Soil component detection method for camellia oleifera nursery

Abstract

The invention relates to the technical field of near-infrared soil detection, in particular to a method for detecting soil components of a camellia oleifera nursery. The method comprises the following steps: obtaining a soil sample, drying, and screening to obtain soil particles with different particle sizes; carrying out spectral analysis to determine an absorbance spectral curve and near-infrared characteristic peak regions of different hydrogen-containing groups, and taking the highest absorbance of the peak regions as a group peak value; determining a change amplitude sequence according to the change of the group peak value; determining a reference peak value according to the absorbance range, and determining an initial sensitivity degree according to the difference between different group peak values and the reference peak value in the variation amplitude sequence; according to the particle size and the water content difference of the soil sample, correcting the initial sensitivity degree to obtain a corrected sensitivity degree; and obtaining a corrected spectrum curve according to the corrected sensitivity degree, performing neural network analysis according to the corrected spectrum curve, and outputting soil component data. The method can effectively realize the influence analysis of different particle sizes on spectral data, and improves the accuracy of soil component analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of near-infrared soil detection technology, and particularly relates to a soil component detection method for an oil tea nursery. BACKGROUND

[0002] The geological conditions and climate conditions of different regions are significantly different, and the selection of suitable tree species and afforestation technology directly affects the afforestation survival rate. Compared with the traditional afforestation technology, the light substrate net bag container seedling afforestation technology has outstanding advantages, and the soil component detection is extremely important for the light substrate net bag container seedling afforestation technology. The mortality rate of the net bag container seedlings without detection can reach 45% (only 15% for traditional ground cultivation), and 70% of the death causes are due to substrate problems. Among them, the detection of hydrogen-containing groups (O-H water, affecting root anoxic root rot and container drainage, N-H nitrogen, affecting oil tea nursery ammonia poisoning and new shoot death, C-H carbon, plant toxicity release due to unrotted, S-H sulfur, root enzyme inactivation and growth stagnation due to mercaptan pollution) is more important, and the hydrogen-containing groups belong to near-infrared (NIR) spectrum (750-2500nm).

[0003] In the process of obtaining the hydrogen-containing group component, the near-infrared spectrum is a prediction model establishment process based on spectral data and chemometrics methods. In the prior art, by collecting a large amount of spectral data of samples with known properties, using a suitable modeling algorithm, a correlation model between spectral data and sample properties can be established. In this way, due to the different particle sizes of the soil, the particles of different components have an imaging influence on the near-infrared spectrum of the soil, which causes the soil particles of different particle sizes to have an influence on the spectral data, resulting in errors in the soil data, causing the established model to be abnormal, and further causing inaccurate soil component detection. SUMMARY

[0004] In order to solve the technical problem that different component particles have an imaging influence on the near-infrared spectrum of the soil, the established model is abnormal, and further causes inaccurate soil component detection, the present application provides a soil component detection method for an oil tea nursery, and the technical scheme adopted is as follows: The present application provides a soil component detection method for an oil tea nursery, and the method comprises the following steps: Obtain soil samples at different sampling times after pruning, dry, grind and screen to obtain soil particles of different particle sizes; perform spectral analysis on the soil particles, determine the absorbance spectrum curve of the near-infrared diffuse reflectance spectrum, determine the near-infrared characteristic peak value region of different hydrogen-containing groups in the absorbance spectrum curve as the peak value region of the corresponding hydrogen-containing group, and take the highest absorbance of the peak value region as the peak value of the corresponding group; Under the same pruning, the change amplitude sequence of each group peak in the spectral curve of different particle sizes is determined based on the change of group peak at different sampling moments; the reference peak is determined based on the absorbance range of the group peak at all sampling moments, and the initial sensitivity of the spectral performance at each particle size is determined based on the difference between the different group peaks and the reference peak in the change amplitude sequence; According to the particle size and the difference in water content of soil samples after different pruning times, the initial sensitivity is corrected to obtain the corrected sensitivity; According to the correction sensitivity, the absorbance spectrum curves of different particle sizes are corrected and analyzed to obtain a corrected spectrum curve. A neural network analysis is performed based on the corrected spectrum curve to output soil composition data.

[0005] Furthermore, the hydrogen-containing groups include methyl, water, primary amine and thiol pollutants.

[0006] Furthermore, the step of determining the change amplitude sequence of each group peak value in the spectral curves of different particle sizes based on the change of the group peak value at different sampling moments includes: The difference between the peak values ​​of the same peak area at the next sampling moment and the previous sampling moment is taken as the change amplitude between the two corresponding sampling moments; All the change amplitudes in the same peak area are combined into a change amplitude sequence according to the time sequence.

[0007] Furthermore, the determination of the reference peak value according to the absorbance extreme difference of the group peak value at all sampling moments includes: Calculate the absorbance range of the same group peak at all sampling times, and take the group peak corresponding to the median of the absorbance range as the reference peak.

[0008] Furthermore, the initial sensitivity of the spectrum performance at each particle size is determined based on the difference between the peak values ​​of different groups and the reference peak value in the variation amplitude sequence, including: The response degree of the peak values ​​of different groups is determined based on the similarity between the peak values ​​of different groups and the reference peak value in the change amplitude sequence; Calculate the absolute value of the difference between the group peak and the reference peak at the same position in the variation amplitude sequence as the element difference; and take the sum of the element differences at all positions as the element reference difference; Calculate the product of the response level and the element reference difference as the sensitivity coefficient of the corresponding group peak; The average of the sensitivity coefficients of all group peaks is taken as the initial sensitivity of the spectrum at the corresponding particle size.

[0009] Furthermore, the response degree of the peak values ​​of different groups is determined based on the similarity between the peak values ​​of different groups and the reference peak value in the change amplitude sequence, including: Calculate the Pearson correlation coefficient between the peak value of any group and the baseline peak value in the change amplitude series; The absolute value of the Pearson correlation coefficient was used as the response level.

[0010] Furthermore, the initial sensitivity is corrected according to the particle size and the difference in water content of soil samples after different pruning times to obtain the corrected sensitivity, including: Determine the correction factor based on the particle size and the moisture content of the soil samples after different pruning times; The product of the correction coefficient and the initial sensitivity is calculated as the corrected sensitivity.

[0011] Furthermore, the correction coefficient is determined based on the particle size and the moisture content of the soil samples after different pruning times, including: Determine the particle size weight according to the mesh number in the soil particle screening process, wherein the higher the mesh number value, the smaller the particle size weight, and the value of the particle size weight is a normalized value; Calculate the mean moisture content of soil samples after different pruning times, and use the ratio of the moisture content of each pruning to the mean moisture content as the moisture content weight; The product of the particle size weight and the water content weight is used as the correction coefficient.

[0012] Furthermore, the correction analysis of the absorbance spectrum curves of different particle sizes is performed according to the correction sensitivity to obtain the corrected spectrum curve, including: The absorbances at different wavelengths in the absorbance spectrum curves at different particle sizes are weighted based on the corrected sensitivity to obtain a corrected spectrum curve.

[0013] Furthermore, the performing of neural network analysis based on the modified spectral curve and outputting soil composition data includes: The corrected spectral curve is used as the input spectral wavelength of nonlinear fitting, and modeling is performed based on the artificial neural network ANN. Through the nonlinear activation of the hidden layer, the output is output from the output layer, and the output result is the soil composition data, where the activation function is the ReLU function.

[0014] The present invention has the following beneficial effects: In an embodiment of the present invention, a benchmark analysis is performed on the amplitude of change in absorbance, and the initial sensitivity of the spectral performance at each particle size is determined according to the difference between the peak values ​​of different groups and the benchmark peak values ​​in the amplitude change sequence, so that the change analysis can be accurately realized. Then, the initial sensitivity is corrected by objective factors such as the particle size and the difference in water content of soil samples after different pruning times, so that the actual spectral deviation can be accurately characterized. The deviation is adjusted to obtain a corrected spectral curve, and a neural network analysis is performed based on the corrected spectral curve to output more accurate and reliable soil composition data. In summary, the present invention can effectively realize the analysis of the impact of soil particles of different particle sizes on spectral data, reduce the error of soil data, and establish a more accurate and objective soil composition analysis model, thereby improving the accuracy of the output soil composition data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A flow chart of a method for detecting soil composition in a tea oil nursery provided by one embodiment of the present invention; Figure 2 A schematic diagram of sampling point distribution provided by one embodiment of the present invention; Figure 3 A schematic diagram of diffuse reflectance spectroscopy analysis provided by one embodiment of the present invention; Figure 4 A schematic diagram of an absorbance spectrum curve provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0017] To further illustrate the technical means and efficacy employed by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a method for detecting soil composition in a tea oil nursery according to the present invention, including its specific implementation, structure, features, and efficacy. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The specific scheme of the soil composition detection method for a tea oil nursery provided by the present invention is described in detail below with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flow chart of a soil composition detection method for a tea oil nursery provided by one embodiment of the present invention, the method comprising: S101: Obtain soil samples at different sampling times after pruning, grind and screen after drying to obtain soil particles of different particle sizes; perform spectral analysis on the soil particles, determine the absorbance spectrum curve of the near-infrared diffuse reflectance spectrum, determine the near-infrared characteristic peak area of ​​the absorbance spectrum curve of different hydrogen-containing groups as the peak area corresponding to the hydrogen-containing group, and take the highest absorbance in the peak area as the peak value of the corresponding group.

[0021] Geological and climatic conditions vary significantly across regions, and the selection of appropriate tree species and planting techniques directly impacts plant survival rates. Compared to traditional planting techniques, lightweight matrix net bag container seedling cultivation offers significant advantages, including higher seedling quality and sufficient nutrient and water content within the container matrix. This significantly enhances tree species' resilience to stress and has been widely adopted and implemented across a wide range of regions.

[0022] Light-substrate net bag container seedling cultivation technology places great emphasis on soil composition testing. The mortality rate of untested net bag container seedlings can reach 45% (compared to only 15% for traditional ground planting), with 70% of this mortality attributed to substrate defects. Of particular importance is the detection of hydrogen-containing groups (OH water, which affects root hypoxia and rot and container drainage; NH nitrogen, which affects ammonia poisoning and new shoot dieback in oil tea seedlings; CH carbon, which can potentially release phytotoxic substances due to lack of composting; and SH sulfur, which causes root enzyme inactivation and growth stagnation due to thiol contamination). Near-infrared spectroscopy is a predictive modeling process based on spectral data and chemometric methods. By collecting spectral data from a large number of samples with known properties and using appropriate modeling algorithms, a correlation model can be established between the spectral data and the sample properties.

[0023] In the embodiment of the present invention, it is first necessary to collect soil samples from the oil tea nursery, that is, soil samples after each pruning, usually collected 15 days after pruning, and tested every two days; and at the same time, the water content of each time is analyzed, and a drying treatment is performed, such as placing the soil sample in a 40-degree constant temperature box for drying, which are all existing soil testing steps.

[0024] Among them, multiple sampling points can be set up in the area, and part of the soil samples can be collected at each sampling point, and then integrated into the soil samples of the oil tea nursery. Figure 2 , Figure 2 A schematic diagram of sampling point distribution provided by one embodiment of the present invention.

[0025] Specifically, under normal circumstances, the oil tea nursery has a weak absorption intensity of hydrogen-containing groups in the soil, and the oil tea seedlings are tree species with obvious apical dominance. If the three-year-old seedlings are not pruned, they are likely to grow into a "broom shape", resulting in a weak trunk and messy side branches, resulting in low light energy utilization. Dense branches are prone to form a high-humidity microenvironment, inducing anthracnose and sooty disease. By cutting off overgrown branches (length > 50 cm) and weak branches, nutrients are concentrated on the fruiting branches, and the production period is advanced by 1-2 years to avoid affecting the later yield.

[0026] After pruning, the tea oil seedlings need to sprout new branches and reorganize their structures, which will affect the changes in various hydrogen-containing groups. For example, due to the reduction of side branches, transpiration decreases, so the OH bond decreases in the short term, and when the new shoot protein is resynthesized, the NH bond increases, and the cellulose deposition of the new branch becomes lignification, which requires structural reorganization, causing the CH bond to decrease. The wound needs to be prevented from diseases and pests, and the synthesis of sulfur-containing defense substances (such as glutathione) increases, resulting in an increase in the absorption of SH bonds. Correspondingly, the increase and decrease in the plant's needs will cause a decrease or increase in the content of various hydrogen-containing groups in the soil.

[0027] Therefore, the change in the content of hydrogen-containing groups in the soil after pruning is more obvious than that during the normal growth period. Therefore, this application uses the soil changes in the period after pruning to construct a model.

[0028] In an embodiment of the present invention, the soil sample can be ground in advance and then screened using sieves of different mesh sizes, such as 40 mesh, 60 mesh, 80 mesh, 100 mesh, 120 mesh, and 140 mesh, based on the NY / T 1121.1-2006 standard to obtain soil particles of various particle sizes.

[0029] It should be noted that the different sizes of soil particles and the different composition of particles will affect the near-infrared spectrum of the soil. Figure 3 , Figure 3 A schematic diagram of diffuse reflectance spectroscopy analysis, provided in accordance with one embodiment of the present invention, illustrates diffusely reflected light. This light, referred to as diffusely reflected light, undergoes multiple transmissions and reflections within the sample before being reflected back from its initial point of entry. This diffusely reflected light carries chemical information about the sample and can therefore be used for spectral analysis.

[0030] After obtaining soil particles of different sizes at each sampling time after the same pruning, the spectral data of the soil samples were collected using a QEPro high-performance spectrometer and a NIRQuest512-2.5 infrared spectrometer. The absorbance Log (1 / R) at each wavelength point of the near-infrared diffuse reflectance spectrum was obtained. With the horizontal axis representing the near-infrared wavelength and the vertical axis representing the absorbance, an absorbance spectrum curve of the near-infrared diffuse reflectance spectrum was constructed. Figure 4 ,Figure 4 A schematic diagram of an absorbance spectrum curve provided by one embodiment of the present invention.

[0031] Hydrogen-containing groups include methyl CH, water OH, primary amine NH, and thiol SH pollutants. Near-infrared spectroscopy is essentially the absorption of the harmonic and sum frequencies of molecular vibrations. The absorption peaks of different hydrogen-containing groups correspond to different peak regions. Therefore, based on this, neural network models can be combined to detect components.

[0032] The peak region is obtained by analyzing the absorption peaks of different actual hydrogen-containing groups. At the same time, in order to facilitate subsequent processing, the highest absorbance in the peak region is taken as the peak value of the corresponding group, and there is no restriction on this.

[0033] S102: Under the same pruning, based on the changes in the group peaks at different sampling moments, determine the change amplitude sequence of each group peak in the spectral curve of different particle sizes; based on the absorbance range of the group peaks at all sampling moments, determine the reference peak; based on the difference between the different group peaks and the reference peak in the change amplitude sequence, determine the initial sensitivity of the spectral performance at each particle size.

[0034] Under the existing technology, soil nutrient modeling is based on nonlinear fitting of spectral curves, but the modeling is based on mixed soils of different particle sizes. On the one hand, the different particle sizes of the soil lead to pores between the particles, which affects the absorbance of the spectrum. Therefore, it is necessary to analyze the fluctuations of the spectral curves of particles of different particle sizes and analyze the sensitivity of changes in soil particles of different particle sizes under each pruning.

[0035] To analyze sensitivity, we first need to arrange the amplitudes of change of each group peak in the spectral curves for different particle sizes. Based on the changes in group peaks at different sampling times, we determine the amplitude sequence of each group peak in the spectral curves for different particle sizes. This involves taking the peak difference between the previous and next sampling times as the amplitude of change between the two sampling times. We then organize all the amplitude changes in the same peak region into an amplitude sequence in a time-sequential manner.

[0036] The amplitude of change is calculated by calculating the difference between the peak values ​​of the same peak region at the later sampling time and the previous sampling time between two adjacent sampling times. The amplitude of change sequence is then formed in time sequence. Each element in the amplitude of change sequence is the first-order difference of the corresponding group peak value.

[0037] Since different soil particle sizes have different sensitivities to changes, in the embodiment of the present invention, in order to eliminate the influence of soil particle size, it is necessary to set a reference peak value at each particle size to facilitate standardization processing when analyzing sensitivity later.

[0038] In an embodiment of the present invention, the reference peak is determined based on the absorbance range of the group peak at all sampling moments, including: calculating the absorbance range of the same group peak at all sampling moments, and taking the group peak corresponding to the median of the absorbance range as the reference peak.

[0039] By using the median method, the benchmark peak can be selected to represent the standardization level at the corresponding particle size, so as to accurately analyze the degree of influence of the peak at different particle sizes.

[0040] Furthermore, in some embodiments of the present invention, the initial sensitivity of the spectral performance at each particle size is determined based on the difference between different group peaks and the benchmark peak in the variation amplitude sequence, including: determining the response degree of different group peaks based on the similarity between different group peaks and the benchmark peak in the variation amplitude sequence; calculating the absolute value of the difference between the group peak and the benchmark peak at the same position in the variation amplitude sequence as the element difference; taking the sum of the element differences at all positions as the element benchmark difference; calculating the product value of the response degree and the element benchmark difference as the sensitivity coefficient of the corresponding group peak; and taking the average of the sensitivity coefficients of all group peaks as the initial sensitivity of the spectral performance at the corresponding particle size.

[0041] The similarity is the degree of similarity in fluctuations between the two sequences. In the embodiment of the present invention, the fluctuation similarity is calculated to achieve labeling processing and obtain the standardized weight of each peak value at each particle size.

[0042] Furthermore, in some embodiments of the present invention, the response degree of different group peaks is determined based on the correlation between different group peaks and the baseline peak in the variation amplitude sequence, including: calculating the Pearson correlation coefficient between any group peak and the baseline peak in the variation amplitude sequence; and taking the absolute value of the Pearson correlation coefficient as the response degree.

[0043] Among them, the Pearson correlation coefficient represents the degree of correlation between the two sequences, and its value is [-1,1]. The larger the absolute value of the Pearson correlation coefficient, the greater the influence of the change amplitude sequence of the corresponding group peak on the change amplitude sequence of the reference peak. That is, when the change amplitude of the reference peak changes, the change amplitude of the corresponding group peak will produce a more matching change effect, and the response degree of the group peak will be greater. Therefore, the absolute value of the Pearson correlation coefficient is directly used as the response degree.

[0044] The element difference is calculated by calculating the absolute value of the difference between the group peak and the reference peak at the same position in the amplitude variation sequence. The sum of the element differences at all positions is taken as the element reference difference. The element reference difference characterizes the overall difference in amplitude variation between the group peak and the reference peak.

[0045] For soil particles of the same size, the greater the variation in each peak across all sampling dates, the higher the sensitivity of that peak at that particle size. Furthermore, since the baseline peak at each particle size is used to standardize the spectral data of all peaks at that particle size, the greater the variation in a peak between different sampling dates compared to the baseline peak at the same particle size, the higher the sensitivity of that peak. Therefore, the product of the response level and the element baseline difference is calculated as the sensitivity coefficient of the corresponding group peak, characterizing the sensitivity of a single group peak. The average of the sensitivity coefficients of all group peaks is then used as the initial sensitivity of the spectral performance at the corresponding particle size.

[0046] The acquisition of the initial sensitivity can preliminarily characterize the sensitivity of the spectral performance of soil particles at the corresponding particle size, but still needs to be further corrected based on the actual soil conditions. Please refer to the subsequent embodiments for details.

[0047] S103: Correcting the initial sensitivity based on the particle size and the difference in water content of the soil samples after different pruning times to obtain a corrected sensitivity.

[0048] Two practical factors are mainly considered to achieve the correction of initial sensitivity. The first is due to the distinction of particle size, and the second is the difference in moisture content of the soil itself under different sampling conditions.

[0049] In the embodiment of the present invention, the moisture content of the soil can be collected at the same time as the soil sample is collected. The humidity 5 cm below the soil surface can be collected as the moisture content using a humidity sensor.

[0050] The initial sensitivity is corrected according to the particle size and the water content of the soil samples after different pruning times to obtain the corrected sensitivity, including: determining a correction coefficient according to the particle size and the water content of the soil samples after different pruning times; and calculating the product of the correction coefficient and the initial sensitivity as the corrected sensitivity.

[0051] Among them, the correction coefficient is used to realize the correction calculation, and the correction coefficient is determined according to the particle size and the moisture content of the soil samples after different pruning times, including: determining the particle size weight according to the mesh number in the soil particle screening process, wherein the higher the mesh number value, the smaller the particle size weight, and the particle size weight value is the normalized value; calculating the average moisture content of the soil samples after different pruning times, and taking the ratio of the moisture content of each pruning to the average moisture content as the moisture content weight; taking the product of the particle size weight and the moisture content weight as the correction coefficient.

[0052] Regarding particle size, it should be noted that small particles have a large surface area and high surface charge under the same mass, and their adsorption capacity for nutrients is far greater than that of large particles (sand particles). When water is lost, free water will preferentially carry away ionic nutrients in the pores of sand particles (such as 、 ) transport, while exchangeable nutrients on the clay surface require a larger hydraulic gradient to desorb. At low flow rates (<1 cm / h), water has ample time to desorb nutrients from the clay; at high flow rates (>5 cm / h), nutrients are primarily washed away from the sand.

[0053] Therefore, among soil particles of different sizes, water will preferentially carry away nutrients in large-sized particles, and the loss rate increases significantly with the increase in water migration speed. The fundamental reason is that the low adsorption capacity and high permeability of large particles make nutrients more easily leached, while small particles (clay particles) act as a "nutrient buffer" due to their strong adsorption capacity and low hydraulic conductivity.

[0054] Since the higher the mesh number of the sieve, the smaller the particle size, its weight is smaller. For example, the sieves used in this application are 40 mesh, 60 mesh, 80 mesh, 100 mesh, 120 mesh, and 140 mesh, so the particle size weights assigned are 1, 0.95, 0.90, 0.85, 0.80, and 0.75, respectively. The particle size weight is adjusted according to actual conditions and is not limited to this.

[0055] Changes in water content can also affect sensitivity. If the water content is high, the flow rate of various soil components is large, so the sensitivity is high. In the embodiment of the present invention, the ratio of the water content of each pruning to the average water content is used as the water content weight.

[0056] In summary, the product of the particle size weight and the water content weight is directly calculated as the correction coefficient, and the initial sensitivity is corrected based on the correction coefficient to obtain the corrected sensitivity. Specifically, the product of the correction coefficient and the initial sensitivity is calculated to obtain the corrected sensitivity.

[0057] S104: According to the correction sensitivity, the absorbance spectrum curves of different particle sizes are corrected and analyzed to obtain a corrected spectrum curve, and a neural network analysis is performed based on the corrected spectrum curve to output soil composition data.

[0058] The corrected sensitivity can accurately characterize the sensitivity at the corresponding particle size, and the spectrum can be corrected based on the corrected sensitivity.

[0059] Furthermore, in some embodiments of the present invention, the absorbance spectrum curves of different particle sizes are corrected and analyzed according to the correction sensitivity to obtain the corrected spectrum curve, including: weighting the absorbance of different wavelengths in the absorbance spectrum curves under different particle sizes based on the correction sensitivity to obtain the corrected spectrum curve.

[0060] After obtaining the corrected sensitivity of each particle size after each pruning, when modeling the hydrogen-containing group, the absorbance of each particle size at each wavelength is used to obtain the corrected absorbance at each wavelength by combining the corrected sensitivity weighting summary, and then the corrected spectrum curve after each pruning is constructed.

[0061] The corrected spectrum curve is used as the input spectrum wavelength of nonlinear fitting, and the artificial neural network ANN is used for modeling, that is, the spectrum wavelength is constructed as the input layer, the nonlinear activation of the hidden layer is passed through, and the output of the output layer is output, and the output result is the soil component data, wherein the activation function is the ReLU function.

[0062] In the embodiment of the present application, the initial sensitivity of the spectrum performance under each particle size is determined according to the difference between the peak value of different groups and the reference peak value in the change amplitude sequence by performing reference analysis on the change amplitude of absorbance, so that the change analysis can be accurately realized, then the initial sensitivity is corrected by the objective factors of particle size and the difference of water content of soil samples after different prunings, so that the actual spectrum deviation can be accurately represented, the deviation is adjusted to obtain the corrected spectrum curve, the neural network analysis is performed based on the corrected spectrum curve, and more accurate and reliable soil component data is output. In summary, the present application can effectively realize the influence analysis of soil particles of different particle sizes on spectrum data, reduce the error of soil data, establish a more accurate and objective soil component analysis model, and thus improve the accuracy of output soil component data.

[0063] It should be noted that the above-mentioned embodiment sequence of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0064] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.

Claims

1. A method for detecting soil composition in an oil-tea camellia nursery, characterized in that: The method comprises: Soil samples were obtained at different sampling times after pruning, dried, ground, and screened to obtain soil particles of different particle sizes; spectral analysis was performed on the soil particles to determine the absorbance spectrum curve of the near-infrared diffuse reflectance spectrum, and the near-infrared characteristic peak area of ​​the absorbance spectrum curve of different hydrogen-containing groups was determined as the peak area corresponding to the hydrogen-containing group, and the highest absorbance in the peak area was taken as the peak value of the corresponding group; Under the same pruning, the change amplitude sequence of each group peak in the spectral curve of different particle sizes is determined based on the change of group peak at different sampling moments; the reference peak is determined based on the absorbance range of the group peak at all sampling moments, and the initial sensitivity of the spectral performance at each particle size is determined based on the difference between the different group peaks and the reference peak in the change amplitude sequence; According to the particle size and the difference in water content of soil samples after different pruning times, the initial sensitivity is corrected to obtain the corrected sensitivity; According to the correction sensitivity, the absorbance spectrum curves of different particle sizes are corrected and analyzed to obtain a corrected spectrum curve. A neural network analysis is performed based on the corrected spectrum curve to output soil composition data.

2. The method for detecting soil composition in an oil-tea camellia nursery according to claim 1, wherein: The hydrogen-containing groups include methyl groups, water, primary amines, and thiol-type contaminants.

3. The method for detecting soil composition in an oil-tea camellia nursery according to claim 1, wherein: The step of determining the change amplitude sequence of each group peak value in the spectral curves of different particle sizes according to the change of the group peak value at different sampling moments includes: The difference between the peak values ​​of the same peak area at the next sampling moment and the previous sampling moment is taken as the change amplitude between the two corresponding sampling moments; All the change amplitudes in the same peak area are organized into a change amplitude sequence according to the time sequence.

4. The method for detecting soil composition in an oil-tea camellia nursery according to claim 1, wherein: Determining the reference peak value based on the absorbance extreme difference of the group peak value at all sampling moments includes: Calculate the absorbance range of the same group peak at all sampling times, and take the group peak corresponding to the median of the absorbance range as the reference peak.

5. The method for detecting soil composition in an oil-tea camellia nursery according to claim 1, wherein: The initial sensitivity of the spectrum at each particle size is determined based on the difference between the peak values ​​of different groups and the reference peak value in the variation amplitude sequence, including: The response degree of the peak values ​​of different groups is determined based on the similarity between the peak values ​​of different groups and the reference peak value in the change amplitude sequence; Calculate the absolute value of the difference between the group peak and the reference peak at the same position in the variation amplitude sequence as the element difference; and take the sum of the element differences at all positions as the element reference difference; Calculate the product of the response degree and the element reference difference as the sensitivity coefficient of the corresponding group peak; The average of the sensitivity coefficients of all group peaks is taken as the initial sensitivity of the spectrum at the corresponding particle size.

6. The method for detecting soil composition in an oil-tea camellia nursery according to claim 5, wherein: Based on the similarity between the peak values ​​of different groups and the reference peak value in the change amplitude sequence, the response degree of the peak values ​​of different groups is determined, including: Calculate the Pearson correlation coefficient between the peak value of any group and the baseline peak value in the change amplitude series; The absolute value of the Pearson correlation coefficient was used as the response level.

7. The method for detecting soil composition in an oil-tea camellia nursery according to claim 1, wherein: The initial sensitivity is corrected according to the particle size and the difference in water content of soil samples after different pruning times to obtain the corrected sensitivity, including: Determine the correction factor based on the particle size and the moisture content of the soil samples after different pruning times; The product value of the correction coefficient and the initial sensitivity is calculated as the corrected sensitivity.

8. The method for detecting soil composition in an oil-tea camellia nursery according to claim 7, wherein: The correction factor is determined based on the particle size and the moisture content of the soil samples after different pruning times, including: Determine the particle size weight according to the mesh number in the soil particle screening process, wherein the higher the mesh number value, the smaller the particle size weight, and the value of the particle size weight is a normalized value; Calculate the mean moisture content of soil samples after different pruning times, and use the ratio of the moisture content of each pruning to the mean moisture content as the moisture content weight; The product of the particle size weight and the water content weight is used as the correction coefficient.

9. The method for detecting soil composition in an oil-tea camellia nursery according to claim 1, wherein: The method of performing correction analysis on the absorbance spectrum curves of different particle sizes according to the correction sensitivity to obtain the corrected spectrum curves includes: The absorbances at different wavelengths in the absorbance spectrum curves at different particle sizes are weighted based on the corrected sensitivity to obtain a corrected spectrum curve.

10. The method for detecting soil composition in an oil-tea camellia nursery according to claim 1, wherein: The performing of neural network analysis based on the modified spectral curve and outputting soil composition data includes: The corrected spectral curve is used as the input spectral wavelength of nonlinear fitting, and modeling is performed based on the artificial neural network ANN. Through the nonlinear activation of the hidden layer, the output is output from the output layer, and the output result is the soil composition data, where the activation function is the ReLU function.

Images