Handheld multi-depth soil parameter detection device for seeding decision

Abstract

The invention discloses a handheld multi-depth soil parameter detection device for seeding decision. Wherein the supporting part inserted into the soil is fixed at the bottom of the handheld central part; the electrode mounting module and the spectrum detection mechanism are mounted on the two sides of the bottom of the supporting part, and the penetrating type conductivity detection assembly is fixed to the lower end of the supporting part; the capacitance information processing module is mounted at the middle lower part of the supporting part; the pressure sensor is fixed in an opening in the lower part of the shell of the handheld pivot part; the pressure sensor is positioned between the top connecting piece and the stress contact surface of the handheld pivot part; in the process that the supporting part penetrates into the soil, the single-chip microcomputer combines depth information provided by the laser ranging module, the numerical value measured by the pressure sensor is directly used as comprehensive penetration resistance, and the soil firmness is represented through the numerical value of the pressure sensor. The four sensing mechanisms of spectrum, capacitance, conductance and pressure are integrated, clutters are eliminated from multiple dimensions by the soil firmness prediction model, and the confidence coefficient is ensured.

Description

Technical Field

[0001] This invention belongs to the technical field of equipment for soil composition analysis, specifically a handheld multi-depth soil parameter detection device for sowing decisions. Background Technology

[0002] Soil is the foundation of crop growth. With the development of precision agriculture and soil carbon sequestration technology, the demand for monitoring soil physicochemical properties is increasing. Soil moisture and nutrient distribution directly determine crop growth status and yield. Therefore, rapidly and accurately obtaining in-situ multi-parameter soil information is of great significance for guiding scientific irrigation and fertilization, assessing soil health, and improving crop yield.

[0003] However, current soil testing technology faces a sharp contradiction between "high single-point accuracy and difficulty in integrating multiple parameters" and "lack of spatiotemporal dimensions in data." First, most existing handheld testing devices have limited functions or combine sensors through simple physical assembly, resulting in significant technical bottlenecks: such as electromagnetic interference (EMC) issues between sensor signals, structural contradictions between optical path structure and probe strength, and contact coupling challenges in in-situ measurements.

[0004] Secondly, and more importantly, traditional portable detectors typically only provide instantaneous readings, lack the ability to identify and filter data, and usually only record information, which can lead to a large number of errors. Summary of the Invention

[0005] To address the problems existing in the background technology, the present invention provides a handheld multi-depth soil parameter detection device for sowing decisions. The technical solution includes: a handheld central part, handheld rods, a support part, a capacitance information processing module, an electrode mounting module, an invasive conductivity detection component, and a spectral detection mechanism. Two handheld rods are fixed to both sides of the handheld central part, and the support part inserted into the soil is fixed to the bottom of the handheld central part. The electrode mounting module and the spectral detection mechanism are installed on both sides of the bottom of the support part, and the invasive conductivity detection component is fixed to the lower end of the support part. The capacitance information processing module is installed in the lower middle part of the support part.

[0006] The support structure includes a central support rod, a top connector, and a bottom connector. These components are fixed sequentially from top to bottom. The top connector passes through a through-hole in the center of the pressure sensor and is secured with a fastening nut. The pressure sensor is fixed in an opening at the bottom of the housing of the handheld pivot. The pressure sensor is positioned between the top connector and the force-bearing contact surface of the handheld pivot. During the penetration of the support structure into the soil, the microcontroller, combined with the depth information provided by the laser ranging module, directly uses the value measured by the pressure sensor as the overall penetration resistance, thus characterizing the soil firmness.

[0007] A microcontroller is installed inside the housing; at the bottom of the housing, a laser ranging module is installed, with the probe of the laser ranging module facing parallel to the support part, so that the laser ranging module, which plays the role of determining depth, is vertically downward aligned with the soil surface.

[0008] The handheld lever includes a right handheld lever and a left handheld lever. The right handheld lever and the left handheld lever pass through pre-set holes on both sides of the housing of the handheld pivot part, and the right handheld lever and the left handheld lever are connected and fixed to the housing through a high-strength threaded structure.

[0009] The spectral detection mechanism is located in a single opening on one side of the bottom of the support section.

[0010] The spectral detection mechanism includes: a spectral mounting housing, a spectral sensor module, and an optical lens; the front end of the spectral mounting housing is designed with a 45° inward chamfer facing downward, which can generate a normal component force during the insertion process, forcibly pushing and pressing the soil particles on the contact surface onto the surface of the optical lens.

[0011] The electrode mounting module includes: an electrode mounting housing and an electrode plate; the electrode plate is mounted in the electrode mounting housing; the capacitance information processing module includes: a capacitance processor mounting housing and a capacitance processor, the capacitance processor and the electrode plate are connected, the electrode plate is mounted in the capacitance processor mounting housing, and an equipotential active shielding layer is provided between the capacitance processor and the capacitance processor mounting housing; the electrode plate is connected to the capacitance processor.

[0012] The equipotential active shielding layer is 2mm wider than the capacitor processor in all four directions, and the two are isolated by a 1.6mm thick FR-4 insulating substrate, which forces the electric field lines at the edge of the capacitor to diverge only towards the soil side.

[0013] The piercing conductivity testing component includes a tail conductor, a middle insulator, and a head conductor. The tail conductor and the head conductor are threaded to the middle insulator from the outside and inside, respectively, and the head conductor is inserted into the soil first.

[0014] The soil firmness prediction model used in the microcontroller identifies and counteracts the interference of porosity changes. The steps for establishing the soil firmness prediction model are as follows:

[0015] Step 1: Sample preparation and soil information collection: Select soil samples with different textures and prepare soil samples with different moisture content gradients. Use this device to collect four raw data at the same time: spectrum, capacitance, conductivity and pressure.

[0016] Step 2, Determination of true physicochemical values ​​of soil: The moisture content was measured by the oven drying method, the organic matter content was measured by the potassium dichromate method, and the bulk density was measured by the ring sampler method to determine the true physicochemical properties of the soil.

[0017] Step 3: Construct a feature space for multi-source heterogeneous soil parameters: Divide the signal sources affecting the inversion of soil physicochemical properties into static environmental factors and dynamic sensing factors, and construct a high-dimensional input vector;

[0018] Step 4: Effective Contact State Gating Based on Quantization Judgment Rules: The microcontroller 106 only marks the current data frame as a "valid sample" and stores it in the buffer when specific quantization judgment rules are met. The quantization judgment rules include the following three rules: Rule 1: Pressure Coupling Threshold Rule; Rule 2: Penetration Stability Rule; and Rule 3: Depth-Logic Verification Rule, where:

[0019] Rule 1, the pressure coupling threshold rule, refers to monitoring the real-time readings of pressure sensors.

[0020] Set the effective coupling pressure threshold. .when If the probe is found to be suspended, not fully compacted, or passing through a large pore, the spectral and capacitance data will have a large air medium error, and the system will automatically discard the data frame.

[0021] Rule 2, the penetration stability rule, calculates the rate of pressure change per unit time. ;when When the impact threshold is exceeded, the device is determined to be in a period of unsteady deformation due to severe impact or contact with hard objects such as rocks. Data recording is suspended until the rate of change returns to the stable range, so as to eliminate the nonlinear interference of dynamic shear deformation on the measurement of soil dielectric constant.

[0022] Rule 3, the depth-logic verification rule, is based on the real-time depth obtained from the laser ranging module 108. Only when The change is monotonically increasing and the increment is Within a reasonable step size range Sampling is only triggered when the user retracts the probe to prevent the same depth data from being recorded repeatedly.

[0023] Step 5: Construct a multidimensional decoupled model of soil parameters based on pressure constraints:

[0024] Construct a training dataset, dividing the valid data frames filtered in step four into a training set and a validation set; referencing relevant domain standards, set the division ratio to 70% for the training set and 30% for the validation set; the method for determining positive samples is: data points that meet the criteria have their label values ​​determined by the laboratory standard method; the input layer dimension is... Including: spectral characteristic principal components Normalized capacitance value Pressure value Electrical conductivity and depth ;

[0025] The quantification discrimination rule is as follows: < ;

[0026] Step 6: Use the trained model to perform real-time calculations on the area to be tested: The host computer mobile terminal receives the raw data packets uploaded by the microcontroller, inputs them into the decoupled model in real time, and outputs the equivalent standard bulk density and moisture content after pressure correction.

[0027] In step six, the grading criteria are output. The grading criteria are divided into five levels: very low <10%, low: 10%-15%, medium: 15%-20%, high: 20%-25%, and very high >25%.

[0028] In step three, static environmental factors This includes geospatial information and preset parameters obtained through a host computer mobile terminal, including: GNSS latitude and longitude coordinates, soil texture type at the current measuring point (selectable sandy soil / loam / clay), user-preset parameters or parameters obtained from cloud maps, and historical basic bulk density range. These factors serve as boundary constraints for the model; dynamic sensing factors... This refers to the high-frequency response signal collected in real time by the device during the soil penetration process, which represents the dynamic sensing factors. It is decomposed into three dimensions of coupled characteristics: electromagnetic response characteristics, mechanical response characteristics, and electrochemical response characteristics.

[0029] Among them, the electromagnetic response characteristics include those acquired by the spectral sensor module. Band reflectivity data, and dielectric constant voltage signal output by the capacitance information processing module. The mechanical response characteristics are the penetration resistance values ​​fed back in real time by the pressure sensor. and resistance change rate ;

[0030] Electrochemical response characteristics are obtained from soil conductivity signals acquired by an invasive conductivity detection module. .

[0031] The beneficial effects of this invention are as follows:

[0032] 1. It integrates four major sensing mechanisms: spectroscopy, capacitance, conductivity, and pressure. The soil firmness prediction model used eliminates clutter from multiple dimensions, ensuring that the final output spatial distribution map of soil parameters has extremely high confidence.

[0033] 2. Traditional shielding layers are often the same size as the electrodes. This invention uses a larger shielding layer to solve the problem found in experiments that shielding layers of the same size cannot completely block backward electric field leakage.

[0034] 3. To address the issue that handheld devices are susceptible to interference from vibration, gaps, and unsteady contact during field operations, this invention does not adopt the traditional all-time recording mode. Instead, it establishes an "effective contact quantification discrimination rule." The microcontroller only marks the data frame at the current moment as a "valid sample" and stores it in the cache when a specific combination of rules is met, thus avoiding such interference. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of an embodiment of a handheld multi-depth soil parameter detection device for seeding decision-making according to the present invention.

[0036] Figure 2 This is a schematic diagram of the support portion and the handheld central portion of the present invention;

[0037] Figure 3 This is a schematic diagram of the spectral detection mechanism of the present invention;

[0038] Figure 4 This is a schematic diagram of the capacitance information processing module of the present invention;

[0039] Figure 5 This is a schematic diagram of the piercing conductivity detection component of the present invention;

[0040] Figure 6 This is a schematic diagram of the back of the support portion of the present invention;

[0041] Figure 7 This is a schematic diagram showing the installation of the host computer in this invention.

[0042] Figure 8 This is a flowchart of the electrical control system for the handheld multi-depth soil parameter detection device of the present invention.

[0043] Figure 9 A flowchart illustrating the modeling process for the predictive model of the device of the present invention.

[0044] 1. Handheld central part 101 Handheld central base 102 Start-up indicator light 103 power module 104 General Expansion Interfaces 105 main power switch 106 microcontroller 107 Fastening Nut 108 Laser Ranging Module 2-hand stick 201 Right-handed stick 202 Left-handed stick 3 Supporting parts 301 Middle Support Rod 302 Top Connector 303 bottom connector 4 Capacitor Information Processing Module 401 Capacitor Processor Mounting Case 402 Capacitor Processor 5-electrode mounting module 501 Electrode Mounting Housing 502 electrode plate 6-Piercing Conductivity Detection Component 601 tail conductor 602 Middle Insulator 603 Head Conductor 7 Spectroscopic Testing Institutions 701 Spectrum Mounting Housing 702 Spectral Sensor Module 703 Optical Lens 8 pressure sensors Detailed Implementation

[0045] The present invention will be further described in detail below with reference to the accompanying drawings.

[0046] like Figures 1-9The embodiment of the present invention shown includes: a handheld central part 1, handheld rods 2, a support part 3, a capacitance information processing module 4, an electrode mounting module 5, an invasive conductivity detection component 6, and a spectral detection mechanism 7. Two handheld rods 2 are fixed to both sides of the handheld central part 1, and the support part 3, which is inserted into the soil, is fixed to the bottom of the handheld central part 1. The electrode mounting module 5 and the spectral detection mechanism 7 are mounted on both sides of the bottom of the support part 3, and the invasive conductivity detection component 6 is fixed to the lower end of the support part 3. The capacitance information processing module 4 is mounted in the lower middle part of the support part 3.

[0047] The support part 3 includes a middle support rod 301, a top connector 302, and a bottom connector 303. The top connector 302, the middle support rod 301, and the bottom connector 303 are fixed sequentially from top to bottom. The top connector 302 passes through the through hole in the middle of the pressure sensor 8 and is fixed by a fastening nut 107. The pressure sensor 8 is fixed in the opening at the bottom of the housing 101 of the handheld pivot part 1. The pressure sensor 8 is precisely positioned between the top connector 302 and the force-bearing contact surface of the handheld pivot part 1. During the process of the support part 3 penetrating the soil, the microcontroller, combined with the depth information provided by the laser ranging module, directly uses the value measured by the pressure sensor 8 as the comprehensive penetration resistance, and uses the pressure sensor value to characterize the soil firmness.

[0048] At the bottom of the housing 101, a laser ranging module 108 is installed. The probe of the laser ranging module 108 is parallel to the support part 3, so that the laser ranging module 108, which plays the role of determining depth, is vertically downward aligned with the soil surface.

[0049] When the user presses down on the hand lever to insert the device into the soil, the laser ranging module 108 measures the vertical distance from the bottom of the housing to the ground surface in real time. The total length of the support rod and sensor is known to be a fixed value. The microcontroller or host computer APP uses the formula The system automatically calculates the precise depth of the sensor in the ground. Simultaneously, an algorithm eliminates interference from abrupt changes in laser ranging caused by ground vegetation; this allows users to precisely pause at a preset depth for measurement, rather than relying solely on experience-based estimations.

[0050] The housing 101 houses a power module 103 and a microcontroller 106. The microcontroller 106 serves as the main control unit and mainly includes an AD conversion module and a wireless communication module (Bluetooth). The power module 103 is responsible for providing a stable operating voltage to the microcontroller 106 and all sensors.

[0051] In this embodiment, the handheld lever 2 includes a right handheld lever 201 and a left handheld lever 202. The right handheld lever 201 and the left handheld lever 202 pass through preset holes on both sides of the housing 101 of the handheld pivot part 1, and are connected and fixed to the housing 101 by a high-strength threaded structure. This double-sided symmetrical design conforms to the ergonomic principle, making it convenient for operators to hold with both hands and apply vertical downward force, ensuring stability when inserted into the soil.

[0052] In this embodiment, a start indicator light 102 is provided on the surface of the housing 101 to indicate that the device has switched from standby to working state; the main power switch 105 is a self-locking switch to control the overall power supply of the entire machine. The material of the middle support rod 301 is 304 stainless steel or high-strength alloy.

[0053] A spectral detection mechanism 7 is located on one side of the single opening at the bottom of the support part 3. The spectral detection mechanism 7 includes a spectral mounting housing 701, a spectral sensor module 702, and an optical lens 703. To address the problem of high light diffuse reflection loss and low signal-to-noise ratio caused by soil surface roughness in in-situ measurements, this invention features a special design for the front end of the spectral mounting housing 701. Specifically, the front end of the spectral mounting housing 701 has a 45° inward chamfer facing downwards (bottom), and the surface of this chamfer is hardened. In actual testing, it was found that ordinary flat-headed or large-angle probes easily compress the soil and create cracks during insertion, leading to external light leakage. The acute-angled wedge structure used in this invention generates a normal force during insertion, forcibly flattening and compacting the soil particles on the contact surface onto the surface of the optical lens 703. This integrated "cutting-compacting" structure not only utilizes the soil itself to form a zero-gap physical seal, creating an absolute underground darkroom environment, but also effectively eliminates random signal fluctuations caused by soil porosity, improving the stability of spectral data by approximately 20%.

[0054] Electrode mounting module 5 includes: electrode mounting housing 501 and electrode plate 502; electrode plate 502 is mounted in electrode mounting housing 501; capacitor information processing module includes: capacitor processor mounting housing 401 and capacitor processor 402, capacitor processor 402 is connected to electrode plate 502, electrode plate 502 is mounted in capacitor processor mounting housing 401, and an equipotential active shielding layer is provided between capacitor processor 402 and capacitor processor mounting housing 401; electrode plate 502 is connected to capacitor processor 402.

[0055] In this embodiment, to address the parasitic capacitance interference of the metal grip stick unique to handheld devices and the edge effect of the soil's non-sensitive area, the present invention employs an asymmetric field constraint structure design; such as Figure 4 and Figure 6The equipotential active shielding layer on the back of the capacitance information processing module 4 shown is 2mm wider than the capacitance processor 402 in all four directions, and the two are isolated by a 1.6mm thick FR-4 insulating substrate. This "large-enclosing-small" size margin design forces the electric field lines at the capacitor edge to diverge only towards the soil side, completely severing the coupling path between the metal support rod 301 and the capacitance measurement. This prevents baseline drift regardless of changes in the user's grip.

[0056] like Figure 5 The piercing conductivity detection component 6 shown includes a tail conductor 601, a middle insulator 602, and a head conductor 603. The tail conductor 601 and the head conductor 603 are threaded to the middle insulator 602 from the outside and inside, respectively, with the head conductor 603 penetrating the soil first. Since the piercing conductivity detection component 6 is located at the bottom, it must withstand the greatest penetration resistance and lateral shear force. Experiments show that traditional adhesive or planar contact connections are prone to breakage at the connection point when encountering hard soil layers. Therefore, this invention innovatively adopts a stress-decoupling embedded load-bearing structure. The middle insulator 602 is made of high-modulus PEEK (polyetheretherketone) material, and its two ends are machined with internal threads deeper than 10 mm. The head conductor 603 and the tail conductor 601 are screwed into the insulator through corresponding threads, allowing the middle insulator to not only provide electrical isolation but also serve as a core load-bearing skeleton to transmit penetration pressure. Tests have shown that the structure can withstand vertical loads greater than 800N without deformation, ensuring structural integrity under harsh working conditions.

[0057] The modeling process of the soil firmness prediction model used in a microcontroller is as follows: Figure 9 As shown, the interference from changes in porosity was identified and counteracted. The steps for establishing the prediction model for water content, organic matter content, and soil firmness are as follows:

[0058] Step 1: Sample Preparation and Soil Information Collection: Select soil samples of different textures. Prepare soil samples with different moisture content gradients using methods such as baking and sieving. Simultaneously collect four types of raw data: spectral, capacitance, conductivity, and pressure using this device.

[0059] Step 2: Determination of True Soil Physicochemical Values: The soil's true physicochemical properties were determined using the oven-drying method to measure moisture content, the potassium dichromate method to measure organic matter content, and the ring sampler method to measure bulk density. This increases the reliability of the control data.

[0060] Step 3: Constructing a feature space for multi-source heterogeneous soil parameters: This invention divides the signal sources that affect the inversion of soil physicochemical properties into static environmental factors and dynamic sensing factors, and constructs a high-dimensional input vector.

[0061] Among them, static environmental factors ( This includes geospatial information and preset parameters obtained through a host computer mobile terminal. Specifically, this includes: GNSS latitude and longitude coordinates, soil texture type at the current measuring point (selectable as sandy soil / loam / clay, user-preset or obtained from a cloud map), and historical basic bulk density range. These factors serve as boundary constraints for the model; dynamic sensing factors ( This refers to the high-frequency response signal collected in real time by the device during soil penetration. This invention innovatively decomposes this into three coupled characteristics: electromagnetic response characteristics, mechanical response characteristics, and electrochemical response characteristics. The electromagnetic response characteristics include those collected by the spectral sensor module 702. Band reflectivity data, and dielectric constant voltage signal output by capacitance information processing module 4. The mechanical response characteristics are determined by the penetration resistance value fed back in real time by pressure sensor 8. and resistance change rate .

[0062] The electrochemical response characteristics are soil conductivity signals acquired by the piercing conductivity detection component 6. .

[0063] Step 4: Effective Contact State Gating Based on Quantitative Judgment Rules: Addressing the issue of handheld devices being susceptible to vibration, gaps, and unsteady contact interference during field operations, this invention does not employ the traditional all-time recording mode. Instead, it establishes "effective contact quantitative judgment rules." The microcontroller 106 only marks the current data frame as a "valid sample" and stores it in the cache when a specific combination of rules is met. The quantitative judgment rules include the following three: Rule 1: Pressure Coupling Threshold Rule; Rule 2: Penetration Stability Rule; and Rule 3: Depth-Logic Verification Rule, wherein:

[0064] Rule 1, the pressure coupling threshold rule, refers to monitoring the real-time readings of pressure sensor 8.

[0065] Set the effective coupling pressure threshold. .when If the probe is found to be suspended, not fully compacted, or passing through large pores (such as wormholes or fissures), the spectral and capacitance data will have a large air medium error, and the system will automatically discard the data frame.

[0066] Rule 2, the penetration stability rule, calculates the rate of pressure change per unit time. .when Exceeding the preset impact threshold If the device is in a period of unsteady deformation due to severe impact or contact with hard objects such as rocks, data recording is suspended until the rate of change returns to a stable range, in order to eliminate the nonlinear interference of dynamic shear deformation on the measurement of soil dielectric constant.

[0067] Rule 3, the depth-logic verification rule, is based on the real-time depth obtained from the laser ranging module 108. Only when The change is monotonically increasing (i.e., inserting downwards) and the increment is... Within a reasonable step size range Sampling is only triggered when the probe is pulled back. This prevents the user from repeatedly recording the same depth data.

[0068] Step 5: Constructing a multidimensional decoupling model for soil parameters based on pressure constraints: To address the cross-sensitivity interference of soil compaction (bulk density) on dielectric constant and spectral reflectance in traditional detection methods, this invention constructs a machine learning prediction model incorporating physical constraint mechanisms. First, a training dataset is constructed, dividing the valid data frames selected in Step 4 into a training set and a validation set. Referring to relevant industry standards, the division ratio is set to 70% for the training set and 30% for the validation set. The method for determining positive samples is provided here: Meeting the quantitative discrimination rules (… < The data points are labeled with true values ​​obtained by the laboratory standard method; this model involves multi-dimensional feature vectors: the input layer dimension is... Specifically, this includes: principal components of spectral features. Normalized capacitance value Pressure value Electrical conductivity and depth The model architecture and key parameters are also provided here.

[0069] In this embodiment, an improved BP neural network is used for nonlinear regression prediction. To achieve the "de-compaction effect," the soil firmness prediction model structure is designed as follows:

[0070] First, the network topology of the soil firmness prediction model adopts a double-hidden-layer structure. The input layer has 5 nodes; the first hidden layer has 12 nodes, and the TanH hyperbolic tangent function is used to accelerate convergence; the second hidden layer has 8 nodes, and the ReLU activation function is used to maintain nonlinear characteristics; the output layer has 3 nodes, corresponding to soil moisture content, organic matter content, and firmness, respectively.

[0071] Secondly, the unique physical constraint mechanism of the soil firmness prediction model is the introduction of a pressure penalty term into the loss function during model training. This is because the pressure value... The soil firmness prediction model is configured to learn this physical law, which is negatively correlated with soil porosity: when When the soil compaction factor increases while the moisture content label remains unchanged, the soil compaction prediction model automatically decreases. The contribution of the weight.

[0072] The core hyperparameters of the soil firmness prediction model are set as follows: the learning rate is initially set to 0.01, and an adaptive decay strategy is adopted. The maximum number of iterations is 2000. The momentum factor is 0.9 to suppress local oscillations. Convergence threshold: .

[0073] Step Six: Multi-depth Spatiotemporal Data Fusion and Visualization Output: The trained model is used to perform real-time calculations on the area under test, generating a spatial distribution map with geographic information.

[0074] First, real-time data stream processing is performed: the host computer mobile terminal receives the raw data packets uploaded by the microcontroller and inputs them into the decoupled model in real time. The system output is no longer a single electrical signal, but an "equivalent standard bulk density and moisture content" after pressure correction, eliminating measurement errors caused by different insertion forces of the operator.

[0075] Secondly, there is three-dimensional coordinate matching: using the mobile terminal's built-in GNSS module to obtain latitude and longitude. Combined with the fine depth obtained by the laser ranging module , build Three-dimensional spatial coordinate system.

[0076] Next is spatial distribution mapping and grading: the natural breakpoint method is used to grade and display the whole field data to intuitively show the differences in soil fertility; the grading standard is: the soil parameters obtained by inversion are divided into 5 levels: very low (<10%), low (10%-15%), medium (15%-20%), high (20%-25%), and very high (>25%).

[0077] The specific workflow of this invention is as follows:

[0078] The user inserts the support part 3 into the handheld hub 1 and locks it in place. The mobile terminal is mounted on the universal expansion interface 104 via the bracket. The user presses the main power switch 105, illuminating the start indicator light 102. The user opens the accompanying app on their mobile phone, and the system automatically establishes a connection via Bluetooth.

[0079] To obtain high-resolution vertical distribution profiles of soil parameters, this device adopts a "depth-triggered continuous recording" working mode.

[0080] The user inserts the probe into the soil at a steady speed, and the laser ranging module 108 calculates the real-time depth at a high frequency (e.g., 50Hz). Simultaneously, to address signal noise caused by unstable contact during dynamic measurement, the microcontroller 106 runs a pressure-based dynamic cleaning algorithm.

[0081] First, the analog signal from pressure sensor 8 is aligned with the signals from the capacitance and spectral sensors at the microsecond level. After data synchronization, a contact quality gating step is performed; the software will monitor the pressure value in real time. When the pressure value is detected When the current probe is below the preset effective coupling threshold, it is determined that the probe is in a pore or slip state, and abnormal spectral and capacitance values ​​at that moment are automatically discarded; at the same time, for the retained effective data points, the real-time penetration speed is considered. Perform adaptive weighted averaging. When the speed... When the speed is fast, the sampling frequency is automatically increased to ensure spatial resolution (e.g., a set of data is recorded every 1 mm).

[0082] Using the above strategy, the device can directly generate smooth, continuous, and contact artifact-free vertical profiles of soil moisture content and firmness without stopping the machine.

[0083] The 106 microcontroller packages the raw sensor data and depth data and sends them to the mobile phone via Bluetooth. The mobile app first uses the phone's built-in GNSS to obtain latitude and longitude, and then merges the "GNSS coordinates + soil depth + soil physicochemical parameters (moisture content, organic matter, etc.)". The app interface not only displays the current values, but also generates a spatial distribution map (heat map) based on geographical location, intuitively showing which areas of the farmland are lacking water or fertilizer, directly assisting in sowing and fertilization decisions.

[0084] After the measurement is completed, remove the device and prepare for the next measurement.

[0085] This invention upgrades a simple soil testing tool into a decision-making system with spatial mapping capabilities by adding laser ranging and a universal mobile phone interface, greatly improving the efficiency and usability of precision agriculture data collection. To further verify the actual effectiveness of this device in complex field environments and its significant advantages over existing technologies, comparative experiments were conducted: First, addressing the cross-sensitivity problem of soil bulk density to dielectric constant measurement, under standard loam conditions with a constant moisture content of 15%, the soil bulk density was gradually increased from 1.1 g / cm³ to 1.6 g / cm³ using a hydraulic device. The experimental results showed that the measurement reading of the traditional pressure-free correction model drifted drastically with increasing compaction, artificially increasing from 14.8% to 21.5%, resulting in an absolute error of +6.7%. However, this invention, using a pressure-constrained decoupling model, identified and offset the interference of porosity changes, with the reading remaining stable in the 14.9%-15.3% range, and the maximum absolute error being only 0.3%, confirming the decisive contribution of the pressure penalty term to achieving the "de-compaction effect." Second, regarding the signal-to-noise ratio of the spectral detection mechanism 7... In comparative testing, comparing the traditional planar window probe with the 45° blade-shaped "cutting-compacting" integrated probe of this invention, data showed that the former had a coefficient of variation (CV) of up to 8.4% due to the tiny air gaps at the contact surface, while the present invention achieved micron-level self-sealing contact by utilizing the normal component force of the blade edge, reducing the CV value to 1.2%, and improving the signal-to-noise ratio by 18dB compared to the control group, effectively solving the problem of stray light interference in in-situ spectral measurements. Finally, through a full-process contact pressure calibration experiment, it was found that when the penetration resistance was less than 20N, the standard deviation of the contact resistance was greater than 200Ω, and the data was in the nonlinear and violently fluctuating region. However, when the pressure exceeded 20N, the rate of change of the contact resistance quickly converged to less than 2%. This data quantitatively verified the scientific validity and necessity of setting the "effective contact gate threshold" to 20N in this invention, ensuring that the final output spatial distribution map of soil parameters has extremely high confidence.

Claims

1. A handheld multi-depth soil parameter detection device for seeding decision-making, characterized in that, include: The handheld central part (1), handheld rods (2), support part (3), capacitance information processing module (4), electrode mounting module (5), piercing conductivity detection component (6), and spectral detection mechanism (7) are provided. Two handheld rods (2) are fixed on both sides of the handheld central part (1), and the support part (3) inserted into the soil is fixed at the bottom of the handheld central part (1). The electrode mounting module (5) and spectral detection mechanism (7) are installed on both sides of the bottom of the support part (3), and the piercing conductivity detection component (6) is fixed at the lower end of the support part (3). The capacitance information processing module (4) is installed in the lower middle part of the support part (3). The support part (3) includes: a middle support rod (301), a top connector (302) and a bottom connector (303). The top connector (302), the middle support rod (301) and the bottom connector (303) are fixed from top to bottom. The top connector (302) passes through the through hole in the middle of the pressure sensor (8) and is fixed by a fastening nut (107). The pressure sensor (8) is fixed in the opening at the bottom of the housing (101) of the handheld pivot part (1). The pressure sensor (8) is positioned between the top connector (302) and the force contact surface of the handheld pivot part (1). During the process of the support part (3) penetrating the soil, the microcontroller combines the depth information provided by the laser ranging module and uses the value measured by the pressure sensor (8) directly as the comprehensive penetration resistance. The pressure sensor value is used to characterize the soil firmness. A microcontroller (106) is installed inside the housing (101); at the bottom of the housing (101), a laser ranging module (108) is installed. The probe of the laser ranging module (108) is parallel to the support part (3), so that the laser ranging module (108) which plays the role of determining depth is vertically downward aligned with the soil surface.

2. The handheld multi-depth soil parameter detection device for seeding decision-making according to claim 1, characterized in that, The handheld lever (2) includes a right handheld lever (201) and a left handheld lever (202). The right handheld lever (201) and the left handheld lever (202) pass through the preset holes on both sides of the housing (101) of the handheld pivot part (1). The right handheld lever (201) and the left handheld lever (202) are connected and fixed to the housing (101) through a high-strength threaded structure.

3. The handheld multi-depth soil parameter detection device for seeding decision-making according to claim 1, characterized in that, The spectral detection mechanism (7) is set in a single opening on one side of the bottom of the support part (3).

4. The handheld multi-depth soil parameter detection device for seeding decision-making according to claim 3, characterized in that, The spectral detection mechanism (7) includes: a spectral mounting housing (701), a spectral sensor module (702), and an optical lens (703); the front end of the spectral mounting housing (701) is designed with an inner chamfer of 45° facing downward, which can generate a normal component force during the insertion process, forcibly pushing and pressing the soil particles on the contact surface onto the surface of the optical lens (703).

5. A handheld multi-depth soil parameter detection device for seeding decision-making according to claim 1, characterized in that, The electrode mounting module (5) includes: an electrode mounting housing (501) and an electrode plate (502); the electrode plate (502) is mounted in the electrode mounting housing (501); the capacitor information processing module includes: a capacitor processor mounting housing (401) and a capacitor processor (402), the capacitor processor (402) and the electrode plate (502) are connected, the electrode plate (502) is mounted in the capacitor processor mounting housing (401), and an equipotential active shielding layer is provided between the capacitor processor (402) and the capacitor processor mounting housing (401); the electrode plate (502) is connected to the capacitor processor (402).

6. A handheld multi-depth soil parameter detection device for seeding decision-making according to claim 5, characterized in that, The equipotential active shielding layer is 2 mm wider than the capacitor processor (402) in all four directions, and the two are isolated by a 1.6 mm thick FR-4 insulating substrate, which forces the electric field lines at the edge of the capacitor to diverge only towards the soil side.

7. A handheld multi-depth soil parameter detection device for seeding decision-making according to claim 1, characterized in that, The piercing conductivity detection assembly (6) includes a tail conductor (601), a middle insulator (602) and a head conductor (603). The tail conductor (601) and the head conductor (603) are threaded to the middle insulator (602) from the outside and the inside, respectively, and the head conductor (603) is inserted into the soil first.

8. A handheld multi-depth soil parameter detection device for seeding decision-making according to claim 1, characterized in that, The soil firmness prediction model used in the microcontroller identifies and counteracts the interference of porosity changes. The steps for establishing the soil firmness prediction model are as follows: Step 1: Sample preparation and soil information collection: Select soil samples with different textures and prepare soil samples with different moisture content gradients. Use this device to collect four raw data at the same time: spectrum, capacitance, conductivity and pressure. Step 2, Determination of true physicochemical values ​​of soil: The soil's true physicochemical properties were determined by measuring moisture content using the oven-drying method, organic matter content using the potassium dichromate method, and bulk density using the ring sampler method. Step 3: Construct a feature space for multi-source heterogeneous soil parameters: Divide the signal sources affecting the inversion of soil physicochemical properties into static environmental factors and dynamic sensing factors, and construct a high-dimensional input vector; Step 4: Effective contact state gating based on quantization discrimination rules: The microcontroller 106 only marks the current data frame as a "valid sample" and stores it in the buffer when a specific quantization discrimination rule is met; The quantification discrimination rules include the following three: Rule 1, the pressure coupling threshold rule; Rule 2, the penetration stability rule; and Rule 3, the depth-logic verification rule, wherein: Rule 1, the pressure coupling threshold rule, refers to monitoring the real-time readings of pressure sensor 8. Set the effective coupling pressure threshold. ;when If the probe is found to be suspended, not fully compacted, or passing through a large pore, the spectral and capacitance data will have a large air medium error, and the system will automatically discard the data frame. Rule 2, the penetration stability rule, calculates the rate of pressure change per unit time. ;when When the impact threshold is exceeded, the device is determined to be in a period of unsteady deformation due to severe impact or contact with hard objects such as rocks. Data recording is suspended until the rate of change returns to the stable range, so as to eliminate the nonlinear interference of dynamic shear deformation on the measurement of soil dielectric constant. Rule 3, the depth-logic verification rule, is based on the real-time depth obtained from the laser ranging module (108). Only when The change is monotonically increasing and the increment is Within a reasonable step size range Sampling is only triggered when the user retracts the probe to prevent the same depth data from being recorded repeatedly. Step 5: Construct a multidimensional decoupled model of soil parameters based on pressure constraints: Construct a training dataset, dividing the valid data frames filtered in step four into a training set and a validation set; referencing relevant domain standards, set the division ratio to 70% for the training set and 30% for the validation set; the method for determining positive samples is: data points that meet the criteria have their label values ​​determined by the laboratory standard method; the input layer dimension is... Including: principal components of spectral features Normalized capacitance value Pressure value Electrical conductivity and depth ; The quantification discrimination rule is as follows: < ; Step 6: Use the trained model to perform real-time calculations on the area to be tested: The host computer mobile terminal receives the raw data packets uploaded by the microcontroller, inputs them into the decoupled model in real time, and outputs the equivalent standard bulk density and moisture content after pressure correction.

9. A handheld multi-depth soil parameter detection device for seeding decision-making according to claim 8, characterized in that, In step six, the grading criteria are output. The grading criteria are divided into 5 levels: very low <10%, low: 10%-15%, medium: 15%-20%, high: 20%-25%, and very high >25%.

10. A handheld multi-depth soil parameter detection device for seeding decision-making according to claim 8, characterized in that, In step three, static environmental factors This includes geospatial information and preset parameters obtained through a host computer mobile terminal, including: GNSS latitude and longitude coordinates, soil texture type at the current measuring point (user-preset or obtained from cloud maps), and historical basic bulk density range; these factors serve as boundary constraints for the model; dynamic sensing factors. This refers to the high-frequency response signal collected in real time by the device during the soil penetration process, which represents the dynamic sensing factors. It is decomposed into three dimensions of coupled characteristics: electromagnetic response characteristics, mechanical response characteristics, and electrochemical response characteristics. Among them, the electromagnetic response characteristics include those collected by the spectral sensor module (702) Band reflectivity data, and dielectric constant voltage signal output by capacitance information processing module (4). The mechanical response characteristics are the penetration resistance values ​​fed back in real time by the pressure sensor (8). and resistance change rate ; Electrochemical response characteristics are obtained from soil conductivity signals acquired by the piercing conductivity detection component (6). .

Images