A soil organic matter movement detection device and method based on multiple spectral channels

Abstract

The application discloses a kind of equipment and method for detecting soil organic matter movement based on multispectral channel, it is related to precision agriculture sensing technical field, method includes: in the process of equipment movement, utilize darkroom plow type probe to construct semi-closed detection space, simultaneously collect multispectral reflection signal, soil moisture content, conductivity and surface texture image;Reflectance signal is converted into absorbance, extracts texture feature and generates texture correction coefficient, carries out texture compensation and water-salt decoupling correction to absorbance, obtains net organic matter spectral characteristics;Select representative sample to carry out in-situ calibration, calculate environmental drift deviation and gain coefficient, and output soil organic matter content detection result or distribution result.Stable darkroom light environment is realized by streamlined black plow body to realize physical noise reduction, humidity and salt interference are weakened by water-salt double-path decoupling, and the detection accuracy and adaptability under different rough surface scenes are improved by texture adaptive compensation.

Description

Technical Field

[0001] This invention relates to the field of precision agriculture sensing technology, specifically to a mobile detection device and method for soil organic matter based on multispectral channels. Background Technology

[0002] Soil organic matter (SOM), as a core characteristic parameter of soil quality and fertility, plays a fundamental supporting role in crop growth prediction, soil improvement, farmland productivity assessment, crop model construction, and precision fertilization management. Accurately obtaining spatial distribution information of soil organic matter is a key prerequisite for achieving efficient utilization of agricultural resources and targeted cultivation of arable land quality. However, existing soil organic matter detection methods have significant limitations in terms of timeliness and spatial resolution.

[0003] Currently, routine determination of soil organic matter mainly relies on traditional laboratory chemical analysis methods, such as the potassium dichromate oxidation-external heating method or the dry burning method. Although these methods offer high accuracy, they are cumbersome, requiring multiple steps including field sampling, sample packaging and transportation, laboratory drying, grinding and sieving, and chemical titration or high-temperature combustion testing. From sampling to data output, it typically takes several days or even longer, and the testing cost increases dramatically with sample density. This high-delay, high-cost operating model cannot support the need for high-density, gridded, and rapid monitoring at the field scale, leading to the common dilemma of "inaccurate, incomplete, and untimely measurements" in farmland nutrient management, which hinders the implementation of agronomic measures such as variable fertilization and precision irrigation.

[0004] In recent years, to meet the demand for rapid detection, some studies have attempted to utilize novel sensing techniques such as visible-near-infrared (Vis-NIR) spectroscopy, mid-infrared spectroscopy, and laser-induced breakdown spectroscopy (LIBS) to retrieve soil properties. Related studies have shown that soil organic matter exhibits significant spectral response characteristics in specific wavelength bands, providing a physical basis for rapid detection. However, existing spectral detection models generally suffer from poor regional adaptability, sensitivity to changes in soil type and moisture content, and significant interference from environmental factors on retrieval accuracy. Furthermore, technical bottlenecks remain in hardware integration and in-situ measurement capabilities, hindering the achievement of truly portable, low-cost, and robust synergistic optimization. Changes in soil moisture content cause drastic nonlinear fluctuations in spectral absorbance (water has strong absorption in the near-infrared region), leading to the misclassification of "wet soil" as "high-organic-matter soil." In saline-alkali land (such as soda saline-alkali soil in Northeast China), salt crystallization leads to enhanced light scattering and increased reflectivity, masking the spectral characteristics of organic matter. Different plots (black soil, sandy soil, grassland) have varying roughness, affecting the diffuse reflection light path.

[0005] Therefore, there is an urgent need to develop an in-situ rapid soil organic matter detection device that can balance detection accuracy, response speed, and operational flexibility. This device should be able to avoid light interference through physical structure and decouple moisture, salinity, and texture through algorithmic models. This paper proposes a portable detection device suitable for in-situ field environments based on multispectral fusion and machine learning modeling techniques. By optimizing the sensing unit and algorithm architecture, environmental noise is effectively suppressed, and the model's cross-regional generalization ability is improved, providing a feasible technical path for achieving high-throughput, low-cost, and real-time monitoring of soil organic matter. Summary of the Invention

[0006] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a soil organic matter mobile detection device and method based on multispectral channels to solve the above-mentioned technical problems.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a mobile soil organic matter detection device based on multispectral channels, comprising: a mounting bracket, a darkroom plow-shaped probe, a multispectral sensing module, an auxiliary correction module, a processor, a memory, and a power supply module;

[0008] The dark chamber plow-shaped probe is installed at the front end of the mounting bracket and is used to connect with the traction equipment. The front end of the dark chamber plow-shaped probe is provided with a cutting part facing the direction of travel, which is used to cut the topsoil and cooperate with the soil on both sides to form a semi-enclosed dark chamber detection space during the movement of the equipment.

[0009] The multispectral sensing module is embedded in the lower front of the darkroom plow-shaped probe. The multispectral sensing module includes at least three light-emitting units with different center wavelengths and photoelectric receiving units corresponding to each light-emitting unit, used to acquire the reflected light signal of the soil in each multispectral channel.

[0010] The auxiliary correction module includes a contact water and salt sensor located below the multispectral sensing module and a macro vision module located above the plow-shaped probe in the darkroom. The contact water and salt sensor is used to acquire the soil volumetric water content and electrical conductivity at the current detection location, and the macro vision module is used to acquire the soil surface texture image at the current detection location.

[0011] The processor is electrically connected to the multispectral sensing module, the contact water and salt sensor, the macro vision module, and the memory, respectively, and is used to receive the data output by each detection module and output the detection results of soil organic matter content.

[0012] The invention is further configured such that the darkroom plow-shaped probe is made of black light-absorbing material and has a streamlined plow body structure. The front is a trapezoid with a narrow top and a wide bottom, and the front part forms a forward-protruding blade structure to cut the topsoil during mobile operation and place the detection surface in the stable soil trench behind the trencher.

[0013] The present invention is further configured such that the dark chamber plow-shaped probe is rigidly connected to the rear of the agricultural machinery traction arm or furrow opener, and performs continuous in-situ detection at a soil penetration depth of 10cm to 15cm.

[0014] The present invention is further configured such that the three center wavelengths of the multispectral sensing module are located at the visible light red end, the near-infrared short-wave band and the near-infrared long-wave band, respectively, preferably 660nm, 850nm and 940nm; the contact water-salt sensor is an interdigital electrode sensor or an FDR probe; and the macro vision module includes a miniature CMOS camera and an illumination unit.

[0015] This invention also provides a method for detecting soil organic matter movement based on multispectral channels, used to implement the above-mentioned soil organic matter movement detection device based on multispectral channels, comprising:

[0016] S1: During the movement of the equipment, the topsoil is cut by the plow-shaped probe in the dark chamber and it is combined with the soil on both sides to form a semi-enclosed dark chamber detection space. The reflected light signal of the current detection position under each multispectral channel is obtained by the multispectral sensing module. The soil volume moisture content and electrical conductivity of the current detection position are obtained by the contact water and salt sensor. The soil surface texture image of the current detection position is obtained by the macro vision module.

[0017] S2: Convert the reflected light signal of each multispectral channel into the absorbance of the corresponding channel;

[0018] S3: Construct a gray-level co-occurrence matrix based on the soil surface texture image, extract texture entropy and contrast, and generate texture correction coefficients based on the extraction results;

[0019] S4: Based on the texture correction coefficient, the light intensity attenuation caused by soil surface roughness is compensated, and the absorbance is corrected by water-salt decoupling based on volume water content and electrical conductivity to generate net organic matter spectral characteristics.

[0020] S5: Before or during the operation, select representative sample points in the plot to be tested for in-situ calibration, obtain the true value of soil organic matter of the representative sample points, and calculate the environmental drift bias Biasday and gain coefficient Gday based on the net organic matter spectral characteristics of the representative sample points and the true value of soil organic matter.

[0021] S6: Based on the environmental drift bias Biasday, gain coefficient Gday, and preset regression coefficient, the net organic matter spectral characteristics of each test sample point in the test plot are corrected, and the soil organic matter content test results or distribution results are output.

[0022] The present invention is further configured such that, in step S3, multi-directional and multi-scale gray-level co-occurrence analysis is performed on the soil surface texture image at the current detection location to extract texture features, which are used to characterize the degree of disorder and gray-level abrupt change of the soil surface. Based on the texture features, a texture correction coefficient is constructed to synergistically compensate for the diffuse reflection optical path change and light intensity attenuation caused by the difference in soil surface roughness.

[0023] The present invention is further configured such that, in step S4, a water-salt dual-path decoupling correction model is constructed to address the different interference mechanisms of moisture and salinity on multispectral absorbance; wherein, a nonlinear correction path is used to compensate for the absorption enhancement caused by moisture, and a linear correction path is used to compensate for the scattering attenuation caused by salinity, and the two correction paths are linked with the texture correction coefficient to remove the environmental disturbance component from the original absorbance and generate net organic matter spectral features for soil organic matter inversion.

[0024] The present invention is further configured such that, in step S5, representative sample points are selected from the current test site to obtain the measured true value of soil organic matter, and the in-situ calibration relationship of the day is established in combination with the net organic matter spectral characteristics of the representative sample points. The deviation calibration amount Biasday and the gain calibration amount Gday, which reflect the drift characteristics of the current working environment, are calculated. Biasday and Gday are used as common calibration parameters for the current test site to perform unified migration calibration on the inversion results of other test sample points of the same site, so as to suppress the overall drift error under different working periods, different surface conditions and different environmental backgrounds.

[0025] The present invention is further configured such that, after step S4 and before step S5, it also includes:

[0026] The stable state of the current detection point is determined based on the absorbance response relationship of the three characteristic bands after texture compensation and water-salt decoupling correction. Specifically, the absorbance difference between two adjacent characteristic bands is taken as the first and second order differences, respectively. The deviation of the absorbance of any characteristic band relative to the response relationship of the other two characteristic bands is taken as the channel deviation. A stable state is defined as follows: when the first and second order differences have the same sign, their absolute ratio is within a preset stable range, and all channel deviations are within a preset deviation tolerance range; a near-stable state is defined as follows: when the first and second order differences have the same sign, but their absolute ratio exceeds the preset stable range and is within a preset near-stable range, or at least one channel deviation exceeds the preset deviation tolerance range but is within a preset deviation correction range; a near-stable state is defined as follows: when the first and second order differences have different signs, or their absolute ratio exceeds the preset near-stable range, or at least one channel deviation exceeds the preset deviation correction range.

[0027] The present invention is further configured such that when the channel is in a stable state, it enters the soil organic matter inversion process; when the channel is in a near-stable state, the current detection point is enhanced and corrected before being re-judged, and participates in the soil organic matter inversion process when the re-judgment meets the preset conditions; when the channel is in a dissociated state, the current detection point is triggered to undergo re-testing, neighbor point substitution or elimination processing, and the current detection point is not used as a representative sample point for in-situ calibration.

[0028] This invention provides a mobile soil organic matter detection device and method based on multispectral channels. During device movement, a plow-shaped probe in a dark chamber cuts through the topsoil and aligns with soil on both sides to form a semi-enclosed dark chamber detection space. A multispectral sensing module acquires the reflected light signals from each multispectral channel at the current detection location. A contact water-salt sensor acquires the soil volumetric water content and conductivity at the current detection location. A macro vision module acquires the soil surface texture image at the current detection location. The reflected light signals from each multispectral channel are converted into the absorbance of the corresponding channel. A gray-level co-occurrence matrix is ​​constructed based on the soil surface texture image, and texture entropy and contrast are extracted. A texture correction coefficient is generated based on the extraction results. The texture correction coefficient is then used to... The method compensates for light intensity attenuation caused by soil surface roughness, and corrects absorbance for water-salt decoupling based on volumetric water content and electrical conductivity to generate net organic matter spectral characteristics. Representative samples from the test plot are selected for in-situ calibration before or during the operation to obtain the true measured values ​​of soil organic matter at these representative samples. Based on the net organic matter spectral characteristics and the true measured values ​​of soil organic matter at these representative samples, the environmental drift bias (Biasday) and gain coefficient (Gday) for the day are calculated. Based on the environmental drift bias (Biasday), gain coefficient (Gday), and preset regression coefficients, the net organic matter spectral characteristics of each test sample point in the test plot are corrected, and the soil organic matter content detection results or distribution results are output. The beneficial effects include:

[0029] 1. Physical noise reduction: The streamlined black plow body design not only minimizes the resistance to movement, but also creates a stable darkroom light environment, which improves the signal-to-noise ratio from a physical perspective.

[0030] 2. Mathematical decoupling: By separating the moisture and salt components in the spectral signal through a mathematical model, the misjudgment caused by the traditional equipment's "from wet to black" is solved. It is particularly suitable for the Northeast black soil region (high moisture) and the Suaeda salsa region (high salinity).

[0031] 3. High adaptability: Combined with texture correction coefficients, the same set of equipment can automatically adapt to detection scenarios with different surface roughness, such as bare ground to grass.

[0032] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0034] Figure 1 This is a three-dimensional schematic diagram of the overall appearance of a mobile soil organic matter detection device based on a multispectral channel, as shown in an exemplary embodiment of the present invention.

[0035] Figure 2 This is an exemplary embodiment of the present invention, illustrating the in-situ detection photoelectric signal flow and water-salt decoupling link diagram of a multi-spectral channel-based soil organic matter mobile detection device.

[0036] Figure 3 The flowchart of the data processing and water-salt calibration algorithm for a soil organic matter movement detection method based on multispectral channels is shown as an exemplary embodiment of the present invention.

[0037] Figure 4 This is a schematic diagram of the full wavelength and first-order conduction of a soil organic matter movement detection method based on multispectral channels, as shown in an exemplary embodiment of the present invention.

[0038] Figure 5 This is a schematic diagram illustrating the simulated original spectrum, continuous removal of the 2200nm absorption region, first derivative of the red-edge sensitive region, band selection, and regression of a soil organic matter movement detection method based on a multi-spectral channel, as an exemplary embodiment of the present invention.

[0039] Figure 6 A comparison of correction curves showing the effects of moisture and salinity on spectral absorbance in a multi-spectral channel-based soil organic matter movement detection method, as an exemplary embodiment of the present invention.

[0040] Figure 7 This is a preliminary internal hardware design diagram of a mobile soil organic matter detection device based on a multispectral channel, as shown in an exemplary embodiment of the present invention. Detailed Implementation

[0041] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0042] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0043] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0044] Example 1:

[0045] A mobile soil organic matter detection device based on multispectral channels, such as Figure 1 and Figure 7 As shown, it includes: a mounting bracket, an anechoic chamber plow-shaped probe, a multispectral sensing module, an auxiliary correction module, a processor, a memory, and a power supply module;

[0046] The dark chamber plow-shaped probe is installed at the front end of the mounting bracket and is used to connect with the traction equipment. The front end of the dark chamber plow-shaped probe is provided with a cutting part facing the direction of travel, which is used to cut the topsoil and cooperate with the soil on both sides to form a semi-enclosed dark chamber detection space during the movement of the equipment.

[0047] The multispectral sensing module is embedded in the lower front of the darkroom plow-shaped probe. The multispectral sensing module includes at least three light-emitting units with different center wavelengths and photoelectric receiving units corresponding to each light-emitting unit, used to acquire the reflected light signal of the soil in each multispectral channel.

[0048] The auxiliary correction module includes a contact water and salt sensor located below the multispectral sensing module and a macro vision module located above the plow-shaped probe in the darkroom. The contact water and salt sensor is used to acquire the soil volumetric water content and electrical conductivity at the current detection location, and the macro vision module is used to acquire the soil surface texture image at the current detection location.

[0049] The processor is electrically connected to the multispectral sensing module, the contact water and salt sensor, the macro vision module, and the memory, respectively, and is used to receive the data output by each detection module and output the detection results of soil organic matter content.

[0050] Specifically, the dark chamber plow-shaped probe: The entire body is made of black light-absorbing material (such as black ABS, anodized aluminum, or carbon fiber), and is approximately 10cm long. The structure is designed as a streamlined plow body, approximately 2.5cm wide at the front and extremely thin on the sides (0.4-0.6cm), with a trapezoidal shape that is narrower at the top and wider at the bottom. The plow-shaped protrusion (blade) faces forward and moves with the field equipment during operation, cutting through the topsoil so that the detection surface is located within a stable furrow behind the furrow opener. The physical contact between the black body and the soil forms a "semi-enclosed dark chamber," effectively shielding more than 90% of external natural light interference. The dark chamber plow-shaped probe is rigidly connected to the agricultural machinery's traction arm or the rear of the furrow opener, and performs continuous in-situ detection at a depth of 10cm to 15cm.

[0051] Multispectral sensing module: Embedded in the lower front of the probe, it includes LED light sources (λ1, λ2, λ3) with three characteristic wavelength bands and corresponding photodiodes. The wavelength band selection is based on SOM sensitivity screening in the previous modeling (e.g., selecting the visible red end, near-infrared short wave, and near-infrared long wave). The three center wavelengths of the multispectral sensing module are located at the visible red end, near-infrared short wave band, and near-infrared long wave band, respectively, preferably 660nm, 850nm, and 940nm. The contact water-salt sensor is an interdigitated electrode sensor or an FDR probe. The macro vision module includes a miniature CMOS camera and an illumination unit.

[0052] The auxiliary correction module includes a contact water-salt sensor and a macro vision module. The contact water-salt sensor is located below the spectral window and uses interdigitated electrodes or an FDR probe to measure soil volumetric water content (θ) and electrical conductivity (EC) in real time. The macro vision module integrates a miniature CMOS camera located on the upper part of the probe and includes a miniature camera and an illumination lamp for acquiring images of soil micro-texture.

[0053] Example 2:

[0054] This exemplary method for detecting soil organic matter movement based on multispectral channels, used to implement the aforementioned soil organic matter movement detection device based on multispectral channels, includes:

[0055] S1: During the movement of the equipment, the topsoil is cut by the plow-shaped probe in the dark chamber and it is combined with the soil on both sides to form a semi-enclosed dark chamber detection space. The reflected light signal of the current detection position under each multispectral channel is obtained by the multispectral sensing module. The soil volume moisture content and electrical conductivity of the current detection position are obtained by the contact water and salt sensor. The soil surface texture image of the current detection position is obtained by the macro vision module.

[0056] S2: Convert the reflected light signal of each multispectral channel into the absorbance of the corresponding channel;

[0057] S3: Construct a gray-level co-occurrence matrix based on the soil surface texture image, extract texture entropy and contrast, and generate texture correction coefficients based on the extraction results;

[0058] S4: Based on the texture correction coefficient, the light intensity attenuation caused by soil surface roughness is compensated, and the absorbance is corrected by water-salt decoupling based on volume water content and electrical conductivity to generate net organic matter spectral characteristics.

[0059] S5: Before or during the operation, select representative sample points in the plot to be tested for in-situ calibration, obtain the true value of soil organic matter of the representative sample points, and calculate the environmental drift bias Biasday and gain coefficient Gday based on the net organic matter spectral characteristics of the representative sample points and the true value of soil organic matter.

[0060] S6: Based on the environmental drift bias Biasday, gain coefficient Gday, and preset regression coefficient, the net organic matter spectral characteristics of each test sample point in the test plot are corrected, and the soil organic matter content test results or distribution results are output.

[0061] The present invention is further configured such that, in step S3, multi-directional and multi-scale gray-level co-occurrence analysis is performed on the soil surface texture image at the current detection location to extract texture features, which are used to characterize the degree of disorder and gray-level abrupt change of the soil surface. Based on the texture features, a texture correction coefficient is constructed to synergistically compensate for the diffuse reflection optical path change and light intensity attenuation caused by the difference in soil surface roughness.

[0062] The present invention is further configured such that, in step S4, a water-salt dual-path decoupling correction model is constructed to address the different interference mechanisms of moisture and salinity on multispectral absorbance; wherein, a nonlinear correction path is used to compensate for the absorption enhancement caused by moisture, and a linear correction path is used to compensate for the scattering attenuation caused by salinity, and the two correction paths are linked with the texture correction coefficient to remove the environmental disturbance component from the original absorbance and generate net organic matter spectral features for soil organic matter inversion.

[0063] The present invention is further configured such that, in step S5, representative sample points are selected from the current test site to obtain the measured true value of soil organic matter, and the in-situ calibration relationship of the day is established in combination with the net organic matter spectral characteristics of the representative sample points. The deviation calibration amount Biasday and the gain calibration amount Gday, which reflect the drift characteristics of the current working environment, are calculated. Biasday and Gday are used as common calibration parameters for the current test site to perform unified migration calibration on the inversion results of other test sample points of the same site, so as to suppress the overall drift error under different working periods, different surface conditions and different environmental backgrounds.

[0064] The present invention is further configured such that, after step S4 and before step S5, it also includes:

[0065] The stable state of the current detection point is determined based on the absorbance response relationship of the three characteristic bands after texture compensation and water-salt decoupling correction. Specifically, the absorbance difference between two adjacent characteristic bands is taken as the first and second order differences, respectively. The deviation of the absorbance of any characteristic band relative to the response relationship of the other two characteristic bands is taken as the channel deviation. A stable state is defined as follows: when the first and second order differences have the same sign, their absolute ratio is within a preset stable range, and all channel deviations are within a preset deviation tolerance range; a near-stable state is defined as follows: when the first and second order differences have the same sign, but their absolute ratio exceeds the preset stable range and is within a preset near-stable range, or at least one channel deviation exceeds the preset deviation tolerance range but is within a preset deviation correction range; a near-stable state is defined as follows: when the first and second order differences have different signs, or their absolute ratio exceeds the preset near-stable range, or at least one channel deviation exceeds the preset deviation correction range.

[0066] The present invention is further configured such that when the channel is in a stable state, it enters the soil organic matter inversion process; when the channel is in a near-stable state, the current detection point is enhanced and corrected before being re-judged, and participates in the soil organic matter inversion process when the re-judgment meets the preset conditions; when the channel is in a dissociated state, the current detection point is triggered to undergo re-testing, neighbor point substitution or elimination processing, and the current detection point is not used as a representative sample point for in-situ calibration.

[0067] Specifically, a multispectral basic model is constructed: based on the Beer-Lambert Law, the reflectance intensity R of the three collected bands is... Converted to absorbance A ( ): In the formula: Indicates at wavelength The absorbance at a given wavelength is a dimensionless parameter used to characterize the degree to which a soil sample absorbs light of that wavelength. represents the incident light intensity emitted by the light source, with the unit of milliwatt (mW), which is controlled by the LED light source drive circuit and outputs stably; represents at the wavelength the reflected light intensity received by the photodiode after being reflected by the soil sample, with the unit of milliwatt (mW); represents at the wavelength the reflectance, which is a dimensionless parameter, and = / , and its value range is from 0 to 1; represents the central wavelength of the i-th spectral channel, with the unit of nanometer (nm), and i = 1, 2, 3 represent three characteristic wavelength bands.

[0068] Soil texture feature extraction: Use a camera to collect images, calculate the gray-level co-occurrence matrix (GLCM) of the images, extract the texture entropy (Entropy) and contrast (Contrast), and calculate the texture correction coefficient ωtex: , where α, β, γ are preset weights of image features.

[0069] The gray-level co-occurrence matrix is a statistical tool for describing the spatial distribution relationship of pixel gray values in an image. Let the soil surface image be , and its gray level is L. Then the gray-level co-occurrence matrix in the direction θ and at the distance d is defined as: , where: i, j are gray levels, and the value range is 0 ≤ i, j < L; d is the pixel spacing, and in this embodiment, d ∈ {1, 2, 4, 8}; θ is the direction angle, and in this embodiment, θ ∈ {0°, 45°, 90°, 135°}; #{·} represents the number of pixel pairs that meet the conditions; D is the image domain.

[0070] The texture entropy is used to measure the randomness of the gray distribution of an image and reflects the complexity of the soil surface. The calculation formula is: , where: ε is a small constant to avoid zero values in logarithmic operations, and in this embodiment, ε = 10 -10 .

[0071] The value range of the texture entropy is [0, 2log2L]. The larger the entropy value, the more complex and irregular the soil surface texture; the smaller the entropy value, the more uniform and smooth the soil surface.

[0072] The texture entropy and contrast of 16 gray-level co-occurrence matrices in 4 directions and 4 distances can be calculated respectively, and their arithmetic mean is taken as the final texture feature value: , .

[0073] Establish a water-salt decoupling correction model: Due to the nonlinear interference of water and salt on the spectral signal, the following correction equation is constructed to obtain the "net organic matter spectral characteristics" ASOM′(λi): The specific correction function is defined as follows: Moisture correction term: Polynomial fitting is used to eliminate the absorption gain caused by moisture: Salt correction term: Linear fitting is used to eliminate scattering attenuation caused by salt. Where θ is the real-time measured water content, EC is the real-time electrical conductivity, and kw1, kw2, ks are the laboratory calibration coefficients for the corresponding wavebands.

[0074] In-situ rapid calibration in the field: Before operation, a small number of samples (N≥3) of the plots to be tested are selected for field measurement to obtain the true value set {SOMreal, θreal, ECreal}, which serves as the baseline data for the prediction model and the parameters for in-situ rapid calibration. The environmental drift bias Biasday and gain coefficient Gday for the day are calculated using the least squares method. Based on the drift bias Biasday and gain coefficient Gday, bias and gain corrections are applied to the spectral signals generated by the measured plots to generate the final prediction model. In the formula: Ci represents the regression coefficients of each band on organic matter preset in the laboratory; SOMpred represents the final output organic matter content. A bias model is established using the least squares method to calculate... : In the formula: n is the number of standard samples (n≥3); j is the sample number; The regression coefficients for each band are preset for the laboratory; This is the absorbance after water and salt correction. A gain model is established using the least squares method to calculate... : ,in: , which is the predicted value (before correction).

[0075] Furthermore, after the processor completes texture compensation and water-salt decoupling correction for the current sample point, it obtains the net organic matter absorbance response values ​​corresponding to the three characteristic bands. , and Based on this, the processor further determines the channel steady state at the current detection point according to the absorbance response relationship of the three characteristic bands. Specifically, it calculates the absorbance step difference between two adjacent characteristic bands, and then... and The difference in absorbance between them is taken as the first-order difference. and The difference in absorbance between the three bands is used as the second-order difference; at the same time, the overall response relationship of the three characteristic bands is combined to determine whether there is an abnormal deviation of a single band relative to the other two bands.

[0076] When the first and second order differences are aligned in direction, the amplitude relationship between the two order differences is within a preset stable range, and none of the three characteristic bands exhibit abnormal deviations exceeding the preset deviation tolerance range, the channel stable state of the current detection point is determined to be a stable state. This state indicates that after texture compensation and water-salt decoupling correction, the three characteristic bands maintain a consistent organic matter spectral response to the sample point, and can stably characterize the net organic matter spectral features of that point.

[0077] When the first and second order differences are in the same direction, but the amplitude relationship between the two order differences exceeds the preset stable range and falls into the preset near-stable range, or when the deviation of at least one of the three characteristic bands exceeds the preset deviation tolerance range but is still within the preset deviation correction range, the channel stable state of the current detection point is determined to be a near-stable state. This state indicates that the overall response of the three characteristic bands of the current sample point still changes in the same direction, but there is a certain degree of residual disturbance locally, which still has value for further correction and re-judgment.

[0078] When the directions of the first and second order differences are inconsistent, or the amplitude relationship between the two order differences exceeds the preset near-coincidence range, or the deviation of at least one of the three characteristic bands exceeds the preset deviation correction range, the channel steady-state of the current detection point is determined to be a decoupled state. This state indicates that the current sample point has a significant response mismatch among the three characteristic bands, making it difficult to directly use as a stable net organic matter spectral feature input for subsequent inversion and calibration processes.

[0079] After determining the stable state of a channel, the processor performs differentiated follow-up processing based on the different stable states of the channels to which the current detection point belongs.

[0080] When the current detection point is determined to be in a stable state, the processor directly uses the net organic matter spectral characteristics corresponding to the detection point as valid input, inputs them into the field in-situ rapid calibration and global mapping process, and uses them to generate the soil organic matter prediction results for the detection point. If the detection point also meets the representative sample selection criteria, it can also be used as a representative sample point to participate in the calculation of Biasday and Gday.

[0081] When a current detection point is determined to be in a near-closely bound state, the processor does not immediately remove the detection point. Instead, it performs enhancement corrections before re-evaluation. These enhancement corrections may include: re-invoking the texture correction results corresponding to the current detection point and readjusting the texture compensation intensity; or re-invoking the moisture content and conductivity data corresponding to the current detection point and updating the water-salt decoupling correction results again. After re-evaluation, if the detection point meets the stable bound state determination criteria, it is considered a valid detection point; if it remains in a near-closely bound state after re-evaluation, it is allowed to participate in soil organic matter inversion or plot distribution result generation, but it is not given priority as a representative sample point for in-situ calibration of Biasday and Gday.

[0082] When the current detection point is determined to be in a disjoint state, the processor triggers an abnormal sample point processing procedure. This procedure includes retesting the current detection point, replacing it with data from neighboring detection points, or directly marking the current detection point as invalid and removing it from the current round of inversion. Simultaneously, disjoint state detection points are not included in the construction of representative samples for Biasday and Gday. This approach avoids interference from abnormal samples with significant channel response mismatches on the daily in-situ rapid calibration results.

[0083] In one feasible embodiment of the present invention, the detection of spring plowing operations in the black soil region of Northeast China is taken as an example:

[0084] In the Northeast black soil region, the soil undergoes freeze-thaw cycles during spring plowing, exhibiting typical high moisture content (fluctuating between 20% and 35%), and some plots also show signs of soda salinization. Traditional spectroscopic detection equipment easily misinterprets the strong light absorption caused by high moisture content as high organic matter content. In this embodiment, the equipment is mounted behind the tractor's furrow opener to conduct continuous in-situ detection on a typical black soil plot, setting up a gridded sampling point with a total of 300 points. The specific detection methods and steps are as follows:

[0085] Equipment installation and darkroom environment construction: such as Figure 1 As shown, the darkroom plow-shaped probe is rigidly connected to the rear of the agricultural machinery's traction arm, with an insertion depth set at 10-15 cm. When the agricultural machinery travels at a speed of 3-5 km / h, the black streamlined plow body cuts through the soil and closely adheres to the soil on both sides, forming an underground semi-enclosed darkroom structure that physically shields against external natural light interference. During the journey, the equipment continuously passes over the 300 test points.

[0086] Multimodal data synchronous acquisition and basic spectral conversion: such as Figure 2 As shown, when the probe passes the k-th sample point (k=1,2...300), the multimodal sensing module on the front of the probe is triggered, emitting a light source of a specific wavelength (e.g., 660nm, 850nm, 940nm), and the photodetector acquires the reflectance R(λi). The initial absorbance A(λi) is calculated according to the Beer-Lambert law, as follows. Figure 6 As shown in the “original absorbance curve”, due to the high moisture content of black soil in spring, the OH bonds of water molecules generate strong absorption, resulting in an abnormal overall rise in the baseline of the original absorbance curve.

[0087] Soil texture feature extraction: Simultaneously, the macro vision module acquires the soil texture image of the current sample point, calculates the gray-level co-occurrence matrix (GLCM) of the image, and extracts the texture entropy and contrast parameters. Addressing the loss of strong light diffuse reflection caused by muddy, blocky textures common in spring-cultivated black soil, the system automatically calculates the texture correction coefficient ωtex for the current sample point according to a formula to compensate for the light intensity attenuation caused by surface roughness.

[0088] Water-salt signal decoupling and mathematical noise reduction: such as Figure 3 As shown, the contact-type water-salt sensor below the probe simultaneously measures the volumetric water content θ and conductivity EC at that point. In this embodiment, it is assumed that θ = 28% and EC = 0.8 mS / cm are measured at a certain sampling point. The processor substitutes θ and EC into a preset water-salt decoupling correction model:

[0089]

[0090] like Figure 4-5 As shown, in the above calculations, the nonlinear quadratic polynomial of water effectively removes the water absorption gain, and the linear salt correction term removes the light scattering attenuation caused by surface salt crystallization. The ASOM′(λi) obtained after correction is the net organic matter spectral characteristic after eliminating water and salt interference (e.g., Figure 4 (As shown in the "Net Organic Matter Absorbance Curve").

[0091] Rapid in-situ calibration and global mapping in the field: Before or during continuous operation, a small number of representative sample points (preferably N=5 sample points in this embodiment) are selected from the 300 sample points to obtain the measured true values ​​SOMpred. Using the true values ​​of these 5 points, combined with ASOM′(λi) and ωtex extracted by the device, the environmental drift deviation Biasday and gain coefficient Gday for the day are calculated using the least squares method. Subsequently, the system automatically maps this calibration coefficient to the remaining 295 moving detection sample points, outputting the high-precision organic matter content SOMpred for the entire plot.

[0092] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A mobile soil organic matter detection device based on multispectral channels, characterized in that, include: Mounting bracket, anechoic chamber plow-shaped probe, multispectral sensing module, auxiliary correction module, processor, memory and power supply module; The dark chamber plow-shaped probe is installed at the front end of the mounting bracket and is used to connect with the traction equipment. The front end of the dark chamber plow-shaped probe is provided with a cutting part facing the direction of travel, which is used to cut the topsoil and cooperate with the soil on both sides to form a semi-enclosed dark chamber detection space during the movement of the equipment. The multispectral sensing module is embedded in the lower front of the darkroom plow-shaped probe. The multispectral sensing module includes at least three light-emitting units with different center wavelengths and photoelectric receiving units corresponding to each light-emitting unit, used to acquire the reflected light signal of the soil in each multispectral channel. The auxiliary correction module includes a contact water and salt sensor located below the multispectral sensing module and a macro vision module located above the plow-shaped probe in the darkroom. The contact water and salt sensor is used to acquire the soil volumetric water content and electrical conductivity at the current detection location, and the macro vision module is used to acquire the soil surface texture image at the current detection location. The processor is electrically connected to the multispectral sensing module, the contact water and salt sensor, the macro vision module, and the memory, respectively, and is used to receive the data output by each detection module and output the detection results of soil organic matter content.

2. The soil organic matter mobile detection device based on multispectral channels according to claim 1, characterized in that, The darkroom plow-shaped probe is made of black light-absorbing material and has a streamlined plow body structure. The front is a trapezoid that is narrow at the top and wide at the bottom, with a forward-protruding blade structure at the front to cut the topsoil during mobile operation and position the detection surface in the stable soil trench behind the trencher.

3. The soil organic matter mobile detection device based on multispectral channels according to claim 1, characterized in that, The darkroom plow-shaped probe is rigidly connected to the rear of the agricultural machinery traction arm or furrow opener, and performs continuous in-situ detection at a soil penetration depth of 10cm to 15cm.

4. The soil organic matter mobile detection device based on multispectral channels according to claim 1, characterized in that, The three center wavelengths of the multispectral sensing module are located at the visible red end, the near-infrared short band, and the near-infrared long band, respectively, preferably 660nm, 850nm, and 940nm; the contact water-salt sensor is an interdigital electrode sensor or an FDR probe; and the macro vision module includes a miniature CMOS camera and an illumination unit.

5. A method for detecting soil organic matter movement based on multispectral channels, used to implement the soil organic matter movement detection device based on multispectral channels as described in any one of claims 1-4, characterized in that, include: S1: During the movement of the equipment, the topsoil is cut by the plow-shaped probe in the dark chamber and it is combined with the soil on both sides to form a semi-enclosed dark chamber detection space. The reflected light signal of the current detection position under each multispectral channel is obtained by the multispectral sensing module. The soil volume moisture content and electrical conductivity of the current detection position are obtained by the contact water and salt sensor. The soil surface texture image of the current detection position is obtained by the macro vision module. S2: Convert the reflected light signal of each multispectral channel into the absorbance of the corresponding channel; S3: Construct a gray-level co-occurrence matrix based on the soil surface texture image, extract texture entropy and contrast, and generate texture correction coefficients based on the extraction results; S4: Based on the texture correction coefficient, the light intensity attenuation caused by soil surface roughness is compensated, and the absorbance is corrected by water-salt decoupling based on volume water content and electrical conductivity to generate net organic matter spectral characteristics. S5: Before or during the operation, select representative sample points in the plot to be tested for in-situ calibration, obtain the true value of soil organic matter of the representative sample points, and calculate the environmental drift bias Biasday and gain coefficient Gday based on the net organic matter spectral characteristics of the representative sample points and the true value of soil organic matter. S6: Based on the environmental drift bias Biasday, gain coefficient Gday, and preset regression coefficient, the net organic matter spectral characteristics of each test sample point in the test plot are corrected, and the soil organic matter content test results or distribution results are output.

6. The method for detecting soil organic matter movement based on multispectral channels according to claim 5, characterized in that, In step S3, multi-directional and multi-scale gray-level co-occurrence analysis is performed on the soil surface texture image at the current detection location to extract texture features, which are used to characterize the degree of disorder and gray-level abrupt change of the soil surface. Based on the texture features, a texture correction coefficient is constructed to collaboratively compensate for the changes in diffuse reflection optical path and light intensity attenuation caused by the difference in soil surface roughness.

7. The method for detecting soil organic matter movement based on multispectral channels according to claim 5, characterized in that, In step S4, a water-salt dual-path decoupling correction model is constructed to address the different interference mechanisms of moisture and salinity on multispectral absorbance. Specifically, a nonlinear correction path is used to compensate for the absorption enhancement caused by moisture, and a linear correction path is used to compensate for the scattering attenuation caused by salinity. The two correction paths are linked with the texture correction coefficient to remove the environmental disturbance component from the original absorbance and generate net organic matter spectral features for soil organic matter inversion.

8. The method for detecting soil organic matter movement based on multispectral channels according to claim 5, characterized in that, In step S5, representative sample points are selected from the current test site to obtain the measured true value of soil organic matter. The in-situ calibration relationship for the day is established in combination with the net organic matter spectral characteristics of the representative sample points. The bias calibration amount Biasday and gain calibration amount Gday, which reflect the drift characteristics of the current working environment, are calculated. Biasday and Gday are used as common calibration parameters for the current test site to perform unified migration calibration on the inversion results of other test sample points of the same site, so as to suppress the overall drift error under different working periods, different surface conditions and different environmental backgrounds.

9. The method for detecting soil organic matter movement based on multispectral channels according to claim 5, characterized in that, After step S4 and before step S5, the following is also included: The stable state of the current detection point is determined based on the absorbance response relationship of the three characteristic bands after texture compensation and water-salt decoupling correction. Specifically, the absorbance difference between two adjacent characteristic bands is taken as the first and second order differences, respectively. The deviation of the absorbance of any characteristic band relative to the response relationship of the other two characteristic bands is taken as the channel deviation. A stable state is defined as follows: when the first and second order differences have the same sign, their absolute ratio is within a preset stable range, and all channel deviations are within a preset deviation tolerance range; a near-stable state is defined as follows: when the first and second order differences have the same sign, but their absolute ratio exceeds the preset stable range and is within a preset near-stable range, or at least one channel deviation exceeds the preset deviation tolerance range but is within a preset deviation correction range; a near-stable state is defined as follows: when the first and second order differences have different signs, or their absolute ratio exceeds the preset near-stable range, or at least one channel deviation exceeds the preset deviation correction range.

10. The method for detecting soil organic matter movement based on multispectral channels according to claim 9, characterized in that, When the channel is in a stable state, it enters the soil organic matter inversion process; when the channel is in a near-stable state, the current detection point is enhanced and corrected before being re-evaluated, and participates in the soil organic matter inversion process when the re-evaluation meets the preset conditions; when the channel is in a dissociated state, the current detection point is triggered to undergo re-testing, neighbor point substitution or elimination, and the current detection point is not used as a representative sample point for in-situ calibration.

Images