obtain the difference, remove the coordinate positions associated with the difference not reaching the judgment benchmark, extract the remaining positions and values ​​and perform array splicing to generate the fertilizer discharge mass matrix; S502: Call the fertilizer discharge mass matrix, extract the total mass value, retrieve the constant fertilizer mass value released by the fertilizer discharger in a single release, perform a division operation between the total fertilizer discharge mass value and the constant fertilizer mass value released in a single release, obtain the node quotient sequence, extract each element in the node quotient sequence and perform a rounding operation to obtain the integer period parameter, compare the integer period parameter with the motor load limit parameter, retain the associated coordinates where the integer period parameter does not exceed the motor load limit parameter, and establish the rotation number vector; S503: For the rotation number vector, extract the integer period parameter, collect the single-cycle pulse number parameter, multiply the integer period parameter and the single-cycle pulse number parameter to obtain the alternating pulse signal sequence, read the set duty cycle reference value, convert the waveform width of the alternating pulse signal sequence according to the duty cycle reference value, extract the converted pulse frequency and compare it with the upper limit value of the hardware frequency, remove the associated nodes whose frequency exceeds the upper limit value, and obtain the fertilization drive command.

[0013] An agricultural fertilization system based on big data analytics includes: The agricultural monitoring module, based on cloud-based farmland nutrient distribution archives, captures near-infrared reflectance and red reflectance spectra using multispectral equipment to identify soil electrical impedance status and establish a multi-source agricultural characterization set. The nutrient deficiency assessment module uses the multi-source agricultural condition characterization set, subtracts the red light reflectance spectrum from the near-infrared reflectance spectrum and divides it by the total sum of the spectra, to assess the vegetation growth status and obtain the source plot characteristic quantities. The fertilization derivation module, based on the source plot characteristics, sets a regional vegetation growth benchmark value, deducts the vegetation growth status, extracts the degree of physiological deficiency, deducts soil electrical conductivity characteristics from the electrical conductivity benchmark, extracts the degree of fertility deficiency, and establishes comprehensive nutrient deficiency characteristics. The fertilizer discharge target conversion module uses the comprehensive nutrient deficiency characteristics to monitor real-time longitude and latitude coordinates and agricultural machinery movement speed, multiplies the movement speed by the mechanical delay time of the fertilizer discharge device, converts the lag displacement span, and obtains the fertilizer discharge touch control front node. The motor drive control module uses the fertilizer discharge touch front node to calculate the comprehensive nutrient deficiency characteristics and multiply them by the fertilizer conversion ratio to convert the total fertilizer discharge mass, adjust the relationship between the total fertilizer discharge mass and the constant fertilizer mass released in a single operation, and obtain the fertilizer application drive command.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, near-infrared and red light reflectance spectra are captured by multispectral equipment and combined with constant voltage probes to identify soil electrical impedance status. Spectral characteristics are calculated to assess vegetation growth status, extract physiological and fertility deficiency levels, and establish comprehensive nutrient deficiency characteristics. Monitoring coordinates and agricultural machinery movement speed are used to convert lag displacement and derive latitudinal offset to obtain fertilizer pre-discharge nodes. The comprehensive nutrient deficiency characteristics are multiplied by the conversion ratio to calculate fertilizer discharge quality, infer the number of motor rotations, and obtain fertilizer drive commands. This accurately senses nutrient differences in different areas of farmland and real-time vegetation needs, breaking through the limitations of traditional fixed-rate mechanical fertilizer discharge. The amount of fertilizer released per application is dynamically adjusted according to the local conditions of the plot, effectively avoiding local fertilizer accumulation or nutrient supply deficiencies. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] Please see Figure 1 This invention provides an agricultural fertilization method based on big data analysis, comprising the following steps: S1: Based on cloud-based farmland nutrient distribution archives, near-infrared reflectance spectra and red reflectance spectra are captured by multispectral equipment to identify soil electrical impedance status, remove interference items that are out of the real-time reflectance spectrum range from the cloud-based farmland nutrient distribution archive records, and establish a multi-source agricultural condition characterization set; S2: Using a multi-source agricultural characterization set, the near-infrared reflectance spectrum is subtracted from the red reflectance spectrum and divided by the total sum of the spectra to assess the vegetation growth status, determine the relationship between the constant voltage applied by the probe and the soil electrical impedance state, convert the surface soil electrical conductivity distribution characteristics, optimize the combination of vegetation growth status and surface soil electrical conductivity distribution characteristics, and obtain the source plot characteristic quantities. S3: Based on the source plot characteristics, the regional vegetation growth benchmark value is set and the vegetation growth status is deducted to extract the degree of physiological deficiency. The electrical conductivity benchmark is deducted and the soil electrical conductivity characteristics are extracted to extract the degree of fertility deficiency. The degree of physiological deficiency and the degree of fertility deficiency are adjusted and weighted and summed according to the set weight coefficient ratio to analyze the global deficiency attributes and establish comprehensive nutrient deficiency characteristics. S4: Utilizing the comprehensive nutrient deficiency characteristics, monitor real-time longitude and latitude coordinates and agricultural machinery movement speed, multiply the movement speed by the mechanical delay time of the fertilizer applicator, calculate the lag displacement span, compare the lag displacement span with the benchmark of the length span of the Earth's meridian, derive the latitude offset, and obtain the fertilizer applicator touch control front node. S5: The fertilizer discharge touch control front node is adopted. The comprehensive nutrient deficiency characteristics are multiplied with the fertilizer conversion ratio to calculate the total fertilizer discharge mass. The relationship between the total fertilizer discharge mass and the constant fertilizer mass released in a single operation is adjusted to infer the number of motor rotations and generate alternating pulse signals. If the real-time coordinates overlap with the node coordinates, the alternating pulse signals are sent to execute the rotation operation and obtain the fertilizer application drive command. The multi-source agricultural characterization set includes a canopy optical image set, a substrate dielectric property library, and a topsoil habitat matrix. Source-end plot characteristics include a greenness index matrix, an ion ionization rate distribution map, and regional phenotypic parameters. Comprehensive nutrient deficiency characteristics include macro-element allocation extremes, micro-element depletion indicators, and expected biochemical compensation amounts. Fertilizer application control pre-nodes include spatial advance triggering points, servo activation timing, and motion compensation margin. Fertilizer application drive commands include hardware control duty cycle, stepper motor parameters, and valve opening and closing frequency.

[0020] Please see Figure 2 The specific steps of S1 are as follows: S101: Based on cloud-based farmland nutrient distribution archives, near-infrared reflectance and red reflectance spectra are captured using multispectral equipment. The coordinate sequence embedded in the nutrient distribution archives is extracted, and the near-infrared reflectance and red reflectance spectra are registered and aligned according to the coordinate sequence. The pixel values ​​of the near-infrared reflectance spectra are subtracted from the pixel values ​​of the red reflectance spectra and divided by the total number of pixels. The spectral bands corresponding to the ratios greater than the set spectral limit values ​​are extracted to obtain the canopy optical map. A multispectral camera array mounted on a drone flying at an altitude of 50 meters continuously captures surface reflectance spectral data at a frequency of 10 frames per second, obtaining raw near-infrared reflectance spectral images and raw red reflectance spectral images in a 1024x1024 pixel two-dimensional matrix format. The photosensitive band of the near-infrared reflectance spectrum is strictly limited to 800 nm to 900 nm, and the photosensitive band of the red reflectance spectrum is strictly limited to 630 nm to 690 nm. This step performs Gaussian filtering noise reduction on the acquired raw image two-dimensional matrix, using a 3x3 convolution kernel to perform a traversal convolution operation on each pixel in the matrix, replacing the center pixel value with the weighted average of the surrounding nine pixel values, and removing isolated noise data caused by device thermal noise. The pre-stored farmland nutrient distribution archives in the cloud server are retrieved, and the coordinate sequences with GPS coordinates are extracted. The denoised near-infrared reflectance spectral matrix and the red reflectance spectral matrix are then registered and aligned pixel-by-pixel according to the latitude and longitude nodes in the coordinate sequences, ensuring absolute consistency of the two spectral data at the same geographic coordinates in the matrix indices. After alignment, for each geographic coordinate index, the corresponding near-infrared reflectance spectral pixel values ​​and red reflectance spectral pixel values ​​are extracted. The spectral reflectance difference is obtained by subtracting the red reflectance spectral pixel value from the near-infrared reflectance spectral pixel value. This spectral reflectance difference is then divided by the sum of the near-infrared and red pixel values ​​at that coordinate point to derive the normalized spectral ratio, thus obtaining the canopy optical map.

[0021] S102: For the canopy optical spectrum, detect the soil electrical impedance state parameter associated with the sampling area, extract the impedance parameter value range fluctuation range, perform difference calculation between the impedance parameter value range fluctuation range and the preset impedance benchmark parameter, record the frequency of impedance difference exceeding the resistance change limit value, assign weights to the pixel coordinate points attached to the canopy optical spectrum according to the frequency of occurrence, remove the associated feature points whose weight values ​​do not reach the lower limit of the weight benchmark, and obtain the substrate dielectric property set; A ground-penetrating array electrode sensor is used to apply an alternating excitation current with a frequency of 1000 Hz to the corresponding coordinate region. Soil resistivity parameters at depths of 15 cm to 30 cm are detected and collected in real time, generating a one-dimensional time series containing 1000 discrete impedance values. This one-dimensional time series is traversed, and the upper limit impedance parameter with the largest value and the lower limit impedance parameter with the smallest value are extracted. The upper limit impedance parameter is subtracted from the lower limit impedance parameter to obtain the specific impedance parameter range. The calculated impedance parameter range is then subtracted from a preset impedance benchmark parameter to obtain the impedance difference. Regarding the setting of the preset impedance benchmark parameter, 20 independent repeated energizing tests were conducted in a control box of homogeneous sandy loam soil with a standard moisture content of 20%. The resistivity range data under steady-state conditions were collected, and the arithmetic mean was calculated. Finally, the preset impedance benchmark parameter was calibrated to be 45 ohms. For example, when the upper limit impedance parameter extracted from a certain sampling point is 120 ohms and the lower limit impedance parameter is 50 ohms, the calculated range of impedance parameter fluctuation is 70 ohms. This is subtracted from the preset impedance reference parameter of 45 ohms, resulting in an impedance difference of 25 ohms. A resistance change limit of 15 ohms is set. The number of events where the absolute value of the cumulative impedance difference is strictly greater than 15 ohms within 100 consecutive monitoring periods is used as the occurrence frequency. Based on the obtained occurrence frequency value, dynamic weight coefficients are assigned to pixel coordinates corresponding to the same geographical location in the canopy optical map. Specifically, the occurrence frequency value is directly multiplied by the basic attenuation constant 0.02 to obtain the weight attenuation amount, and then the initial weight of 1.0 is subtracted from this weight attenuation amount to obtain the final assigned weight. For example, if the occurrence frequency is 10 times, the weight attenuation amount is 0.2, and the final assigned weight is 0.8. The lower limit of the weight is set to 0.5. The final assigned weight of all associated pixel coordinate points is checked item by item. The cache of associated feature point data with a final assigned weight value strictly less than 0.5 is directly deleted and cleared, and high-quality points that meet the weight requirements are retained. The average impedance data corresponding to all remaining high-quality points are structured and encapsulated according to latitude and longitude coordinates to obtain the substrate dielectric property set.

[0022] Table 1 - Frequency and Weight Allocation of Sampling Point Electrical Impedance Monitoring: ; Table 1 shows the impedance data fluctuations and weight allocation details of the four discrete sampling points during the continuous monitoring period. Sampling point number 3 will be directly removed because its weight is less than 0.5.

[0023] S103: Call the substrate dielectric property set, read the built-in resistance feature value and optical pixel vector, concatenate the resistance feature value and optical pixel vector according to the coordinate index order to establish a feature matrix, calculate the covariance value between each column vector in the feature matrix, remove the column vectors with covariance values ​​less than the data correlation threshold, aggregate the remaining column vectors and perform dimensional format conversion to establish a multi-source agricultural condition characterization set. The system reads the 1D built-in resistance feature value sequence contained in the substrate dielectric property set and simultaneously calls the corresponding multi-dimensional optical pixel vector set in the same coordinate system of the canopy optical map. Following the ascending index order of latitude and longitude coordinates, the 1D built-in resistance feature value at each spatial coordinate point is horizontally concatenated with the 3D optical pixel vector to construct a 2D feature matrix with 4 columns and the number of rows equal to the total number of sampling points. For the established 2D feature matrix, arbitrary two column vector data are extracted. The Pearson correlation coefficient formula is applied, firstly calculating the deviation data sequence of each of the two column vectors relative to the mean; then multiplying the corresponding terms of the two sets of deviation data sequences one by one and summing them globally to obtain the covariance numerator; then calculating the sum of squares of the two sets of deviation data sequences and taking the square root to obtain the standard deviation product as the denominator; finally, dividing the covariance numerator by the standard deviation multiplied by the denominator to obtain the covariance correlation value between the two column vectors. A correlation threshold of 0.25 was set. The covariance correlation values ​​of all column vectors within the matrix were compared, and redundant column vectors with covariance correlation values ​​strictly less than 0.25 were identified and eliminated, retaining only the core data columns with strong correlations. The remaining column vectors underwent a dimensionality transformation, employing Z-score standardization logic. The arithmetic mean and standard deviation of all elements within each column vector were calculated. The arithmetic mean was subtracted from the value of each specific element in the vector to obtain the difference value. This difference value was then divided by the standard deviation, uniformly mapping values ​​with different physical dimensions to a dimensionless standard normal distribution interval with a mean of 0 and a standard deviation of 1. The standardized column vectors were then reassembled and merged along the column direction to establish a multi-source agricultural condition characterization set.

[0024] Please see Figure 3 The specific steps of S2 are as follows: S201: Based on the multi-source agricultural condition characterization set, near-infrared reflectance spectrum and red reflectance spectrum are extracted. The difference between the near-infrared spectrum pixel value and the red spectrum pixel value is subtracted from the red spectrum pixel value. The difference is divided by the sum of the two spectral pixel values ​​to obtain the vegetation feature numerical sequence. The arithmetic mean of the vegetation feature numerical sequence is calculated. The associated coordinate sites within the vegetation feature numerical sequence that are greater than the arithmetic mean are extracted. The coordinate sites and associated pixel values ​​are aggregated to generate a greenness feature matrix. Perform element-wise difference calculations, subtracting the corresponding red spectral pixel value from the near-infrared spectral pixel value at the current index to obtain the absolute difference data. Simultaneously, sum the near-infrared and red spectral pixel values ​​at the same index to obtain the total sum of spectral pixels. Use the calculated absolute difference data as the dividend and the total sum of spectral pixels as the divisor to perform a division operation, obtaining the vegetation feature value at a single coordinate point. Traverse all coordinate system nodes within the entire representation set, arranging all independently calculated vegetation feature values ​​in their original coordinate order to obtain a complete sequence of vegetation feature values. For example, if the extracted near-infrared spectral pixel value for the first coordinate is 0.8 and the red spectral pixel value is 0.2, subtracting 0.2 from 0.8 gives an absolute difference of 0.6. Adding 0.8 and 0.2 gives a total sum of 1.0. Dividing 0.6 by 1.0 gives the vegetation feature value of 0.6 for that coordinate. After obtaining the entire sequence, all values ​​within the sequence are summed, and the sum is divided by the total number of elements in the sequence to calculate the arithmetic mean of the vegetation feature numerical sequence. The extracted spatial coordinates of locations that meet the high greenness condition and their corresponding vegetation feature pixel values ​​are then aggregated in a structured manner to generate a greenness feature matrix.

[0025] S202: For the greenness feature matrix, a constant voltage value is applied to the monitoring probe, and the soil electrical impedance state sequence at the corresponding coordinate is extracted simultaneously. The constant voltage value is divided by the impedance value in the soil electrical impedance state sequence to obtain the node conductivity parameter. The median of the conductivity parameter is taken as the conductivity reference. The difference between the conductivity parameter and the conductivity reference is calculated, and the coordinates of the positions where the difference exceeds the mean of all differences are selected. The conductivity parameters are vector-stitched according to the coordinate order to establish the impedance state distribution map. A constant voltage of 12 volts is applied to the subsurface probe at each geospatial coordinate location recorded in the greenness feature matrix. During the steady-state maintenance phase of the applied voltage, soil electrical impedance state sequence data returned from each coordinate location is synchronously extracted via an analog-to-digital converter interface. Using the constant voltage of 12 volts as the numerator, each actual impedance value recorded in the soil electrical impedance state sequence is used as the denominator for division operations to obtain a sequence of nodal conductivity parameters characterizing the ion conduction capacity of that location. For example, if the impedance value extracted by the probe at a certain coordinate is 40 ohms, the constant voltage of 12 is divided by the impedance value of 40, and the nodal conductivity parameter at that location is calculated to be 0.3 Siemens. All the obtained nodal conductivity parameter sequences are sorted in ascending order from smallest to largest, and the value located at the very center of the sequence is extracted as the conductivity reference parameter. Each nodal conductivity parameter is subtracted from the extracted conductivity reference parameter in turn, and the deviation difference between the two is calculated. Subsequently, the absolute values ​​of all deviation differences are summed and divided by the total number of points to calculate the mean of all deviations. Each original deviation difference is compared with this mean, and the coordinates of anomalous conduction locations associated with deviations significantly greater than the mean are selected. These selected anomalous conduction location coordinates are then sorted according to the spatial physical order of the farmland from north to south, and the first and last ends of the corresponding node conductivity parameters are concatenated according to this coordinate order to establish an impedance state distribution map.

[0026] S203: Call the impedance state distribution map, extract the node conductivity parameters and location coordinates, match the vegetation feature values ​​under the corresponding coordinates in the greenness feature matrix, perform array splicing of the vegetation feature values ​​and node conductivity parameters according to coordinate index, construct a multi-dimensional feature vector group, calculate the Euclidean distance values ​​between each column vector in the multi-dimensional feature vector group, remove the associated discrete vector data whose Euclidean distance values ​​are greater than the median distance, aggregate the remaining vector sequence, and obtain the source plot feature quantity; Based on the extracted two-dimensional location coordinate index, vegetation feature values ​​recorded under the same coordinate system are retrieved and matched in the greenness feature matrix. The extracted single-dimensional vegetation feature values ​​and single-dimensional nodal electrical conductivity parameters are then horizontally arrayed and stitched together according to the same coordinate index, fusing the two isolated scalar data into a single feature vector of length 2. This process is repeated for all matching coordinate points to construct a multi-dimensional feature vector group containing a large amount of two-dimensional feature data. For all column vector elements within this multi-dimensional feature vector group, the spatial dispersion is calculated using the Euclidean distance operation logic described in the text. Specifically, two column vectors are arbitrarily selected, and the square of their difference in the first and second feature dimensions is calculated. These two squared values ​​are then summed, and the square root operation is performed to derive the Euclidean distance between the two column vectors. For example, the first vector data consists of vegetation feature 0.6 and electrical conductivity parameter 0.3, and the second vector data consists of vegetation feature 0.8 and electrical conductivity parameter 0.4. Calculating the square of 0.6 minus 0.8 gives 0.04, and calculating the square of 0.3 minus 0.4 gives 0.01. Adding 0.04 and 0.01 gives 0.05. Taking the square root of 0.05 gives the Euclidean distance value of 0.223. After calculating the Euclidean distance for all vector pairs, the median distance is extracted after sorting all distance values. Comparing the average Euclidean distance value of each vector with the median distance, discrete isolated vector data associated with Euclidean distance values ​​strictly greater than the median distance are directly removed. The remaining highly clustered vector sequence after removal is vertically aggregated and compressed along the row direction to obtain the source plot feature values.

[0027] Please see Figure 4 The specific steps of S3 are as follows: S301: Based on the source plot feature quantity, extract the vegetation growth status sequence, read the regional vegetation growth benchmark value, subtract the value of each element in the vegetation growth status sequence from the vegetation growth benchmark value to obtain the difference sequence, process the absolute value of each difference in the difference sequence, remove the coordinate points whose absolute value is less than the growth deviation limit value, extract the remaining coordinate points and the corresponding differences, and perform vector splicing to obtain the physiological deficiency assessment matrix. A one-dimensional vegetation growth status sequence recording the historical cumulative photosynthetic amount of crops was extracted, and a regionally set vegetation growth benchmark value directly issued by the regional agricultural management database was read. This benchmark value was established based on statistical analysis of the past three years' growth data from five high-yield demonstration farms at the same latitude, with the median of vegetation growth characteristics at the jointing stage set as a fixed parameter value of 0.75. The values ​​of each specific element within the vegetation growth status sequence were extracted one by one. Using the regionally set vegetation growth benchmark value of 0.75 as the minuend, each extracted element value was subtracted from the sequence to obtain a difference sequence reflecting the degree of growth lag. All positive and negative differences within this difference sequence were uniformly converted to absolute values ​​to eliminate the influence of sign direction, yielding the pure deviation amplitude. A growth deviation threshold of 0.10 was set. The values ​​after absolute value conversion were compared with this threshold, and spatial coordinates associated with absolute values ​​strictly less than 0.10 were directly eliminated, indicating that the growth in this area was normal and required no intervention. The remaining coordinate sites indicating significant growth retardation after screening, along with their corresponding absolute values ​​of growth deviations, are concatenated into a one-dimensional vector to obtain a key physiological deficit assessment matrix. For example, if the growth status sequence element extracted from a certain coordinate point has a value of 0.55, subtracting 0.55 from the baseline growth value of 0.75 yields a difference of 0.20, the absolute value of which is also 0.20. Since 0.20 is strictly greater than the set growth deviation threshold of 0.10, this coordinate and the deviation value of 0.20 are retained and concatenated into the physiological deficit assessment matrix.

[0028] S302: Call the physiological deficiency assessment matrix, extract coordinate sites, retrieve soil electrical conductivity characteristic parameters associated with coordinate sites, retrieve the pre-stored soil electrical conductivity benchmark threshold, calculate the difference between the soil electrical conductivity benchmark threshold and the soil electrical conductivity characteristic parameters, i.e., calculate the soil electrical conductivity compensation difference, compare the soil electrical conductivity compensation difference with the electrical conductivity fluctuation limit value, record the node index corresponding to the soil electrical conductivity compensation difference exceeding the electrical conductivity fluctuation limit value, extract the soil electrical conductivity compensation difference according to the node index and sort it in descending order to establish a fertility deficiency characterization vector; The soil conductivity compensation difference is calculated using the following formula: ; in, Representing the Soil conductivity compensation difference at each coordinate point The node number subscript identifier represents the extracted coordinate point. Representing the The device has a pre-stored soil conductivity reference threshold at each coordinate point. Representing the Each coordinate point is associated with soil electrical conductivity characteristic parameters. Represents the influence coefficient of soil pore structure. Represents the soil depth index; Extracted soil electrical conductivity characteristic parameters Influence coefficient on soil pore structure and soil depth index These three parameters are multiplied sequentially to obtain the dynamic conductivity attenuation value; subsequently, the pre-stored soil conductivity reference threshold value is retrieved. Subtracting the calculated dynamic conductivity attenuation value from the above calculation yields the preliminary conductivity deviation. Finally, this preliminary conductivity deviation is compared with the basic conductivity environmental compensation offset. By performing summation operations, the final soil conductivity compensation difference value can be derived. For example, the equipment has a pre-stored soil conductivity reference threshold. For a Siemens 0.8, at a certain coordinate point Soil electrical conductivity characteristic parameters The influence coefficient of soil pore structure is 0.5 Siemens. Set to 1.2, soil depth index The basic conductivity environmental compensation offset is 0.5. The value is 0.1 (Siemens). Substituting into the aforementioned logic, firstly, multiply 0.5, 1.2, and 0.5 sequentially to calculate the dynamic conductivity attenuation value of 0.3; then, subtract the dynamic conductivity attenuation value of 0.3 from the baseline threshold of 0.8 to obtain the preliminary conductivity deviation of 0.5; finally, add the preliminary deviation of 0.5 to the environmental compensation offset of 0.1 to calculate the final coordinate point. Soil conductivity compensation difference The value is 0.6 Siemens. The calculated soil conductivity compensation differences will be used... The difference in soil conductivity compensation is compared with the preset conductivity fluctuation limit value of 0.4 Siemens. When the difference is 0.6 Siemens, which is significantly greater than the conductivity fluctuation limit value of 0.4 Siemens, the corresponding node index number is recorded. Extract the node indexes of all records that exceed the standard, along with the corresponding soil conductivity compensation difference. Then, perform a position rearrangement operation based on the descending order of the difference values ​​to establish a fertility deficiency representation vector.

[0029] Table 2 - Calculation and Judgment of Nodal Conductivity Compensation: ; As shown in Table 2, the compensation difference obtained by substituting the three consecutive index nodes into the logical deduction is listed in detail. Since the difference between the nodes of index 12 and 13 is greater than 0.4, Siemens will be extracted and the vector will be sorted in descending order according to the size relationship between 0.6 and 0.48.

[0030] S303: For the fertility deficiency characterization vector, extract the differences of each item arranged in descending order, match the differences associated with the same coordinate positions in the physiological deficiency assessment matrix, perform a weighted summation operation on the two sets of differences according to the set weight coefficient ratio, obtain the node fusion feature value, compare the node fusion feature value with the nutrient compensation limit value, retain the coordinates associated with the node fusion feature value that has not reached the nutrient compensation limit value, and obtain the comprehensive nutrient deficiency feature. Based on the current node spatial coordinates, the absolute difference in growth deviation associated with the same coordinate position in the matching physiological deficiency assessment matrix is ​​retrieved. The obtained conductivity compensation difference and the absolute difference in growth deviation are then weighted and summed according to a set weighting coefficient. This weighting coefficient allocation scheme is based on principal component analysis experiments; the weighting coefficient assigned to the conductivity compensation difference is 0.6, and the weighting coefficient assigned to the absolute difference in growth deviation is 0.4. The calculation logic is as follows: multiply the conductivity compensation difference by 0.6 to obtain the first product, multiply the absolute difference in growth deviation by 0.4 to obtain the second product, and add the first product and the second product to derive the node fusion feature value. For example, if the extracted conductivity compensation difference for a certain node is 0.6 Siemens and the absolute difference in growth deviation is 0.20, multiplying 0.6 by 0.6 yields the first product 0.36, and multiplying 0.20 by 0.4 yields the second product 0.08. Adding 0.36 and 0.08 gives a node fusion feature value of 0.44. A nutrient compensation threshold of 0.60 is set, and the calculated node fusion feature values ​​are compared with this threshold. Spatial coordinates associated with node fusion feature values ​​strictly less than the 0.60 nutrient compensation threshold are extracted and retained, while oversaturated data exceeding this threshold are discarded, thus obtaining the comprehensive nutrient deficiency characteristics.

[0031] Please see Figure 5 The specific steps of S4 are as follows: S401: Extract the built-in regional location of comprehensive nutrient deficiency characteristics, monitor the real-time longitude and latitude coordinates corresponding to the regional location, collect the agricultural machinery movement speed parameter, retrieve the mechanical delay time value of the built-in fertilizer applicator, multiply the agricultural machinery movement speed parameter and the mechanical delay time value of the fertilizer applicator to obtain the product value sequence, extract the reference elements with values ​​greater than the displacement limit in the product value sequence, aggregate the elements and their corresponding longitude and latitude coordinates, and establish the lag displacement span vector; The system utilizes a vehicle-mounted dual-frequency positioning receiver to monitor real-time longitude and latitude coordinates for each location. It reads the agricultural machinery's movement speed parameter transmitted from the chassis via the underlying bus interface. It then retrieves the mechanical delay time value of the fertilizer applicator, stored internally in the control equipment. A direct multiplication operation is performed between the acquired agricultural machinery movement speed parameter and the fertilizer applicator mechanical delay time value to obtain a product sequence reflecting the physical lag distance. For example, if the acquired agricultural machinery movement speed parameter is 1.5 meters per second, and the retrieved internal fertilizer applicator mechanical delay time value is 2.0 seconds, multiplying 1.5 and 2.0 yields a product value of 3.0 meters for the lag distance at that instant. A displacement limit benchmark of 1.0 meter is pre-established. Elements representing excess distances within the product value sequence whose calculated values ​​are strictly greater than the 1.0-meter displacement limit benchmark are extracted to establish a lag displacement span vector.

[0032] S402: Read the built-in element values ​​and corresponding latitude coordinates of the hysteresis displacement span vector, retrieve the benchmark value of the Earth's meridian span length, divide the element values ​​by the benchmark value of the Earth's meridian span length to obtain the benchmark ratio parameter, perform array splicing on the benchmark ratio parameter according to the latitude coordinate sequence, calculate the deviation values ​​of each parameter in the array from the standard latitude interval, remove the discrete coordinate items whose deviation values ​​exceed the offset judgment threshold, aggregate the remaining item coordinates and corresponding deviation values ​​to obtain the latitude offset matrix; The built-in geographic mapping constant library is searched to obtain the baseline value for the Earth's meridian span length, which is fixed at 111,132 meters per degree in the test area. The extracted physical lag distance element value is used as the dividend, and the baseline value of the Earth's meridian span length (111,132) is used as the divisor to perform a division operation, calculating the baseline ratio parameter for latitude and longitude conversion. For example, if the extracted distance element value is 3.0 meters, dividing 3.0 by 111,132 yields a baseline ratio parameter of approximately 0.000027 degrees. Following the sorting sequence of the original latitude coordinate column, all the converted baseline ratio parameters are subjected to a one-dimensional array stitching operation. The array is traversed, and the difference between each baseline ratio parameter within the array and the preset standard latitude interval deviation value is calculated. An offset threshold of 0.000050 degrees is set. Extremely discrete coordinate items associated with an absolute difference value strictly greater than this offset threshold of 0.000050 degrees are eliminated to prevent extreme data interference caused by instantaneous satellite signal drift. The remaining reasonable coordinate items and their corresponding standard deviation values ​​are then aggregated and reorganized in memory to obtain a two-dimensional latitude offset matrix.

[0033] S403: Extract the coordinates and corresponding deviation values ​​of the latitude offset matrix, match the corresponding site feature parameters in the comprehensive nutrient deficiency features, sum the deviation values ​​and site feature parameters to obtain the node trigger weight parameter, compare the node trigger weight parameter with the preset trigger threshold, retain the spatial coordinates corresponding to the node trigger weight parameter being greater than the preset trigger threshold, arrange the spatial coordinates in ascending order along the agricultural machinery travel sequence, extract the front site coordinate data array, and obtain the fertilizer discharge touch front node; By cross-referencing spatial locations, comprehensive nutrient characteristic parameters of sites coinciding with the aforementioned coordinates are retrieved and extracted. The calculated baseline deviation value is then added directly to the comprehensive nutrient characteristic parameter to obtain the node triggering weight parameter used to control response priority. For example, if the extracted deviation value is 0.000027 and the matched characteristic parameter is 0.44, adding 0.000027 to 0.44 yields a node triggering weight parameter of approximately 0.440027. A preset triggering threshold is retrieved. This threshold is set by conducting spraying delay tests under different soil viscosities, taking the minimum critical value where the test response rate reaches above 95%, and is set to 0.30. All generated node triggering weight parameters are compared one by one with the preset triggering threshold of 0.30. Data below this threshold is discarded, and spatial coordinate data corresponding to node triggering weight parameters strictly greater than 0.30 are retained. According to the planned travel route sequence of the agricultural machinery chassis, these retained spatial coordinate points are sorted in ascending order of occurrence from earliest to latest. Starting from the head of the sequence, extract the coordinate data slices of the first 5 sequentially arranged sites to form an array, and lock these 5 nodes as the front nodes of the fertilizer discharge touch control.

[0034] Please see Figure 6 The specific steps of S5 are as follows: S501: Extract the coordinate sequence of the fertilizer discharge touch front node record, retrieve the comprehensive nutrient deficiency feature parameter of the corresponding position, call up the fertilizer conversion ratio parameter built into the device, multiply the comprehensive nutrient deficiency feature parameter with the fertilizer conversion ratio parameter to obtain the total fertilizer discharge mass value, subtract the mass lower limit benchmark from the total fertilizer discharge mass value to obtain the difference, remove the coordinate positions associated with the difference not reaching the judgment benchmark, extract the remaining positions and values ​​and perform array splicing to generate the fertilizer discharge mass matrix; The system retrieves the comprehensive nutrient deficiency characteristic parameters stored at the corresponding geographical location and calls up the fertilizer conversion ratio parameter built into the device's microcontroller. This parameter is defined by spreading a compound fertilizer sample within a standard 1-square-meter area, recording the actual number of grams of fertilizer required for vegetation to recover to the baseline, and setting the conversion ratio parameter to a fixed 120 grams per characteristic unit. The retrieved comprehensive nutrient deficiency characteristic parameter is multiplied by the fertilizer conversion ratio parameter to obtain the total fertilizer discharge mass requirement for guiding mechanical movement. For example, if the comprehensive nutrient deficiency characteristic parameter at a certain coordinate is 0.44 and the conversion ratio parameter is 120, multiplying 0.44 by 120 yields a total fertilizer discharge mass value of 52.8 grams at that point. The total fertilizer discharge mass value is then subtracted from the set lower limit benchmark of 10.0 grams to obtain the judgment difference. The coordinate points with minute fluctuations associated with the judgment difference being strictly less than 0 (i.e. the original value did not reach the lower limit judgment benchmark) are removed to avoid high-frequency ineffective start-stop wear of mechanical devices. The remaining spatial points that meet the quality distribution requirements and the corresponding total mass values ​​are extracted and a two-dimensional array splicing operation is performed to generate the fertilizer discharge quality demand matrix.

[0035] Table 3 - Total Fertilizer Discharge Mass Requirements and Threshold Filtering Table: ; As shown in Table 3, the mass values ​​obtained by substituting the first 3 preceding nodes into the multiplication calculation logic are displayed. Among them, node 2 was judged to be removed and no instruction was issued because its conversion mass of 6.0 grams did not meet the 10-gram benchmark.

[0036] S502: Call the fertilizer discharge mass matrix, extract the total mass value, retrieve the constant fertilizer mass value released by the fertilizer dispenser in a single release, perform a division operation between the total fertilizer discharge mass value and the constant fertilizer mass value released in a single release, obtain the node quotient sequence, extract each element in the node quotient sequence and perform a rounding operation to obtain the integer period parameter, compare the integer period parameter with the motor load limit parameter, retain the associated coordinates of the integer period parameter that do not exceed the motor load limit parameter, and establish the rotation number vector; The stepper motor fertilizer dispenser releases a constant fertilizer mass per rotation, calibrated to 15.0 grams per revolution based on on-site groove volume measurement. The extracted total fertilizer mass is used as the dividend, and the constant fertilizer mass released per revolution (15.0 grams) is used as the divisor. A division operation is performed to calculate a sequence of node quotients in floating-point format. For example, the total fertilizer mass extracted for node 1 is 52.8 grams; dividing it by 15.0 grams yields a node quotient of 3.52. Each floating-point element in the node quotient sequence is rounded down to the nearest integer, discarding the decimal part to obtain an integer period parameter recognizable by the hardware. For the quotient of 3.52, rounding results in an integer period parameter of 3 revolutions. The calculated integer period parameters are compared with a preset motor load limit parameter. This limit parameter, based on the motor's continuous operation overheat protection requirements, sets a maximum continuous output of 10 revolutions per cycle, retaining associated coordinate data where the integer period parameter is strictly less than or equal to 10 revolutions, thus constructing a rotation revolution vector.

[0037] S503: For the rotation number vector, extract the integer period parameter, collect the single-rotation pulse number parameter, multiply the integer period parameter and the single-rotation pulse number parameter to obtain the alternating pulse signal sequence, read the set duty cycle reference value, convert the waveform width of the alternating pulse signal sequence according to the duty cycle reference value, extract the converted pulse frequency and compare it with the hardware frequency upper limit value, remove the associated nodes whose frequency exceeds the upper limit value, and obtain the fertilizer application drive command; The driver's underlying configuration file is read to collect the number of pulses required for the drive motor to complete one revolution. This parameter is set to 200 pulses based on the motor's step angle of 1.8 degrees. A direct multiplication operation is performed between the integer period parameter and the number of pulses per revolution parameter to derive the total number of alternating pulse signals needed for controlling the port level transitions. For example, if the integer period parameter of a certain node is 3 revolutions, multiplying it by the number of pulses per revolution parameter of 200 yields a total of 600 alternating pulse signals required for this action. The duty cycle reference value of 0.5 set by the microcontroller is read, and the aforementioned number of alternating pulse signals undergoes a symmetrical high-low level waveform width conversion operation according to this duty cycle reference value, ensuring that the high-level duration and low-level duration are equal. The converted pulse output frequency value is extracted and directly compared with the hardware motherboard's specified frequency limit of 2000 Hz. Nodes with calculated output frequencies strictly exceeding the 2000 Hz limit are eliminated. The remaining pulse sequence, verified by all electrical parameters, is packaged and converted into underlying hexadecimal machine code to obtain the fertilizer application drive instruction.

[0038] Please see Figure 7 An agricultural fertilization system based on big data analysis includes: The agricultural monitoring module, based on cloud-based farmland nutrient distribution archives, captures near-infrared reflectance and red reflectance spectra using multispectral equipment to identify soil electrical impedance status and establish a multi-source agricultural characterization set. The nutrient deficiency assessment module uses a multi-source agricultural condition characterization set, subtracts the red light reflectance spectrum from the near-infrared reflectance spectrum and divides it by the total sum of the spectra, assesses the vegetation growth status, and obtains the source plot characteristic quantities. The fertilization derivation module, based on the source plot characteristics, sets a regional vegetation growth benchmark value, deducts the vegetation growth status, extracts the degree of physiological deficiency, deducts soil electrical conductivity characteristics from the electrical conductivity benchmark, extracts the degree of fertility deficiency, and establishes comprehensive nutrient deficiency characteristics. The fertilizer discharge target conversion module utilizes comprehensive nutrient deficiency characteristics to monitor real-time longitude and latitude coordinates and agricultural machinery movement speed. It multiplies the movement speed with the mechanical delay time of the fertilizer discharger to convert the lag displacement span and obtain the fertilizer discharge touch control front node. The motor drive control module uses a fertilizer discharge touch front node to calculate the comprehensive nutrient deficiency characteristics and multiply them by the fertilizer conversion ratio to convert the total fertilizer discharge mass, adjust the relationship between the total fertilizer discharge mass and the constant fertilizer mass released in a single cycle, and obtain the fertilizer application drive command.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.

Claims

1. An agricultural fertilization method based on big data analysis, characterized in that, Includes the following steps: S1: Based on cloud-based farmland nutrient distribution archives, near-infrared reflectance and red reflectance spectra are captured by multispectral devices to identify soil electrical impedance status and establish a multi-source agricultural condition characterization set; S2: Using the multi-source agricultural condition characterization set, the near-infrared reflectance spectrum is subtracted from the red reflectance spectrum and divided by the total sum of the spectra to assess the vegetation growth status, determine the relationship between the constant voltage applied by the probe and the soil electrical impedance state, and obtain the source plot characteristic quantities. S3: Based on the source plot characteristics, the regional vegetation growth benchmark value is set and the vegetation growth status is deducted to extract the degree of physiological deficiency. The electrical conductivity benchmark is deducted and the soil electrical conductivity characteristics are extracted to extract the degree of fertility deficiency. The degree of physiological deficiency and the degree of fertility deficiency are adjusted and weighted and summed according to the set weight coefficient ratio to establish a comprehensive nutrient deficiency characteristic. S4: Utilizing the comprehensive nutrient deficiency characteristics, monitor the real-time longitude and latitude coordinates and the agricultural machinery movement speed, multiply the movement speed by the mechanical delay time of the fertilizer applicator, calculate the lag displacement span, compare the lag displacement span with the Earth's meridian crossing length benchmark, derive the latitude offset, and obtain the fertilizer applicator touch control front node. S5: Using the aforementioned fertilizer discharge touch front node, the total mass of fertilizer discharge is calculated by multiplying the comprehensive nutrient deficiency characteristics with the fertilizer conversion ratio, adjusting the relationship between the total mass of fertilizer discharge and the constant fertilizer mass released in a single cycle, inferring the number of motor rotations, generating alternating pulse signals, and obtaining the fertilizer application drive command. The specific steps of S2 are as follows: S201: Based on the multi-source agricultural condition characterization set, extract near-infrared reflectance spectrum and red reflectance spectrum, subtract the difference between the near-infrared spectrum pixel value and the red spectrum pixel value, divide the difference by the sum of the two spectral pixel values, obtain the vegetation feature numerical sequence, calculate the arithmetic mean of the vegetation feature numerical sequence, extract the associated coordinate sites within the vegetation feature numerical sequence that are greater than the arithmetic mean, converge the coordinate sites and associated pixel values, and generate a greenness feature matrix. S202: For the greenness feature matrix, a constant voltage value is applied to the monitoring probe, and the soil electrical impedance state sequence at the corresponding coordinate is extracted simultaneously. The constant voltage value is divided by the impedance value in the soil electrical impedance state sequence to obtain the node conductivity parameter. The median of the conductivity parameter is taken as the conductivity reference. The difference between the conductivity parameter and the conductivity reference is calculated, and the coordinates of the positions where the difference exceeds the mean of all differences are selected. The conductivity parameters are vector-stitched according to the coordinate order to establish the impedance state distribution map. S203: Call the impedance state distribution map, extract the node conductivity parameters and location coordinates, match the vegetation feature values ​​under the corresponding coordinates in the greenness feature matrix, perform array splicing of the vegetation feature values ​​and node conductivity parameters according to coordinate index, construct a multi-dimensional feature vector group, calculate the Euclidean distance values ​​between each column vector in the multi-dimensional feature vector group, remove the associated discrete vector data whose Euclidean distance values ​​are greater than the median distance, aggregate the remaining vector sequence, and obtain the source plot feature quantity.

2. The agricultural fertilization method based on big data analysis according to claim 1, characterized in that, The multi-source agricultural characterization set includes a canopy optical image set, a substrate dielectric property library, and a topsoil habitat matrix. The source-end plot characteristics include a greenness index matrix, an ion ionization rate distribution map, and regional phenotypic parameters. The comprehensive nutrient deficiency characteristics include macro-element allocation extreme values, micro-element depletion indicators, and expected biochemical compensation amounts. The fertilizer application control pre-node includes a spatial advance trigger point, servo activation timing, and motion compensation margin. The fertilizer application drive commands include hardware control duty cycle, stepper transmission parameters, and valve opening and closing frequency.

3. The agricultural fertilization method based on big data analysis according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Based on cloud-based farmland nutrient distribution archives, near-infrared reflectance and red reflectance spectra are captured using multispectral equipment. The coordinate sequence embedded in the nutrient distribution archives is extracted, and the near-infrared reflectance and red reflectance spectra are registered and aligned according to the coordinate sequence. The pixel values ​​of the near-infrared reflectance spectra are subtracted from the pixel values ​​of the red reflectance spectra and divided by the total number of pixels. The spectral bands corresponding to the ratios greater than the set spectral limit values ​​are extracted to obtain the canopy optical map. S102: For the canopy optical spectrum, detect the soil electrical impedance state parameter associated with the sampling area, extract the impedance parameter value range fluctuation range, perform difference calculation between the impedance parameter value range fluctuation range and the preset impedance reference parameter, record the frequency of impedance difference exceeding the resistance change limit value, assign weights to the pixel coordinate points attached to the canopy optical spectrum according to the frequency of occurrence, remove the associated feature points whose weight values ​​do not reach the lower limit of the weight reference, and obtain the substrate dielectric property set; S103: Call the substrate dielectric property set, read the built-in resistance feature value and optical pixel vector, perform matrix concatenation of the resistance feature value and optical pixel vector according to the coordinate index order to establish a feature matrix, calculate the covariance value between each column vector in the feature matrix, remove the associated column vectors whose covariance value is less than the data correlation threshold, aggregate the remaining column vectors and perform dimensional format conversion to establish a multi-source agricultural condition characterization set.

4. The agricultural fertilization method based on big data analysis according to claim 3, characterized in that, The specific steps for S3 are as follows: S301: Based on the source plot feature quantity, extract the vegetation growth status sequence, read the regional vegetation growth benchmark value, subtract the value of each element in the vegetation growth status sequence from the vegetation growth benchmark value to obtain the difference sequence, process the absolute value of each difference in the difference sequence, remove the coordinate points whose absolute value is less than the growth deviation limit value, extract the remaining coordinate points and the corresponding differences and perform vector splicing to obtain the physiological deficiency assessment matrix. S302: Call the physiological deficiency assessment matrix, extract coordinate sites, retrieve soil electrical conductivity characteristic parameters associated with coordinate sites, retrieve the pre-stored soil electrical conductivity benchmark threshold, calculate the difference between the soil electrical conductivity benchmark threshold and the soil electrical conductivity characteristic parameters, i.e., calculate the soil electrical conductivity compensation difference, compare the difference with the electrical conductivity fluctuation limit value, record the node index corresponding to the difference exceeding the electrical conductivity fluctuation limit value, extract the difference according to the node index and sort it in descending order to establish a fertility deficiency characterization vector; S303: For the fertility deficiency characterization vector, extract the differences of each item arranged in descending order, match the differences associated with the same coordinate positions in the physiological deficiency assessment matrix, perform a weighted summation operation on the two sets of differences according to the set weight coefficient ratio, obtain the node fusion feature value, compare the node fusion feature value with the nutrient compensation limit value, retain the coordinates associated with the node fusion feature value that has not reached the nutrient compensation limit value, and obtain the comprehensive nutrient deficiency feature.

5. The agricultural fertilization method based on big data analysis according to claim 4, characterized in that, Extract the physiological deficiency weight coefficient and fertility deficiency weight coefficient included in the set weight coefficient ratio, and multiply the correlation difference of the same coordinate position in the physiological deficiency assessment matrix with the physiological deficiency weight coefficient to obtain the physiological feature weight product value. The differences of each item arranged in descending order are multiplied by the fertility deficiency weight coefficient to obtain the fertility feature weight product value. The physiological feature weight product value is added to the fertility feature weight product value to obtain the node fusion transition feature value.

6. The agricultural fertilization method based on big data analysis according to claim 5, characterized in that, The specific steps of S4 are as follows: S401: Extract the built-in regional locations of the comprehensive nutrient deficiency features, monitor the real-time longitude and latitude coordinates corresponding to the regional locations, collect the agricultural machinery movement speed parameters, retrieve the mechanical delay time value of the built-in fertilizer applicator, multiply the agricultural machinery movement speed parameters and the mechanical delay time value of the fertilizer applicator to obtain the product value sequence, extract the relevant elements in the product value sequence whose values ​​are greater than the displacement limit benchmark, aggregate the elements and their corresponding longitude and latitude coordinates, and establish the lag displacement span vector; S402: Read the value of the element and the corresponding latitude coordinate of the hysteresis displacement span vector, retrieve the reference value of the span length of the Earth's meridian, divide the element value by the reference value of the span length of the Earth's meridian to obtain the reference ratio parameter, perform array splicing on the reference ratio parameter according to the latitude coordinate sequence, calculate the deviation value of each parameter in the array from the standard latitude interval, remove the discrete coordinate items associated with deviation values ​​exceeding the offset judgment threshold, aggregate the coordinates of the remaining items and the corresponding deviation values ​​to obtain the latitude offset matrix; S403: Extract the coordinates and corresponding deviation values ​​of the latitude offset matrix, match the corresponding site feature parameters in the comprehensive nutrient deficiency features, sum the deviation values ​​and site feature parameters to obtain the node trigger weight parameter, compare the node trigger weight parameter with the preset trigger threshold, retain the spatial coordinates corresponding to the node trigger weight parameter being greater than the preset trigger threshold, arrange the spatial coordinates in ascending order along the agricultural machinery travel sequence, extract the front site coordinate data array, and obtain the fertilizer discharge touch front node.

7. The agricultural fertilization method based on big data analysis according to claim 6, characterized in that, Extract the preset feature scaling factor and preset offset weighting factor, multiply the site feature parameter with the preset feature scaling factor, and obtain the value of the mapped feature parameter. The deviation value is multiplied by the preset offset weighting coefficient to obtain the weighted offset feature value. The mapping feature parameter value is then summed with the weighted offset feature value to obtain the trigger evaluation benchmark parameter.

8. The agricultural fertilization method based on big data analysis according to claim 6, characterized in that, The specific steps of S5 are as follows: S501: Extract the coordinate sequence of the fertilizer discharge touch front node, retrieve the comprehensive nutrient deficiency feature parameter of the corresponding position, retrieve the fertilizer conversion ratio parameter built into the device, multiply the comprehensive nutrient deficiency feature parameter with the fertilizer conversion ratio parameter to obtain the total fertilizer discharge mass value, subtract the mass lower limit benchmark from the total fertilizer discharge mass value to obtain the difference, remove the coordinate positions associated with the difference not reaching the judgment benchmark, extract the remaining positions and values ​​and perform array splicing to generate the fertilizer discharge mass matrix; S502: Call the fertilizer discharge mass matrix, extract the total mass value, retrieve the constant fertilizer mass value released by the fertilizer discharger in a single release, perform a division operation between the total fertilizer discharge mass value and the constant fertilizer mass value released in a single release, obtain the node quotient sequence, extract each element in the node quotient sequence and perform a rounding operation to obtain the integer period parameter, compare the integer period parameter with the motor load limit parameter, retain the associated coordinates where the integer period parameter does not exceed the motor load limit parameter, and establish the rotation number vector; S503: For the rotation number vector, extract the integer period parameter, collect the single-cycle pulse number parameter, multiply the integer period parameter and the single-cycle pulse number parameter to obtain the alternating pulse signal sequence, read the set duty cycle reference value, convert the waveform width of the alternating pulse signal sequence according to the duty cycle reference value, extract the converted pulse frequency and compare it with the upper limit value of the hardware frequency, remove the associated nodes whose frequency exceeds the upper limit value, and obtain the fertilization drive command.

9. An agricultural fertilization system based on big data analysis, characterized in that, The system is used to implement the agricultural fertilization method based on big data analysis as described in any one of claims 1-8, and the system comprises: The agricultural monitoring module, based on cloud-based farmland nutrient distribution archives, captures near-infrared reflectance and red reflectance spectra using multispectral equipment to identify soil electrical impedance status and establish a multi-source agricultural characterization set. The nutrient deficiency assessment module uses the multi-source agricultural condition characterization set, subtracts the red light reflectance spectrum from the near-infrared reflectance spectrum and divides it by the total sum of the spectra, assesses the vegetation growth status, determines the relationship between the constant voltage applied by the probe and the soil electrical impedance state, and obtains the characteristic quantities of the source plot. The fertilization derivation module, based on the source plot characteristics, sets a regional vegetation growth benchmark value, deducts the vegetation growth status, extracts the degree of physiological deficiency, deducts soil electrical conductivity characteristics from the electrical conductivity benchmark, extracts the degree of fertility deficiency, and establishes comprehensive nutrient deficiency characteristics. The fertilizer discharge target conversion module uses the comprehensive nutrient deficiency characteristics to monitor real-time longitude and latitude coordinates and agricultural machinery movement speed. It multiplies the movement speed by the mechanical delay time of the fertilizer discharge device to calculate the lag displacement span. It compares the lag displacement span with the Earth's meridian crossing length benchmark to derive the latitude offset and obtain the fertilizer discharge touch control front node. The motor drive control module uses the aforementioned fertilizer discharge touch front node to calculate the total mass of fertilizer discharge by multiplying the comprehensive nutrient deficiency characteristics with the fertilizer conversion ratio, adjusting the relationship between the total mass of fertilizer discharge and the constant mass of fertilizer released per cycle, and obtaining the fertilizer application drive command.

Images