A Stepping Time Domain Reflectometry System and Method for In-situ Rapid Measurement of Available Soil Nutrients

Abstract

A step-by-step time-domain reflection soil available nutrient in-situ rapid measurement system and method, belonging to the field of soil detection technology, the system includes an in-situ multi-layer probe device, an information acquisition control box, and a client that are interconnected, the in-situ multi-layer probe device is connected to the information acquisition control box, the information acquisition control box includes an interconnected signal transceiver processing module, a data acquisition control module, a 4G / 5G transmission module, and a power supply module, and the information acquisition control box is connected to the client operating system via the Internet. The present invention installs an in-situ multi-layer probe device to monitor the difference in the real and imaginary parts of the soil dielectric constant during crop growth, and combines principal component analysis with a neural network algorithm to model, and regularly, in-situ, and one-time joint determination of multi-layer soil related parameters.

Description

Technical Field

[0001] The present invention belongs to the technical field of soil detection, and particularly relates to a step-by-step time-domain reflectometry in-situ rapid measurement system and method for available soil nutrients. Background Art

[0002] Accurate judgment of soil nutrient status is the basis for land managers to carry out operations such as crop fertilization management and soil improvement. Currently, there are mainly three methods for detecting soil nutrients: First, field soil samples are collected, air-dried, and then tested in the laboratory. The results can only represent the nutrient content of the soil at the time of sampling. Moreover, due to the long processes of sampling, extraction, and analysis and measurement, the immediate changes in soil nutrient content cannot be obtained in a timely manner, resulting in a serious lag in the effectiveness of data application. Finally, empirical data has to be used to complete the main production management process, and the nutrient test data is only used as reference data or a unified judgment standard. Second, remote sensing monitoring is used to extract spectral characteristic indicators highly correlated with soil nutrients from soil spectra and crop canopy spectra to invert soil nutrients and establish an inversion model for analysis and calculation. Its advantage is that it can obtain information in real time and quickly. Its disadvantages are as follows: First, the soil environment is complex, and different results may be obtained even with the same input and the same inversion model. Second, it is difficult to find a general soil nutrient monitoring and inversion model applicable to different regions, different soil types, different soil textures, and different farming measures. Third, the result accuracy is limited. The reflection spectrum of water even exceeds the influence on the soil reflection spectrum, further increasing the practical limitations of this method. Finally, the measurement depth is also concentrated within 0 - 20 cm, in the shallow soil layer, while crop roots can extend into the ground 2 meters deep depending on the variety. Third, soil sensors based on electromagnetic principles are used, and capacitive in-situ soil sensors are used for measurement. Its advantage is that real-time measurement is achieved. Its disadvantages are as follows: First, most sensors are needle-type, and soil profiles need to be excavated. This invasive installation method destroys the in-situ soil structure and crop roots, resulting in greater influence of measurement results by spatial heterogeneity and a decrease in result accuracy. Second, it is developed based on capacitance, frequency domain, or single-frequency domain time domain methods, with low precision, low resolution, incomplete analysis of frequency domain information, and poor accuracy of measurement results, and is not generally recognized in the industry. Third, the texture of the soil, genetic classification, specific regional planting systems, different climate conditions, etc. will all affect the measurement results, and a large number of calibration tests are required for sensor calibration before it can be used. Finally, the measurement indicators are few, only including moisture, salinity, macronutrient indicators, or single-element indicators.

[0003] The soil body is a three-phase system composed of soil particles, pore liquid, and air. The main influencing factors for the real-time measurement of available soil nutrients are porosity (n), water content (w), ionic composition of pore water, and pore water conductivity (EC w)(and temperature (T)). Because there are many influencing factors, there is no reported theory or method for in-situ dynamic multi-layer soil monitoring of soil available nutrients that can simultaneously consider the above factors and cause no damage to the soil structure. Summary of the Invention

[0004] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a stepped time domain reflectometry in-situ rapid measurement system and method for soil available nutrients. By in-situ installing and arranging a time domain reflectometer and a temperature detection integrated probe, the differences in the real and imaginary parts of the soil dielectric constant during the growth process of crops are monitored. Through the combination of principal component analysis and neural network algorithm for modeling, the relevant parameters of multi-layer soil are jointly measured in a timed, in-situ, and one-time manner, thus fundamentally solving the problem of continuous monitoring of root soil data required for crop whole growth period nutrition management and soil quality management.

[0005] To achieve the above object, the main technical solutions adopted by the present invention include:

[0006] A stepped time domain reflectometry in-situ rapid measurement system for soil available nutrients, the system includes an in-situ multi-layer probe device, an information acquisition and control box, and a client that are connected to each other. The in-situ multi-layer probe device is connected to the information acquisition and control box. The information acquisition and control box includes a signal transceiver and processing module, a data acquisition and control module, a 4G / 5G transmission module, and a power supply module that are connected to each other. The information acquisition and control box is connected to the client operating system through the Internet; the in-situ multi-layer probe device includes an outer sleeve and an inner sleeve that are sleeved with each other. The bottom of the outer sleeve is a conical structure. A measurement hole is evenly and equidistantly opened from top to bottom on one side wall of the outer sleeve. Transverse installation grooves are evenly and equidistantly opened in the inner sleeve from top to bottom. The opening position of the installation groove matches the position of the measurement hole. A time domain reflectometer and a temperature detection integrated probe are arranged in the installation groove. The time domain reflectometer and the temperature detection integrated probe are connected to one end of a magnetic spring, and the other end of the magnetic spring is connected to the inner side wall of the installation groove. The magnetic spring is connected to a power supply. The time domain reflectometer and the temperature detection integrated probe are connected to the information acquisition and control box.

[0007] Further, the time domain reflectometer and the temperature detection integrated probe include a TDR sensor probe and a patch type soil temperature sensor. The TDR sensor probe and the patch type soil temperature sensor are cast into an integral structure by epoxy resin.

[0008] Further, a rubber pad covers the outside of the test hole of the outer sleeve.

[0009] Further, the time domain reflectometer has 1024 sets of stepped frequency signals in the frequency range of 1M - 1GHz.

[0010] A rapid measurement method for in-situ available nutrients in soil using a step-by-step time-domain reflectometry device, comprising the following steps:

[0011] Step S01: Place the in-situ multi-layer probe device in the soil, energize the magnetic spring, and extend the time-domain reflectometer and the temperature detection integrated probe into the soil;

[0012] Step S02: The data acquisition and control module of the information acquisition control box receives instructions from the client operating system. The data acquisition and control module simultaneously sends commands to the signal transceiver and processing module and the temperature sensor. The signal transceiver and processing module sequentially generates 1024 groups of point-frequency continuous wave signals step by step. The stepped frequency signals are transmitted into the soil through the in-situ multi-layer probe device. When encountering soils in different states, signal reflections occur. Then, through directional coupling, the test signal is separated from the reflected response signal to form a reflected received signal. When a 1024-point frequency scan is completed, the obtained series of incident representative signal received data, reflected received signal received data, and voltage signals collected by the temperature sensor are transmitted to the intelligent client operating system through the data acquisition and control module and the transmission module;

[0013] Step S03: The operating system represents the 1024 groups of stepped frequencies from high to low of the TDR transmission frequency as f H 、f L , then the high-low frequency difference is the frequency bandwidth . According to different bandwidths, the center frequencies f C of different bandwidths are obtained. Then, the center frequency f C〡B〡 of the frequency bandwidth B can be expressed as:

[0014]

[0015] The frequency-domain signal is converted into the center time domain of different bandwidths through inverse Fourier transform ;

[0016] Step S04: Measure the soil volumetric water content;

[0017] Step S05: Measure the soil conductivity;

[0018] Step S06: Measure the available nutrient data through principal component analysis and neural network model building.

[0019] Furthermore, in the step S02, the sweep frequency range of the 1024 groups of point-frequency continuous wave signals sequentially generated step by step by the signal transceiver and processing module is within the frequency range of 1M - 1GHz.

[0020] Furthermore, in the step S02, the incident representative signal and the reflected received signal are respectively received by their respective receivers, and the received data is stored.

[0021] Further, in the step S04, the measurement of the soil volumetric water content is specifically as follows: the central time domain in the measurement of the soil volumetric water content , the low-high frequency difference thereof is 30 MHz to 1 GHz. Through the Topp empirical formula, based on the measurement of the apparent dielectric constant, the soil volumetric water content is calculated according to the following formula ,

[0022] .

[0023] In the formula, a 1 , a 2 , a 3 are polynomial coefficients, and b is an adjustment constant.

[0024] Further, in the step S05, the measurement of the soil conductivity is specifically as follows: the central time domain in the measurement of the soil conductivity , the low-high frequency difference thereof is 1 MHz to 1 GHz;

[0025] Obtain the apparent dielectric constant. The speed of the electromagnetic wave propagating along the cable is inversely proportional to the square root of the apparent dielectric constant of the surrounding medium. The apparent dielectric constant is converted into the formula:

[0026]

[0027] where K a is the apparent dielectric constant of the medium, v is the speed of the electromagnetic wave propagating in the medium; c is the speed of the electromagnetic wave propagating in a vacuum, with the unit of centimeter / nanosecond; 2L is the distance that the electromagnetic wave propagates back and forth on the probe, with the unit of centimeter; d(t) 2 -d(t) 1 is the propagation time, with the unit of nanosecond;

[0028] Obtain the soil conductivity according to the following formula

[0029]

[0030] In the formula: K a is the apparent dielectric constant of the medium, V 1 is the voltage at the starting end of the soil of the central time domain probe, and V 2 is the voltage at the end of the soil of the central time domain probe.

[0031] Further, the step S06 includes the following steps:

[0032] (1) Input the central time domain data groups and temperature data of each time and soil layer monitored by the soil TDR sensor and the temperature sensor during the 3414 experiment;

[0033] (2)Input the data of available nutrient content measured by chemical methods of soil samples;

[0034] (3)Central time domain preprocessing: Preprocess the central frequency by using methods such as derivation and / or Savitzky-Golay smoothing and / or multiplicative scatter correction (MSC) and / or standard normal variate (SNV); Use robust principal component analysis to remove abnormal samples of soil samples;

[0035] (4)Characteristic central time domain extraction: Select characteristic bands by using successive projections algorithm (SPA) and / or uninformative variable elimination algorithm (UVE) and / or genetic algorithm (GA) methods, extract the characteristic central time domain for model establishment; Use robust principal component analysis to remove abnormal samples of soil samples;

[0036] (5)Multivariate calibration: Use Kennard-Stone algorithm to divide the calibration set and test set for establishing the soil nutrient content model, and use partial least squares method (PLS) to establish the soil nutrient content model; Use robust principal component analysis to remove abnormal samples of soil samples;

[0037] (6)Accuracy judgment: Perform accuracy discrimination calculation on the soil available nutrient model after central time domain interpretation to determine the coefficient of determination r 2 and relative analysis error RPD value, and judge r 2 and RPD value. If it meets the experimental allowable range, proceed to the next step. If not, return to step 1 to start over, and calculate the coefficient of determination r 2 and relative analysis error RPD value again, and select the best model;

[0038] (7)Complete the modeling;

[0039] (8)Prepare for prediction: Input the stepped frequency data of unknown soil samples;

[0040] (9)Prediction: Predict the available nutrient content value through the established model of the characteristic central frequency of soil samples and their available nutrient content and the stepped frequency data of unknown soil samples;

[0041] (10)Complete the prediction.

[0042] The beneficial effects of the present invention are as follows: A stepped time domain reflectometry soil available nutrient in-situ rapid measurement system and method provided by the present invention, by in-situ installing and arranging an in-situ multi-layer probe device with a time domain reflectometry (TDR) and temperature integrated sensor, monitors the differences in the real and imaginary parts of the soil dielectric constant during the growth process of crops. By combining principal component analysis and neural network algorithm for modeling, it jointly measures multi-layer soil related parameters in a timed, in-situ and one-time manner, thus fundamentally solving the needs of crop whole growth period nutrient management and soil quality management, and realizing continuous monitoring of root zone soil data. Description of the Drawings

[0043] Figure 1 Schematic structural diagram of the step - by - step time - domain reflectometry in - situ rapid measurement system for available soil nutrients of the present invention;

[0044] Figure 2 Schematic diagram of the test principle of the present invention;

[0045] Figure 3 Schematic structural diagram of the in - situ multi - layer probe device;

[0046] Figure 4 Schematic structural diagram when the integrated probe of the time - domain reflectometer and temperature detector does not extend;

[0047] Figure 5 Schematic structural diagram when the integrated probe of the time - domain reflectometer and temperature detector extends;;

[0048] Figure 6 Schematic structural diagram of the integrated probe of the time - domain reflectometer and temperature detector;

[0049] Figure 7 Fitting result of the calibration set of available phosphorus in soil in Example 1;

[0050] Figure 8 Fitting result of the test set of available phosphorus in soil in Example 1;

[0051] Figure 9 Migration curve of available phosphorus in soil during the whole growth period of corn in Example 1.

[0052] Components in the figure: 1 is the outer sleeve, 2 is the inner sleeve, 3 is the TDR sensor probe, 4 is the patch - type soil temperature sensor, 5 is the magnetic spring, 6 is the measurement hole, 7 is the installation groove, 8 is the cone, 9 is the power supply plug of the magnetic spring, 10 is the sensor coaxial cable and data line. Detailed Embodiment

[0053] In order to better explain the present invention for easy understanding, the present invention will be described in detail below with reference to the drawings through specific embodiments.

[0054] As Figure 1-2 shown, the present invention is a step - by - step time - domain reflectometry in - situ rapid measurement system for available soil nutrients. The system includes an in - situ multi - layer probe device, an information acquisition and control box, and a client, which are connected to each other. The in - situ multi - layer probe device is connected to the information acquisition and control box. The information acquisition and control box includes a signal transceiver and processing module, a data acquisition and control module, a 4G / 5G transmission module, and a power supply module, which are connected to each other. The information acquisition and control box is connected to the client operating system through the Internet; As Figure 3-6As shown, the in-situ multi-layer probe device includes an outer sleeve 1 and an inner sleeve 2 which are connected to each other. The bottom of the outer sleeve is a cone 8. A measuring hole 6 is evenly and equidistantly opened on one side wall of the outer sleeve from top to bottom. The inner sleeve is evenly and equidistantly opened with a transverse mounting groove 7 from top to bottom. The opening position of the mounting groove 7 matches the position of the measuring hole 6. A time domain reflectometer and temperature detection integrated probe is arranged in the mounting groove. The time domain reflectometer and temperature detection integrated probe is connected to one end of a magnetic spring 5. The other end of the magnetic spring 5 is connected to the inner side wall of the mounting groove 7. The time domain reflectometer and temperature detection integrated probe is connected to a magnetic spring power supply plug 9 reserved on the ground through a cable. A handheld power supply can be used to provide power supply. The time domain reflectometer and temperature detection integrated probe includes a TDR sensor probe 3 and a patch type soil temperature sensor 4. The TDR sensor probe 3 and the patch type soil temperature sensor 4 are cast into an integrated structure through epoxy resin. The test hole 6 of the outer sleeve is covered with a rubber pad. When the magnetic spring is extended, it drives the TDR sensor probe of the time domain reflectometer and temperature detection integrated probe and the patch-type soil temperature sensor to move toward the soil, push open the rubber pad of the reserved hole of the outer tube, and push it into the soil. When the test is completed or the probe needs to be replaced, the magnetic spring is retracted to drive the time domain reflectometer and temperature detection integrated probe to retract, and the rubber pad can prevent soil from entering the installation groove. The time domain reflectometer and temperature detection integrated probe are connected to the information acquisition control box through the sensor coaxial cable and data cable 10. The time domain reflectometer has 1024 sets of step frequency signals in the frequency range of 10M-1GHz.

[0055] Step S01: placing the in-situ multi-layer probe device in the soil, energizing the magnetic spring, and extending the time domain reflectometer and temperature detection integrated probe toward the soil;

[0056] Use an electric direct pressure original soil drilling drill adapted to the tubular soil sensor to drill a hole with an outer diameter and depth consistent with the outer tube size of the multi-layer probe in-situ installation device in the plot to be measured, and install the outer tube of the multi-layer probe in-situ installation device with the cone part facing downward in the soil hole; the needle-type TDR and temperature integrated sensor adopts a split installation device, which consists of an outer tube and an inner tube. Several measuring holes (one every 10 cm) are reserved on the outer tube, and rubber pads are installed outside the measuring holes. The inner tube includes a sensor and a magnetic spring sensor propulsion system. One end of the magnetic spring sensor propulsion system is fixed to the wall of the inner tube, and the other end is fixed to the needle-type TDR and temperature integrated probe, and the other end of the probe is aligned with the reserved hole in the inner tube. Then place the inner tube in the outer tube, with the reserved holes of the inner tube and the outer tube in the same position. Use a handheld power supply to power the magnetic spring through the reserved cable port on the tube cap. The magnetic spring is in an extended state, and the probe is pushed forward to push open the rubber pad of the reserved hole of the outer tube and push it into the soil. If the probe needs to be replaced or removed, reverse power is supplied to the magnetic spring, and the magnetic spring is in a contracted state. The probe is withdrawn from the soil into the inner tube, the inner tube is taken out, and the probe is replaced.

[0057] S02: The data acquisition and control module of the information acquisition control box uses the received instructions from the intelligent client operating system. The data acquisition and control module simultaneously sends the commands to the signal transceiver and processing module and the temperature sensor. The signal transceiver and processing module sequentially generates 1024 groups of point frequency continuous wave signals step by step, with the frequency sweep range within the range of 1M - 1GHz. The stepped frequency signals are transmitted into the soil. Each single frequency signal generates a representative signal of the incident signal through the coupling generation mechanism. The direct signal is transmitted along the coaxial cable to the probe at the end, and when encountering different states of the soil, signal reflection occurs. Then, through directional coupling, the test signal is separated from the reflected response signal to form a reflected received signal. The incident representative signal and the reflected received signal are respectively received by their respective receivers, and the received data is stored. When a 1024 - point frequency scan is completed, the obtained series of incident representative signal received data, reflected received signal received data, and the voltage signal collected by the temperature sensor are transmitted to the intelligent client operating system through the data acquisition and control module and the transmission module.

[0058] S03: The operating system represents the 1024 groups of stepped frequencies from high to low of the TDR emission frequency as f H 、f L , then the high - low frequency difference is the frequency bandwidth , and different center frequencies f C of different bandwidths are obtained according to different bandwidths.

[0059]

[0060] Through the inverse Fourier transform, the frequency - domain signal is converted into the center time - domain of different bandwidths .

[0061] S04: Measure the soil volumetric water content;

[0062] The center time - domain in the measurement of soil volumetric water content , and its low - high frequency difference is from 30MHz to 1GHz.

[0063] Through the TOPP empirical formula, based on the measurement of the apparent dielectric constant, the soil volumetric water content is calculated ,

[0064]

[0065] S05: Measure the soil conductivity;

[0066] The center time - domain in the measurement of soil conductivity , and its low - high frequency difference is from 1MHz to 1GHz.

[0067] First, obtain the apparent dielectric constant. The velocity of electromagnetic wave propagation along the cable is inversely proportional to the square root of the apparent dielectric constant of the surrounding medium. The apparent dielectric constant is then converted into the formula:

[0068]

[0069] Among them, K a is the apparent dielectric constant of the medium, v is the velocity of electromagnetic wave propagation in the medium; c is the velocity of electromagnetic wave propagation in vacuum, with the unit of centimeter / nanosecond; 2L is the distance that the electromagnetic wave travels back and forth on the probe, with the unit of centimeter; d(t) 2 -d(t) 1 is the propagation time, with the unit of nanosecond.

[0070] Obtain the soil conductivity according to the following formula

[0071]

[0072] In the formula: K a is the apparent dielectric constant of the medium, V 1 is the voltage at the starting end of the soil of the central time domain probe, V 2 is the voltage at the ending end of the soil of the central time domain probe;

[0073] S06: Through principal component analysis and neural network model building, measure the available nutrient data;

[0074] The steps are as follows:

[0075] (1) Input the central time domain data groups and temperature data of each time and soil layer monitored by the soil TDR sensor and temperature sensor during the 3414 experiment;

[0076] (2) Input the available nutrient content data measured by chemical methods of soil samples;

[0077] (3) Central time domain preprocessing: Preprocess the central frequency by using methods such as derivation and / or Savitzky-Golay smoothing and / or multiplicative scatter correction (MSC) and / or standard normal variate (SNV); Use robust principal component analysis to remove abnormal samples of soil samples;

[0078] (4) Characteristic central time domain extraction: Use the successive projections algorithm (SPA) and / or uninformative variable elimination algorithm (UVE) and / or genetic algorithm (GA) methods to select characteristic bands and extract the characteristic central time domain. Facilitate the establishment of a more accurate model; Use robust principal component analysis to remove abnormal samples of soil samples;

[0079] (5) Multivariate calibration: The Kennard-Stone algorithm is used to divide the calibration set and the test set for establishing the soil nutrient content model, and the partial least squares method (PLS) is used to establish the soil nutrient content model; the robust principal component analysis method is used to eliminate the abnormal samples of the soil samples;

[0080] (6) Precision judgment: Perform precision discrimination calculation on the soil available nutrient model after central time domain interpretation to determine the coefficient of determination r 2 and the relative analysis error RPD value, and judge r 2 and the RPD value. If it meets the experimental allowable range, proceed to the next step. If not, return to step 1 and start over, and then calculate the coefficient of determination r 2 and the relative analysis error RPD value, and select the best model;

[0081] (7) Complete the modeling;

[0082] (8) Prepare for prediction: Input the stepped frequency data of the unknown soil sample;

[0083] (9) Prediction: Predict the available nutrient content value of the unknown soil sample through the established model of the characteristic central frequency of the soil sample and its available nutrient content and the stepped frequency data of the unknown soil sample;

[0084] (10) Complete the prediction.

[0085] Example 1

[0086] Select a brown soil plot in Shenyang City, Liaoning Province, and plant corn. Set 14 treatments according to the soil testing and formulated fertilization 3414 experiment (3414 refers to 3 factors of nitrogen, phosphorus, and potassium, 4 levels, and 14 treatments, which is a field test scheme recommended for soil testing and formulated fertilization), and have 3 replicates for each treatment, and conduct the cultivation of the plot to be tested:

[0087]

[0088] "3414" refers to 3 factors of nitrogen, phosphorus, and potassium, 4 levels, and 14 treatments. The meaning of the 4 levels: Level 0 means no fertilization; Level 2 means the local optimal fertilization rate; Level 1 = Level 2 × 0.5; Level 3 = Level 2 × 1.5 (this level is the over-fertilization level).

[0089] After the spring thawing and before planting, a stepped-frequency continuous wave time domain reflectometry (Stepped-Frequency Continuous Wave SFCW-TDR) and temperature integrated sensor probe is deployed at the representative central position of the experimental field through an in-situ multi-layer probe device, and the soil temperature and the frequency domain and time domain information of each frequency band output per hour are set to be measured;

[0090] As the 3414 experiment progresses, before and after each irrigation and fertilization or other important agricultural operations (no less than 10 times throughout the year), soil samples are collected at positions more than 1 meter away from the sensor layout positions but with similar representativeness. The sampling depths are 0 - 10 cm, 10 - 20 cm, 20 - 30 cm, and 30 - 40 cm. The samples are sent to the laboratory for measuring soil available nutrients using extractants to obtain soil available nutrient attribute data. The measurement indicators and methods are shown in the following table:

[0091]

[0092] Convert the original experimental data into csv format and add it to the modeling module.

[0093] Automatically run the TDR time domain interpretation process within a specific temperature range.

[0094] Taking soil available phosphorus as an example, in its time domain interpretation process, the genetic algorithm is used to extract the characteristic center time domain, and the partial least squares method is used to establish a quantitative model for the available phosphorus of all soil samples with a bandwidth of every 1 - 10 frequency points in the full center time domain.

[0095] First, remove the abnormal samples from the soil samples, and eliminate the samples with significantly abnormal available phosphorus contents. Finally, 1068 soil samples are obtained. Then, the genetic algorithm is used to repeatedly iterate 5 times for the full center time domain of all soil samples, and the determination coefficient r 2 is calculated each time after extracting the characteristic center time domain, and is compared with the determination coefficient r 2 calculated from the full center time domain. See the following table:

[0096]

[0097] After using the genetic algorithm to extract the characteristic center time domain from the full center time domain of soil samples, the obtained determination coefficient r 2 is closer to 1 than the full center time domain, with a better modeling effect, and the correlation coefficient of iteration number 2 is the highest. Therefore, the characteristic time domain selected by iteration number 2 is chosen as the modeling variable.

[0098] Finally, before using the partial least squares method to establish the model, the SPXY algorithm is used to classify the soil samples. According to the ratio of 2:1, 712 calibration set soil samples and 356 validation set soil samples are selected and numbered accordingly.

[0099] Use the partial least squares method to establish a calibration model between the characteristic time domain of the soil sample calibration set and its available phosphorus content, and obtain its determination coefficient r 2 The calculation formula of the determination coefficient r 2 is as follows:

[0100]

[0101] where n is the number of samples in the calibration set (or test set), and y i is the actual value of the i-th sample in the calibration set (or test set), is the predicted value of the i-th sample in the calibration set (or test set), is the average value of all actual values in the calibration set (or test set), and the r 2 value is closer to 1, the better the model performance.

[0102] Figure 7 is the fitting result of the soil available phosphorus calibration set in this implementation, and its determination coefficient r 2 is 0.9097. The determination coefficient r 2 is relatively high, and the modeling effect is good.

[0103] Predicting the total phosphorus content of unknown soil samples

[0104] Estimate the available phosphorus content of the test set according to the established calibration model for soil available phosphorus content, obtain the estimated value, and then compare it with the measured actual value of the available phosphorus content, and calculate the determination coefficient r 2 and the relative analysis error RPD value. The relative analysis error RPD value is the ratio of the sample standard deviation SD to the prediction standard error RMSEP, and its calculation formula is as follows:

[0105]

[0106] where n is the number of samples in the test set, and y i is the actual value of the i-th sample in the test set, is the predicted value of the i-th sample in the test set, is the average value of all actual values in the test set. When RPD > 2.5, the calibration model is an excellent model and can be used for excellent prediction.

[0107] Figure 8 is the fitting result of the soil available phosphorus test set in this implementation, Figure 9 is the migration curve of soil available phosphorus during the whole growth period of corn in this example. Its determination coefficient r 2 is 0.9403, the correlation coefficient is relatively high, and the modeling effect is good; the relative analysis error RPD value is 4.071. When RPD > 2.5, the calibration model is a good model and can quantitatively predict the soil available phosphorus content.

[0108] The present invention is developed and designed based on the principle that when the TDR sensor is subjected to an electromagnetic field applied to the soil to be tested, the soil will exhibit electrochemical characteristics, that is, under step frequency sampling, the soil will experience differences in the real and imaginary parts of the dielectric constant caused by electronic displacement polarization, ion polarization and steering polarization. Because there are charged colloids and a variety of secondary minerals in the soil that constitute the clay structure of the soil, the soil and its solution form a complex Donan system, so it is difficult to solve it by traditional instrument correction plus empirical formulas. The present invention monitors the differences in the real and imaginary parts of the soil dielectric constant during the growth of crops by installing and deploying multi-layer time domain reflectometers and temperature detection integrated probes in situ, and continuously measures the soil temperature and effective nutrient data of multiple soil layers at one time through the combination of principal component analysis and neural network algorithm modeling, thereby fundamentally solving the problem of continuous monitoring of soil nutrient data that needs to be solved in the nutritional management of crops throughout the growth period and soil quality management.

[0109] The scope of the in-situ rapid measurement of soil available nutrients of the present invention includes the combined determination of soil moisture content in multiple soil layers, soil conductivity, pore water conductivity, total salt content, temperature, and the content of available nitrogen, phosphorus, and potassium of a large number of elements as well as calcium, magnesium, iron, manganese, copper, and zinc.

[0110] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A step-by-step time-domain reflectometry method for rapid in-situ measurement of soil available nutrients. Features: The step-by-step time-domain reflection soil available nutrient in-situ rapid measurement system is adopted, which includes the following steps: Step S01: placing the in-situ multi-layer probe device in the soil, energizing the magnetic spring, and extending the time domain reflectometer and temperature detection integrated probe toward the soil; Step S02: The data acquisition control module of the information acquisition control box receives the instruction of the client operating system, and the data acquisition control module sends the command to the signal transceiver processing module and the temperature sensor at the same time. The signal transceiver processing module generates 1024 sets of step frequency continuous wave signals in sequence, and the step frequency signals are transmitted to the soil through the in-situ multi-layer probe device. When encountering soil in different states, signal reflection is generated, and then the test signal and the reflection response signal are separated by directional coupling to form a reflection reception signal; when a 1024-set step frequency scan is completed, the obtained series of incident representative signal reception data, reflection reception signal reception data and voltage signals collected by the temperature sensor are transmitted to the intelligent client operating system through the data acquisition control module and the transmission module; Step S03: The operating system represents 1024 sets of step frequencies of the TDR emission frequency from high to low as f H , f L . Then the high-low frequency difference is the frequency bandwidth . Different center frequencies f C of different bandwidths are obtained according to different bandwidths. Then the center frequency f C〡B〡 of the frequency bandwidth B is expressed as: ； Convert the frequency-domain signal into the central time domain with different bandwidths through inverse Fourier transform ; Step S04: measuring soil volume water content; In the step S04, the measurement of the soil volume water content is specifically as follows: the center time domain in the measurement of the soil volume water content , the low-high frequency difference thereof is 30 MHz to 1 GHz. Through the Topp empirical formula, based on the measurement of the apparent dielectric constant, the soil volume water content is calculated according to the following formula , ； where a 1 , a 2 , a 3 are the coefficients of the polynomial terms, and b is the adjustment constant; Step S05: measuring soil conductivity; In the step S05, the measurement of the soil conductivity is specifically as follows: the center time domain in the soil conductivity measurement , and the difference between its low and high frequencies is from 1 MHz to 1 GHz; Obtain the apparent dielectric constant. The speed at which electromagnetic waves propagate along the cable is inversely proportional to the square root of the apparent dielectric constant of the surrounding medium. The apparent dielectric constant is converted into the formula: ； Among them, K a is the apparent dielectric constant of the medium, v is the propagation speed of the electromagnetic wave in the medium; c is the propagation speed of the electromagnetic wave in vacuum, with the unit of centimeter / nanosecond; 2L is the distance that the electromagnetic wave propagates back and forth on the probe, L is the one-way propagation distance of the electromagnetic wave on the probe once, with the unit of centimeter; d(t) 2 -d(t) 1 is the propagation time, with the unit of nanosecond; The soil conductivity is obtained according to the following formula :[[]]END]] ； Where: K a is the apparent dielectric constant of the medium, V 1 is the voltage at the starting end of the central time domain probe in the soil, V 2 is the voltage at the ending end of the central time domain probe in the soil; Step S06: measuring the available nutrient data through principal component analysis and neural network modeling; The step S06 comprises the following steps: (1) Input the central time domain data set and temperature data of each time and soil layer monitored by the soil TDR sensor and temperature sensor during the 3414 test; (2) Input the data of available nutrient content of soil samples determined by chemical methods; (3) Central time domain preprocessing: Use derivative and / or SG smoothing and / or multivariate scatter correction MSC and / or variable standardization SNV to preprocess the central frequency; use robust principal component analysis to eliminate abnormal samples of soil samples; (4) Extraction of characteristic center time domain: Use the continuous projection algorithm SPA and / or the uninformative variable elimination algorithm UVE and / or the genetic algorithm GA method to select characteristic bands and extract the characteristic center time domain for model building; use the robust principal component analysis method to eliminate abnormal samples of soil samples; (5) Multivariate correction: The Kennard-Stone algorithm was used to divide the calibration set and test set for establishing the soil nutrient content model, and the partial least squares method (PLS) was used to establish the soil nutrient content model; the robust principal component analysis method was used to eliminate abnormal soil samples; (6) Precision judgment: Calculate the determination coefficient r for precision discrimination of the soil available nutrient model after central time-domain interpretation 2 and the relative analysis error RPD value, and judge r 2 and the RPD value. If it meets the experimental allowable range, proceed to the next step. If not, return to step 1 and start over, and calculate the determination coefficient r again 2 and the relative analysis error RPD value, and select the best model; (7) Complete modeling; (8) Prepare prediction: input unknown soil sample step frequency data; (9) Prediction: Predict the effective nutrient content of soil samples using the established soil sample characteristic center frequency and effective nutrient content model and the step frequency data of unknown soil samples; (10) Complete the forecast; The described step-by-step time domain reflectometry in-situ rapid measurement system for soil available nutrients includes an in-situ multi-layer probe device, an information acquisition and control box, and a client, which are interconnected. The in-situ multi-layer probe device is connected to the information acquisition and control box. The information acquisition and control box includes a signal transceiver and processing module, a data acquisition and control module, a 4G / 5G transmission module, and a power supply module, which are interconnected. The information acquisition and control box is connected to the client operating system through the Internet. The in-situ multi-layer probe device includes an outer sleeve and an inner sleeve that are sleeved on each other. The bottom of the outer sleeve is a conical structure. Measuring holes are evenly and equidistantly arranged from top to bottom on one side wall of the outer sleeve. Horizontally arranged mounting grooves are evenly and equidistantly arranged from top to bottom on the inner sleeve. The opening positions of the mounting grooves match the positions of the measuring holes. A time domain reflectometer and temperature detection integrated probe is arranged in the mounting groove. The time domain reflectometer and temperature detection integrated probe is connected to one end of a magnetic spring, and the other end of the magnetic spring is connected to the inner side wall of the mounting groove. The magnetic spring is connected to a power supply. The time domain reflectometer and temperature detection integrated probe is connected to the information acquisition and control box.

2. The step-by-step time domain reflectometry in-situ rapid measurement method for soil available nutrients according to claim 1, characterized in that: In step S02, the signal transceiver and processing module sequentially generates 1024 sets of stepped frequency continuous wave signals with a frequency sweep range in the range of 1M - 1GHz.

3. The step-by-step time domain reflectometry in-situ rapid measurement method for soil available nutrients according to claim 1, characterized in that: In step S02, the incident representative signal and the reflected received signal are respectively received by their respective receivers, and the received data is stored.

4. The step-by-step time domain reflectometry in-situ rapid measurement method for soil available nutrients according to claim 1, characterized in that: The time domain reflectometer and temperature detection integrated probe includes a TDR sensor probe and a patch type soil temperature sensor. The TDR sensor probe and the patch type soil temperature sensor are cast into an integrated structure by epoxy resin.

5. The step-by-step time domain reflectometry in-situ rapid measurement method for soil available nutrients according to claim 1, characterized in that: A rubber pad covers the outside of the test hole of the outer sleeve.

6. The step-by-step time domain reflectometry in-situ rapid measurement method for soil available nutrients according to claim 1, characterized in that: The time domain reflectometer has 1024 sets of stepped frequency signals in the frequency range of 1M - 1GHz.

Images