Intelligent Detection Method and System for Ecological Nutrient Soil Based on Feature Extraction

Abstract

The present invention relates to the technical field of organic matter detection, and specifically relates to an intelligent detection method and system for ecological nutrient soil based on feature extraction. Multiple portions of nutrient soil are randomly selected from the same production batch of ecological nutrient soil, and then the spectral curves of each portion of nutrient soil are obtained respectively. Then, based on the differences in the spectral curves of different nutrient soils in different wavelength bands, the organic matter content of each portion of nutrient soil in the current production batch is analyzed and compared with the analysis of the spectral curves of the reference nutrient soil, so as to obtain the classification of the organic matter content of the ecological nutrient soil in the current production batch. The inspection of organic matter in the present invention relies on hyperspectral technology, which can effectively reduce the detection time and make the detection of the organic matter content in nutrient soil more accurate. The present invention also considers the influence of salinity on organic matter, and corrects the classification data of the organic matter content before analyzing the organic matter content, further improving the accuracy of the organic matter content detection.

Description

Technical Field

[0001] The present invention relates to the technical field of organic matter detection, and in particular to an intelligent detection method and system for ecological nutrient soil based on feature extraction. Background Art

[0002] Soil organic matter is an important component of the solid phase of soil and one of the main sources of plant nutrition. Broadly speaking, soil organic matter refers to all carbon-containing organic substances existing in soil in various forms, including various animal and plant residues, microorganisms in soil, and various organic substances decomposed and synthesized by them. Narrowly speaking, soil organic matter generally refers to a special, complex, and relatively stable high-molecular organic compound (humic acid) formed by the action of microorganisms on organic residues.

[0003] For the detection of organic matter content, in the prior art, one method is to use the potassium dichromate volumetric method for the determination of soil organic matter. Under the condition of external heating (oil bath temperature is 180 °C, boiling for 5 minutes), soil organic matter (carbon) is oxidized with a potassium dichromate-sulfuric acid solution of a certain concentration, and the remaining potassium dichromate is titrated with ferrous sulfate. The content of organic carbon is calculated from the amount of potassium dichromate consumed, and the operation is relatively complex. Another method is to use the silver chloride method, which is a method for accurately determining soil salinity and is applicable to the case where the salinity content in soil is relatively low. This method requires the preparation of silver chloride reagent, mixing the soil sample with the reagent, and judging the salinity content in the soil through the reaction of mercuric dichloride and silver oxide. The above two methods for judging the organic matter content mainly judge by setting a fixed threshold range of the organic matter content in the nutrient soil, and the judgment of the organic matter content is not accurate enough.

[0004] In view of this, the present invention is specifically proposed. Summary of the Invention

[0005] In order to solve the technical problem that when detecting the traditional organic matter content, the judgment of the organic matter content in the nutrient soil is not accurate enough by setting a fixed threshold range of the organic matter content in the nutrient soil. The purpose of the present invention is to provide an intelligent detection method and system for ecological nutrient soil based on feature extraction, and the specific technical solutions adopted are as follows:

[0006] First aspect, the present invention provides an intelligent detection method for ecological nutrient soil based on feature extraction. The detection method includes: randomly selecting Z portions of ecological nutrient soil with the same weight from the ecological nutrient soil of the current production batch; analyzing the z-th portion of ecological nutrient soil to obtain the organic matter content classification of the z-th portion of ecological nutrient soil; traversing the Z portions of ecological nutrient soil to obtain Z numbers of organic matter content classifications, and taking the organic matter content classification with the largest number of classifications as the organic matter content classification of the ecological nutrient soil of the current production batch; Z is a positive integer, and z ∈ [1, Z]; analyzing the z-th portion of ecological nutrient soil to obtain the organic matter content classification of the z-th portion of ecological nutrient soil, specifically including: using hyperspectral imaging technology to obtain the spectral curve to be measured of the z-th portion of ecological nutrient soil and the spectral curves of multiple reference nutrient soils; extracting features from the spectral curve to be measured in the spectral curve to be measured to obtain the to-be-measured organic matter characteristic band reflecting the organic matter information in the z-th portion of ecological nutrient soil; extracting features from the reference spectral curves in the multiple reference spectral curves to obtain the reference organic matter characteristic band reflecting the organic matter information in the reference nutrient soil; analyzing the to-be-measured organic matter characteristic band and the reference organic matter characteristic band to obtain the similarity degree of the organic matter content between the z-th portion of ecological nutrient soil and the multiple reference nutrient soils; taking the organic matter content classification of the reference nutrient soil corresponding to the maximum value of the similarity degree of the organic matter content as the organic matter content classification of the z-th portion of ecological nutrient soil.

[0007] The present invention first randomly selects multiple portions of ecological nutrient soil from the ecological nutrient soil of the current production batch, and then based on hyperspectral imaging technology, obtains the spectral curves of each portion of ecological nutrient soil respectively. Then, according to the differences in the spectral curves of hyperspectral detection of different nutrient soils in different bands, the organic matter content of each portion of ecological nutrient soil in the current production batch is analyzed, and the analysis result of the ecological nutrient soil in the current production batch is compared with the analysis of the reference nutrient soil, so as to obtain the organic matter content classification of the ecological nutrient soil in the current production batch. The inspection of organic matter in the present invention relies on hyperspectral technology, which can effectively reduce the detection time and is also more accurate in detecting the organic matter content in the nutrient soil.

[0008] Second aspect, the present invention provides an intelligent detection method for ecological nutrient soil based on feature extraction. The detection method includes: randomly selecting Z portions of ecological nutrient soil with the same weight from the ecological nutrient soil of the current production batch; without considering the influence of salt, analyzing the z-th portion of ecological nutrient soil to obtain the first classification of the organic matter content of the z-th portion of ecological nutrient soil; considering the influence of salt, analyzing the z-th portion of ecological nutrient soil to obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil; traversing Z portions of ecological nutrient soil to obtain 2Z classification quantities of organic matter content, and taking the classification of organic matter content with the largest number of classifications as the classification of the organic matter content of the ecological nutrient soil of the current production batch; Z is a positive integer, z ∈ [1, Z]; analyzing the z-th portion of ecological nutrient soil specifically includes: Step 1, using hyperspectral imaging technology to obtain the spectral curve to be measured of the z-th portion of ecological nutrient soil and the reference spectral curves of multiple reference nutrient soils; Step 2, by extracting features from the spectral curve to be measured in the spectral curve to be measured, obtaining the characteristic band of organic matter to be measured reflecting the organic matter information in the z-th portion of ecological nutrient soil; by extracting features from the reference spectral curves in the multiple reference spectral curves, obtaining the characteristic band of reference organic matter reflecting the organic matter information in the reference nutrient soil; by analyzing the characteristic band of organic matter to be measured and the characteristic band of reference organic matter, obtaining the similarity degree of organic matter content between the z-th portion of ecological nutrient soil and the multiple reference nutrient soils; taking the classification of the organic matter content of the reference nutrient soil corresponding to the maximum similarity degree of organic matter content as the first classification of the organic matter content of the z-th portion of ecological nutrient soil; Step 3, according to the influence degree of salt in the z-th portion of ecological nutrient soil on the organic matter content, optimizing the first classification of the organic matter content of the z-th portion of ecological nutrient soil to obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil.

[0009] Further, in the Step 3, according to the influence degree of salt in the z-th portion of ecological nutrient soil on the organic matter content, optimizing the first classification of the organic matter content of the z-th portion of ecological nutrient soil specifically includes: according to the influence degree of salt in the z-th portion of ecological nutrient soil on the organic matter content, correcting the spectral curve to be measured, and based on the corrected spectral curve to be measured, then performing Step 2, and taking the first classification of the organic matter content of the z-th portion of ecological nutrient soil obtained in Step 2 as the second classification of the organic matter content of the z-th portion of ecological nutrient soil.

[0010] Further, the method for obtaining the influence degree of the salt content on the organic matter content in the z-th portion of ecological nutrient soil includes: extracting features from the measured spectral curve graph to obtain a salt characteristic band reflecting the salt information in the z-th portion of ecological nutrient soil; analyzing the measured organic matter characteristic band, the salt characteristic band, the similarity degree of the organic matter content, and the first classification of the organic matter content of the z-th portion of ecological nutrient soil to obtain the influence degree of the salt content on the organic matter content in the z-th portion of ecological nutrient soil.

[0011] Further, the method for feature extraction includes: performing spectral transformation processing on the spectral curve and performing correlation analysis on each band, and determining the bands that pass the significant test at the P = 0.01 level and combine the stepwise regression VIF < 10 as the organic matter characteristic band and the salt characteristic band. Among them, the spectral transformation processing includes: first-order differential, envelope removal, logarithmic transformation, and multiplicative scatter correction.

[0012] Further, the organic matter content in the reference nutrient soil is known, and the organic matter contents of the multiple reference nutrient soils are different.

[0013] Further, the multiple reference nutrient soils are respectively nutrient soil samples with organic matter contents of 1%, 2%, 3%, 4%, 5%, and 6%.

[0014] Thirdly, the present invention provides an intelligent detection system for ecological nutrient soil based on feature extraction. The detection system includes: an organic matter classification module: used to randomly select Z portions of ecological nutrient soil with the same weight from the ecological nutrient soil of the current production batch; analyze the z-th portion of ecological nutrient soil to obtain the organic matter content classification of the z-th portion of ecological nutrient soil; traverse the Z portions of ecological nutrient soil to obtain Z organic matter content classification quantities, and take the organic matter content classification with the largest classification quantity as the organic matter content classification of the ecological nutrient soil of the current production batch; Z is a positive integer, and z ∈ [1, Z]; an organic matter analysis module: used to analyze the z-th portion of ecological nutrient soil to obtain the organic matter content classification of the z-th portion of ecological nutrient soil, specifically including: using hyperspectral imaging technology to obtain the spectral curve to be measured of the z-th portion of ecological nutrient soil and the spectral curves of multiple reference nutrient soils; extracting features from the spectral curve to be measured in the spectral curve to be measured to obtain the to-be-measured organic matter characteristic band reflecting the organic matter information in the z-th portion of ecological nutrient soil; extracting features from the reference spectral curves in the multiple reference spectral curves to obtain the reference organic matter characteristic band reflecting the organic matter information in the reference nutrient soil; analyzing the to-be-measured organic matter characteristic band and the reference organic matter characteristic band to obtain the similarity degree of the organic matter content between the z-th portion of ecological nutrient soil and the multiple reference nutrient soils; taking the organic matter content classification of the reference nutrient soil corresponding to the maximum value of the similarity degree of the organic matter content as the organic matter content classification of the z-th portion of ecological nutrient soil.

[0015] Fourth aspect, the present invention provides an intelligent detection system for ecological nutrient soil based on feature extraction. The detection system includes: an organic matter classification module: randomly select Z portions of ecological nutrient soil with the same weight from the ecological nutrient soil of the current production batch; without considering the influence of salt, analyze the z-th portion of ecological nutrient soil to obtain the first classification of the organic matter content of the z-th portion of ecological nutrient soil; considering the influence of salt, analyze the z-th portion of ecological nutrient soil to obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil; traverse the Z portions of ecological nutrient soil to obtain 2Z classification quantities of organic matter content, and take the classification of organic matter content with the largest number of classifications as the classification of the organic matter content of the ecological nutrient soil of the current production batch; Z is a positive integer, and z ∈ [1, Z]; an organic matter analysis module: used to analyze the z-th portion of ecological nutrient soil, specifically including: using hyperspectral imaging technology to obtain the spectral curve to be measured of the z-th portion of ecological nutrient soil and the reference spectral curves of multiple reference nutrient soils; by extracting features from the spectral curve to be measured in the spectral curve to be measured, obtain the to-be-measured organic matter characteristic band reflecting the organic matter information in the z-th portion of ecological nutrient soil; by extracting features from the reference spectral curves in the multiple reference spectral curves, obtain the reference organic matter characteristic band reflecting the organic matter information in the reference nutrient soil; by analyzing the to-be-measured organic matter characteristic band and the reference organic matter characteristic band, obtain the similarity degree of the organic matter content between the z-th portion of ecological nutrient soil and the multiple reference nutrient soils; take the classification of the organic matter content of the reference nutrient soil corresponding to the maximum value of the similarity degree of the organic matter content as the first classification of the organic matter content of the z-th portion of ecological nutrient soil; a classification optimization module: used to optimize the first classification of the organic matter content of the z-th portion of ecological nutrient soil according to the influence degree of salt in the z-th portion of ecological nutrient soil on the organic matter content, and obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil.

[0016] The present invention has the following beneficial effects:

[0017] The present invention analyzes the organic matter content of the current ecological nutrient soil to be measured according to the differences in the spectral curves of hyperspectral detection of different ecological nutrient soils in different bands, and compares the analysis results with the analysis of the reference nutrient soil, so as to obtain the classification of the organic matter content of the current ecological nutrient soil to be measured. Compared with the prior art, the present invention reduces the operation time and operation complexity by setting a fixed threshold range of the organic matter content in the nutrient soil for comparison and judgment, and the detection of the organic matter content in the ecological nutrient soil to be measured is also more accurate.

[0018] Since the increase in salt content will weaken the spectral response degree of organic matter and interfere with the spectral characteristics of organic matter, thereby affecting the prediction accuracy of organic matter content, the present invention also considers the influence of salt on organic matter. After correcting the classification data of organic matter content, the analysis of organic matter content is carried out to obtain the final classification of the proportion of organic matter, further improving the accuracy of the detection of organic matter content. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0020] Figure 1 Flowchart of an intelligent detection method for ecological nutrient soil based on feature extraction provided by a possible embodiment of the present invention;

[0021] Figure 2 Flowchart of an intelligent detection method for ecological nutrient soil based on feature extraction provided by another possible embodiment of the present invention;

[0022] Figure 3 Schematic diagram of an intelligent detection system for ecological nutrient soil based on feature extraction provided by a possible embodiment of the present invention;

[0023] Figure 4 Schematic diagram of an intelligent detection system for ecological nutrient soil based on feature extraction provided by another possible embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following will, in conjunction with the drawings and preferred embodiments, detail the specific implementation manners, structures, features and effects of an intelligent detection method and system for ecological nutrient soil based on feature extraction proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0026] The following will specifically describe the specific solutions of an intelligent detection method and system for ecological nutrient soil based on feature extraction provided by the present invention in conjunction with the drawings.

[0027] Soil organic matter is an important component of the solid phase of soil and one of the main sources of plant nutrition. It can promote the growth and development of plants, improve the physical properties of soil, promote the activities of microorganisms and soil organisms, promote the decomposition of nutrients in soil, and improve the fertility retention and buffering properties of soil. The following embodiments of the present invention analyze the organic matter content of the current nutrient soil according to the difference in the spectral curves of the hyperspectral detection of different nutrient soils in different bands, and finally determine the organic matter content of the nutrient soil, which greatly reduces the loss of time and improves the recognition accuracy of the organic matter content of the nutrient soil.

[0028] See also Figure 1 , which shows a flow chart of an intelligent detection method for ecological nutrient soil based on feature extraction in an embodiment of the present invention, in order to solve the technical problem that the judgment of the organic matter content in the nutrient soil is not accurate enough by setting a fixed threshold range for the organic matter content in the nutrient soil for comparison and judgment during traditional organic matter content detection.

[0029] First, Z portions of ecological nutrient soil of the same weight are randomly selected from the ecological nutrient soil of the current production batch; the ecological nutrient soil of the current production batch is the object for which the organic matter content classification judgment of the scheme of the present invention is required, that is, the ecological nutrient soil to be tested. Then, the selected Z portions of ecological nutrient soil are analyzed respectively.

[0030] Specifically, the zth portion of ecological nutrient soil is analyzed to obtain the classification of the organic matter content of the zth portion of ecological nutrient soil. The analysis method includes: using hyperspectral imaging technology to detect the Z portions of ecological nutrient soil respectively, and obtaining the Z portions of ecological nutrient soil spectrum curve graphs, each portion of the ecological nutrient soil spectrum curve graph includes a number of spectrum curves of the spectrum spectrum, and the arithmetic average of the spectrum curves of the several spectrums, and then each portion of the spectrum curve graph of the ecological nutrient soil to be tested corresponds to a spectrum curve. Using the same hyperspectral imaging technology method mentioned above, a plurality of reference nutrient soils with different organic matter ratios are analyzed in the same way to obtain a plurality of reference spectral curve graphs of a plurality of reference nutrient soils. Here, the organic matter composition and proportion of the reference nutrient soil are known, and nutrient soil samples with known organic matter contents of 1%, 2%, 3%, 4%, 5%, and 6% are selected as reference nutrient soils.

[0031] Then, feature extraction is performed based on the spectral curves, which is the process of selecting characteristic bands. By performing feature extraction on the spectral curves of the soil samples to be measured in the spectral curve graph of the soil samples to be measured, the characteristic bands of the organic matter to be measured reflecting the organic matter information in the z-th ecological nutrient soil are obtained. By performing feature extraction on the reference spectral curves in multiple reference spectral curve graphs, the reference characteristic bands of the organic matter reflecting the organic matter information in the reference nutrient soil are obtained. By analyzing the characteristic bands of the organic matter to be measured and the reference characteristic bands of the organic matter, the similarity degree of the organic matter content between the z-th ecological nutrient soil and multiple reference nutrient soils is obtained; the organic matter content classification of the reference nutrient soil corresponding to the maximum value of the similarity degree of the organic matter content is classified as the organic matter content classification of the z-th ecological nutrient soil.

[0032] Finally, using the above analysis method, traverse the Z ecological nutrient soils to be measured in turn, obtain Z classification quantities of the organic matter content, and classify the organic matter content classification with the largest number of classification quantities as the organic matter content classification of the ecological nutrient soil in the current production batch. Z is a positive integer, and z ∈ [1, Z].

[0033] In the embodiment of the present invention, multiple ecological nutrient soils are randomly selected from the ecological nutrient soils in the current production batch, and then based on the hyperspectral imaging technology, the spectral curves of each ecological nutrient soil are obtained respectively. Then, according to the differences in the spectral curves of the hyperspectral detection of different nutrient soils in different bands, the organic matter content of each ecological nutrient soil in the current production batch is analyzed, and the analysis results of the ecological nutrient soil in the current production batch are compared with the analysis of the reference nutrient soil, so as to obtain the organic matter content classification of the ecological nutrient soil in the current production batch. The inspection of the organic matter in the present invention relies on the hyperspectral technology, which can effectively reduce the detection time and is also more accurate in detecting the organic matter content in the nutrient soil.

[0034] In the embodiment of the present invention, according to the differences in the spectral curves of the hyperspectral detection of different ecological nutrient soils in different bands, the organic matter content of the current ecological nutrient soil to be measured is analyzed, and the analysis results are compared with the analysis of the reference nutrient soil, so as to obtain the organic matter content classification of the current ecological nutrient soil to be measured. Compared with the prior art, the embodiment of the present invention reduces the operation time and operation complexity by setting a fixed threshold range for the organic matter content in the nutrient soil for comparison and judgment, and is also more accurate in detecting the organic matter content in the ecological nutrient soil to be measured.

[0035] In another possible embodiment, according to the relationship between the organic matter and the salt in the spectrum, the content of the organic matter in the current nutrient soil is further analyzed to further optimize the accuracy of the organic matter content in the ecological nutrient soil to be measured. Based on the same inventive concept as the above embodiment, specifically, as Figure 2The flowchart of an intelligent detection method for ecological nutrient soil based on feature extraction, which shows another possible embodiment of the present invention. The detection method includes: randomly selecting Z portions of ecological nutrient soil with the same weight from the current production batch of ecological nutrient soil; without considering the influence of salt, analyzing the z-th portion of ecological nutrient soil to obtain the first classification of the organic matter content of the z-th portion of ecological nutrient soil; considering the influence of salt, analyzing the z-th portion of ecological nutrient soil to obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil; traversing Z portions of ecological nutrient soil to obtain 2Z classification quantities of organic matter content, and taking the classification of organic matter content with the largest number of classifications as the classification of the organic matter content of the ecological nutrient soil in the current production batch; Z is a positive integer, and z ∈ [1, Z];

[0036] Among them, the analysis of the z-th portion of ecological nutrient soil specifically includes:

[0037] Step 1: Using hyperspectral imaging technology, obtain the spectral curve graph to be measured of the z-th portion of ecological nutrient soil and the spectral curve graphs of multiple reference nutrient soils;

[0038] Step 2: By extracting features from the spectral curve to be measured in the spectral curve graph to be measured, obtain the characteristic band of organic matter to be measured reflecting the organic matter information in the z-th portion of ecological nutrient soil. By extracting features from the reference spectral curves in multiple reference spectral curve graphs, obtain the characteristic band of reference organic matter reflecting the organic matter information in the reference nutrient soil. By analyzing the characteristic band of organic matter to be measured and the characteristic band of reference organic matter, obtain the similarity degree of organic matter content between the z-th portion of ecological nutrient soil and multiple reference nutrient soils. Take the classification of the organic matter content of the reference nutrient soil corresponding to the maximum value of the similarity degree of organic matter content as the first classification of the organic matter content of the z-th portion of ecological nutrient soil;

[0039] Step 3: Optimize the first classification of the organic matter content of the z-th portion of ecological nutrient soil according to the influence degree of salt on the organic matter content in the z-th portion of ecological nutrient soil to obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil. According to the influence degree of salt on the organic matter content in the z-th portion of ecological nutrient soil, correct the spectral curve to be measured, and then execute Step 2 based on the corrected spectral curve to be measured. Take the first classification of the organic matter content of the z-th portion of ecological nutrient soil obtained in Step 2 as the second classification of the organic matter content of the z-th portion of ecological nutrient soil. Among them, the method for obtaining the influence degree of salt on the organic matter content in the z-th portion of ecological nutrient soil includes: by extracting features from the spectral curve graph to be measured, obtain the characteristic band of salt reflecting the salt information in the z-th portion of ecological nutrient soil; by analyzing the characteristic band of organic matter to be measured, the characteristic band of salt, the similarity degree of organic matter content, and the first classification of the organic matter content of the z-th portion of ecological nutrient soil, so as to obtain the influence degree of salt on the organic matter content in the z-th portion of ecological nutrient soil.

[0040] Since the increase in salt content will weaken the spectral response degree of organic matter and interfere with the spectral characteristics of organic matter, thereby affecting the prediction accuracy of the organic matter content, the embodiments of the present invention consider the influence of salt, correct the classified data of the organic matter content, and then analyze the organic matter content to obtain the final classification of the organic matter ratio, further improving the accuracy of the detection of the organic matter content.

[0041] The method for feature extraction involved in the embodiments of the present invention includes: performing spectral transformation processing on the spectral curve, and performing correlation analysis band by band, and determining the organic matter characteristic band and the salt characteristic band by passing the significant test at the P = 0.01 level and combining the bands with VIF < 10 in stepwise regression. Among them, the spectral transformation processing includes: first-order derivative, envelope removal, logarithmic transformation, and multiplicative scatter correction.

[0042] P refers to the P-value, which is a concept in hypothesis testing. The P-value represents the probability of observing the current sample or a more extreme sample under the condition that the null hypothesis is true. "P = 0.01" refers to the significance level used in the significance test, that is, only when the P-value is less than or equal to 0.01, we will consider the result to be statistically significant. VIF represents the variance inflation factor. During the variable selection process, the variance inflation factor (VIF) will be considered. VIF < 10 is considered that there is no serious multicollinearity between variables. Using statistical methods (such as t-test or ANOVA) to test whether the correlation between bands is significant at the significance level of P = 0.01 is a commonly known means for those skilled in the art and will not be elaborated here.

[0043] The detection of organic matter relies on hyperspectral technology, which can effectively reduce the detection time. The traditional judgment of the organic matter content mainly judges by setting a fixed threshold range of the organic matter content in the nutrient soil for comparison, and the judgment of the organic matter content is not accurate enough. The embodiments of the present invention analyze the organic matter content of the current nutrient soil according to the differences in the hyperspectral detection spectral curves of different nutrient soils in different bands. The embodiments of the present invention aim at the disadvantage that when detecting the traditional organic matter content, the judgment of the organic matter content in the nutrient soil is not accurate enough by setting a fixed threshold range of the organic matter content in the nutrient soil for comparison. Analyze the bands reflecting the organic matter content to obtain the gap between the organic matter content in the current nutrient soil and the organic matter content of the other selected nutrient soils with qualified organic matter content. And further analyze the content of organic matter in the current nutrient soil according to the relationship between organic matter and salt in the spectrum. The embodiments of the present invention analyze the organic matter content of the current nutrient soil according to the differences in the hyperspectral detection spectral curves of different nutrient soils in different bands. It reduces the operation time and operation complexity, and the detection of the organic matter content in the nutrient soil is also more accurate.

[0044] In another possible embodiment, the organic matter content of the current nutrient soil is analyzed based on the differences in the hyperspectral detection spectral curves of different nutrient soils in different bands.

[0045] Step A: Obtain the spectral curve graph of the nutrient soil of the production batch to be measured through hyperspectral technology. Generally, spectral collection is carried out indoors, and then the spectral data is preprocessed. In the embodiment of the present invention, the instrument used for soil spectral collection is a portable ground object hyperspectral spectrometer ASD FieldSpec 3 (Analytical Spectral Devices, USA). The spectral range is set to 350 - 2500 nm, and there are 2151 output bands. The spectral collection is carried out in a dark room. The soil sample is placed in a black sample dish. Before measurement, it is calibrated with a standard white board. The surface of the soil sample is slightly scraped flat to make its surface as flat as possible. 10 spectral curves are collected repeatedly each time, and a total of 30 spectral curves are collected. The arithmetic mean is obtained using the built-in software ViewspecPro of the hyperspectral spectrometer to obtain the final spectral data. Randomly weigh Z (Z is set to 100) 50 - gram nutrient soil samples from the nutrient soil produced in the same batch for the above hyperspectral detection and analysis.

[0046] Step B: Analyze the characteristics in the spectral curve.

[0047] 1. Obtain the bands reflecting organic matter and salts in the nutrient soil to be measured. Different bands in the spectral curve reflect different components in the substance. The embodiment of the present invention mainly analyzes the organic matter in the nutrient soil. Therefore, the bands mainly selected for analysis are those reflecting the organic matter content. The original spectral curve of the organic matter content of the nutrient soil to be measured is subjected to transformation processing (mainly including four spectral transformation processes of first - order differential (FD), continuum removal (CR), logarithmic transformation (Log), and multiplicative scatter correction (MSC) for the original spectral curve R). The correlation analysis is carried out band by band. The bands that pass the significant test at the P = 0.01 level and combine with the step - by - step regression VIF < 10 are determined as characteristic bands. Finally, it is found that the organic matter characteristic bands are mainly concentrated in the ranges of 428 - 694, 769 - 1374, 1426 - 1506, 1719 - 1955, and 2046 - 2386 nm. The characteristic bands of salts are obtained by using a similar method, and it is found that some bands overlap with the organic matter bands.

[0048] 2. Analysis of organic matter content. Obtain the manifestations of nutrient soils with different proportions of organic matter in hyperspectral data. The raw materials of nutrient soils usually include humic acid, peat soil, perlite, sand, wood chips, etc. Nutrient soils are mainly made by mixing various raw materials in a certain proportion. Different qualities of the raw materials used will result in different organic matter contents in the produced nutrient soils. The organic matter content in the nutrient soils used for different plants or different states of plants should be different. Now, determine the content ratios of peat soil, perlite, sand, wood chips, etc., excluding humic acid which affects the organic matter content, and select nutrient soil samples with known organic matter contents of 1%, 2%, 3%, 4%, 5%, and 6% as references. Detect the selected nutrient soil samples with a hyperspectrometer to obtain the corresponding hyperspectral curves.

[0049] Analyze the proportion of organic matter content. Generally, the organic matter content in soil is 1% - 5%, and about 2% is sufficient for growing crops. If the soil organic matter reaches 5%, it is considered a very fertile organic matter level. Considering that soil organic matter is an important component of the solid phase of soil, a major source of plant nutrition, it can promote plant growth and development, improve the physical properties of soil, promote the activities of microorganisms and soil organisms, promote the decomposition of nutrient elements in soil, and enhance the fertilizer retention and buffering capacity of soil, of course, the higher the organic matter content, the better. However, plants have different demands for organic matter under different circumstances, so a reasonable classification of the organic matter content in the currently detected nutrient soil should be made. In the spectral curve, it is manifested that the smaller the difference between the curve of the currently measured nutrient soil and the spectral curve of the reference nutrient soil with a known organic matter content in the selected organic matter wavelength band, the more consistent the organic matter content of the currently detected nutrient soil is with the classification of this reference nutrient soil. (In the spectral curve obtained by the hyperspectrometer, the smaller the reflectance, the better the absorption ability of a certain component in the substance to the light wave of the wavelength band, and the higher the content of this component.)

[0050] Thus, the similarity degree W between the organic matter content of the current nutrient soil z (i.e., the z-th measured ecological nutrient soil) and the organic matter content of the reference nutrient soil m (i.e., the m-th reference nutrient soil) can be obtained. z,m It is:

[0051]

[0052] In the above formula, k i,j,z is the slope of the i-th wave point on the spectral curve of the current nutrient soil z in the j-th organic matter wavelength band (the organic matter wavelength band refers to the organic matter characteristic wavelength band obtained in the above step A). k i,j,m is the slope of the i-th wave point on the spectral curve of the m-th reference nutrient soil in the j-th organic matter wavelength band. S j,z is the size of the area enclosed by the curve of the spectral curve of the current nutrient soil z and the x-axis in the j-th organic matter wavelength band. S j,mis the area enclosed by the curve of the m-th reference nutrient soil spectral curve and the x-axis in the j-th organic matter band.

[0053] is the difference between the slopes of the spectral curves of the current nutrient soil z and the reference nutrient soil m at all wave points in all organic matter bands. The smaller the value, the closer the organic matter content of the current nutrient soil z is to that of the reference nutrient soil m. is the reciprocal of. The larger the value, the closer the organic matter content of the current nutrient soil z is to that of the reference nutrient soil m. n j represents the number of organic matter bands, n i represents the number of wave points. To ensure that the denominator is not zero, a tuning parameter ∈ is added to the denominator, ∈ = 0.01.

[0054] is the difference between the areas enclosed by the spectral curves of the current nutrient soil z and the reference nutrient soil m and the x-axis in all organic matter bands. The smaller the value, the closer the organic matter content of the current nutrient soil z is to that of the reference nutrient soil m.

[0055] is the reciprocal of. The larger the value, the closer the spectral curve of the current nutrient soil z is to that of the reference nutrient soil m, and the closer the organic matter content of the current nutrient soil z is to that of the reference nutrient soil m.

[0056] Take the reference nutrient soil m with the highest similarity degree W z,m in organic matter content to the current nutrient soil z as the classification of the organic matter content of the current nutrient soil z.

[0057] Consider the influence of salt on the organic matter content. The spectral responses of organic matter and salt in the visible light band range are significant and the spectral response variation laws in the whole band are basically the same, indicating that their sensitive bands may overlap. As the salt content increases, the prediction accuracy of the organic matter content gradually decreases. This is because the soil spectral reflectance increases non-linearly with the increase of the salt content, while the presence of organic matter will reduce the soil spectral reflectance. The effects of the two on the spectral reflectance are exactly opposite, and the sensitive bands of salt and organic matter overlap in the visible light range. Therefore, the increase of the salt content will weaken the spectral response degree of organic matter, interfere with the spectral characteristics of organic matter, and thus affect the prediction accuracy of the organic matter content. The salt content in the nutrient soil should be detected, and when the salt content is higher, the actual organic matter content in the soil is higher than that shown by the spectral curve. However, at the same time, when the organic matter content is higher, the influence degree of salt on the organic matter content will become smaller.

[0058] From this, the influence degree R of the salt in the current nutrient soil z on the organic matter content can be obtained zis:

[0059]

[0060] In the above formula, S j ′ is the size of the area enclosed by the curve and the x-axis in the j-th salt band where the spectral curve of the current nutrient soil z does not overlap with the organic matter band. is the sum of S j ′ of all salt bands in the current nutrient soil z that do not overlap with the organic matter band. The smaller the value, the more salt content in the current nutrient soil z. f j,min ′ is the minimum reflectance in the j-th salt band where the spectral curve of the current nutrient soil z does not overlap with the organic matter band. is the sum of f j,min ′ of all salt bands in the current nutrient soil z that do not overlap with the organic matter band. The smaller the value, the more salt content in the current nutrient soil z and the greater the impact on the organic matter content in the current nutrient soil z. W z,m is the similarity degree between the organic matter content of the current nutrient soil z and the organic matter content of the reference nutrient soil m. p is the classification of the organic matter content of the current nutrient soil z. When W z,m , p are larger, the organic matter content is larger and the influence of the salt content on the organic matter content is smaller.

[0061] Normalize R z to get r z , whose value range is [0, 1].

[0062] In the spectral curve obtained by the hyperspectral spectrometer, the smaller the reflectance, the better the absorption ability of a certain component in the substance to the light wave of the band and the higher the content of this component. When r z is larger, the influence degree of the salt is larger, the content of the organic matter should be higher than that shown by the spectral curve, and the reflectance of each wave point in the organic matter band should be smaller. Through r z , modify and restore the reflectance of each wave point in the organic matter band of the currently detected nutrient soil z. Thus, the reflectance value f j,i ′ corresponding to the restored wave point of the i-th wave point in the j-th organic matter band is:

[0063] f j ′ ,i = f j,i × (1 - r z )

[0064] In the above formula, f j,i is the reflectance value corresponding to the i-th wave point in the j-th organic matter band before restoration.

[0065] The data is corrected, and after the data is restored, the data unaffected by salts is obtained. The spectral curves after restoration are compared in terms of the proportion of organic matter content, and the final classification of the organic matter proportion is obtained.

[0066] Step C: Process multiple portions of nutrient soil in the current production batch. Extract multiple other samples of the same batch of the current nutrient soil, perform the above analysis on multiple nutrient samples, and use the classification with the largest number obtained as the organic matter classification of the nutrient soil samples in the current production batch. Process it appropriately according to the organic matter classification of the nutrient soil in the current batch to make it more in line with the needs of different plants or plants at different stages.

[0067] Based on the same inventive concept as the above method embodiment, the embodiment of the present invention also provides an intelligent detection system for ecological nutrient soil based on feature extraction, as Figure 3 shown. The detection system specifically includes:

[0068] Organic matter classification module: used to randomly select Z portions of ecological nutrient soil with the same weight from the ecological nutrient soil in the current production batch; analyze the z-th portion of ecological nutrient soil to obtain the organic matter content classification of the z-th portion of ecological nutrient soil; traverse the Z portions of ecological nutrient soil to obtain Z numbers of organic matter content classifications, and use the organic matter content classification with the largest number of classifications as the organic matter content classification of the ecological nutrient soil in the current production batch; Z is a positive integer, and z ∈ [1, Z].

[0069] Organic matter analysis module: used to analyze the z-th portion of ecological nutrient soil to obtain the organic matter content classification of the z-th portion of ecological nutrient soil, specifically including: using hyperspectral imaging technology to obtain the spectral curve graph to be measured of the z-th portion of ecological nutrient soil and the spectral curve graphs of multiple reference nutrient soils; extracting features from the spectral curve to be measured in the spectral curve graph to be measured to obtain the to-be-measured organic matter characteristic band reflecting the organic matter information in the z-th portion of ecological nutrient soil; extracting features from the reference spectral curves in the multiple spectral curve graphs of reference nutrient soils to obtain the reference organic matter characteristic band reflecting the organic matter information in the reference nutrient soil; analyzing the to-be-measured organic matter characteristic band and the reference organic matter characteristic band to obtain the similarity degree of organic matter content between the z-th portion of ecological nutrient soil and multiple reference nutrient soils; using the organic matter content classification of the reference nutrient soil corresponding to the maximum value of the similarity degree of organic matter content as the organic matter content classification of the z-th portion of ecological nutrient soil.

[0070] In a possible embodiment, considering the influence of salts in the nutrient soil on organic matter, the embodiment of the present invention also provides an intelligent detection system for ecological nutrient soil based on feature extraction, as Figure 4 shown. The detection system specifically includes:

[0071] The detection system includes: an organic matter classification module: randomly select Z portions of ecological nutrient soil with the same weight from the ecological nutrient soil of the current production batch; without considering the influence of salt, analyze the z-th portion of ecological nutrient soil to obtain the first classification of the organic matter content of the z-th portion of ecological nutrient soil; considering the influence of salt, analyze the z-th portion of ecological nutrient soil to obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil; traverse the Z portions of ecological nutrient soil to obtain 2Z organic matter content classification quantities, and take the organic matter content classification with the largest number of classifications as the organic matter content classification of the ecological nutrient soil of the current production batch; Z is a positive integer, and z ∈ [1, Z].

[0072] An organic matter analysis module: used to analyze the z-th portion of ecological nutrient soil, specifically including: using hyperspectral imaging technology to obtain the spectral curve graph to be measured of the z-th portion of ecological nutrient soil and the spectral curve graphs of multiple reference nutrient soils; by extracting the features of the spectral curve to be measured in the spectral curve graph to be measured, obtain the characteristic band of organic matter to be measured reflecting the organic matter information in the z-th portion of ecological nutrient soil; by extracting the features of the reference spectral curves in the multiple spectral curve graphs of reference nutrient soils, obtain the characteristic band of reference organic matter reflecting the organic matter information in the reference nutrient soil; by analyzing the characteristic band of organic matter to be measured and the characteristic band of reference organic matter, obtain the similarity degree of organic matter content between the z-th portion of ecological nutrient soil and multiple reference nutrient soils; take the organic matter content classification of the reference nutrient soil corresponding to the maximum value of the similarity degree of organic matter content as the first classification of the organic matter content of the z-th portion of ecological nutrient soil.

[0073] A classification optimization module: used to optimize the first classification of the organic matter content of the z-th portion of ecological nutrient soil according to the influence degree of salt on the organic matter content in the z-th portion of ecological nutrient soil to obtain the second classification of the organic matter content of the z-th portion of ecological nutrient soil.

[0074] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the superiority or inferiority 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, multitasking and parallel processing are also possible or may be advantageous.

[0075] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

Claims

1. An intelligent detection method for ecological nutrient soil based on feature extraction, characterized in that: The detection method comprises: Randomly select Z portions of ecological nutrient soil of the same weight from the current production batch of ecological nutrient soil; without considering the influence of salt, analyze the zth portion of ecological nutrient soil to obtain the first classification of organic matter content of the zth portion of ecological nutrient soil; considering the influence of salt, analyze the zth portion of ecological nutrient soil to obtain the second classification of organic matter content of the zth portion of ecological nutrient soil; traverse the Z portions of ecological nutrient soil to obtain 2Z organic matter content classification numbers, and take the organic matter content classification with the largest number of classifications as the organic matter content classification of the current production batch of ecological nutrient soil; Z is a positive integer, ; The analysis of the zth ecological nutrient soil includes: Step 1, using hyperspectral imaging technology to obtain a spectrum curve graph to be tested of the zth ecological nutrient soil and multiple reference spectrum curve graphs of multiple reference nutrient soils; Step 2, by extracting features from the spectral curve to be tested in the spectral curve graph to be tested, a characteristic band of organic matter to be tested reflecting the organic matter information in the zth ecological nutrient soil is obtained; by extracting features from the reference spectral curves in the multiple reference spectral curve graphs, a reference organic matter characteristic band reflecting the organic matter information in the reference nutrient soil is obtained; By analyzing the characteristic band of the organic matter to be measured and the characteristic band of the reference organic matter, the similarity of the organic matter content between the zth ecological nutrient soil and the plurality of reference nutrient soils is obtained; the organic matter content of the reference nutrient soil corresponding to the maximum value of the similarity of the organic matter content is classified as the first classification of the organic matter content of the zth ecological nutrient soil; Step three, according to the influence of salt on the organic matter content in the zth portion of ecological nutrient soil, optimizing the first classification of the organic matter content of the zth portion of ecological nutrient soil to obtain the second classification of the organic matter content of the zth portion of ecological nutrient soil; In the step three, according to the influence of salt on the organic matter content in the zth portion of ecological nutrient soil, the first classification of the organic matter content of the zth portion of ecological nutrient soil is optimized, specifically including: According to the influence of salt on the organic matter content in the zth portion of ecological nutrient soil, the spectral curve to be measured is corrected, and based on the corrected spectral curve to be measured, step 2 is performed again, and the first classification of the organic matter content of the zth portion of ecological nutrient soil obtained in step 2 is used as the second classification of the organic matter content of the zth portion of ecological nutrient soil; The method for obtaining the influence degree of salt on organic matter content in the zth ecological nutrient soil comprises: By performing feature extraction on the spectrum curve graph to be measured, a salt characteristic band reflecting the salt information in the z-th ecological nutrient soil is obtained; By analyzing the characteristic band of the organic matter to be tested, the characteristic band of salt, the similarity of the organic matter content and the first classification of the organic matter content of the zth ecological nutrient soil, the influence of salt on the organic matter content in the zth ecological nutrient soil is obtained.

2. The method for intelligent detection of ecological nutrient soil based on feature extraction according to claim 1 is characterized in that: The feature extraction method comprises: The spectral curve was subjected to spectral transformation processing, and correlation analysis was performed band by band. The bands with VIF<10 were determined as organic matter characteristic bands and salt characteristic bands through a significant test at the P=0.01 level and combined with stepwise regression.

3. The method for intelligent detection of ecological nutrient soil based on feature extraction according to claim 2 is characterized in that: The spectrum transformation process includes: first-order differentiation, envelope removal, logarithmic transformation and multivariate scatter correction.

4. The method for intelligent detection of ecological nutrient soil based on feature extraction according to claim 1 is characterized in that: The organic matter content in the reference nutrient soil is known, and the organic matter content of the multiple reference nutrient soils is different.

5. The method for intelligent detection of ecological nutrient soil based on feature extraction according to claim 1 is characterized in that: The multiple reference nutrient soils are nutrient soil samples with organic matter contents of 1%, 2%, 3%, 4%, 5% and 6% respectively.

6. An intelligent detection system for ecological nutrient soil based on feature extraction, the detection system is used to execute the method of claim 1, characterized in that: The detection system comprises: Organic matter classification module: randomly select Z portions of ecological nutrient soil of the same weight from the ecological nutrient soil of the current production batch; analyze the zth portion of ecological nutrient soil without considering the influence of salt, and obtain the first classification of organic matter content of the zth portion of ecological nutrient soil; analyze the zth portion of ecological nutrient soil considering the influence of salt, and obtain the second classification of organic matter content of the zth portion of ecological nutrient soil; traverse the Z portions of ecological nutrient soil to obtain 2Z organic matter content classification numbers, and take the organic matter content classification with the largest number of classifications as the organic matter content classification of the current production batch of ecological nutrient soil; Z is a positive integer, ; Organic matter analysis module: used to analyze the zth ecological nutrient soil, specifically including: using hyperspectral imaging technology to obtain the spectrum curve diagram to be measured of the zth ecological nutrient soil and multiple reference spectrum curve diagrams of multiple reference nutrient soils; by extracting features from the spectrum curve to be measured in the spectrum curve diagram to be measured, obtaining the characteristic band of organic matter to be measured that reflects the organic matter information in the zth ecological nutrient soil; by extracting features from the reference spectrum curves in the multiple reference spectrum curve diagrams, obtaining the reference organic matter characteristic band that reflects the organic matter information in the reference nutrient soil; by analyzing the characteristic band of organic matter to be measured and the reference organic matter characteristic band, obtaining the degree of similarity of organic matter content between the zth ecological nutrient soil and the multiple reference nutrient soils; classifying the organic matter content of the reference nutrient soil corresponding to the maximum value of the similarity of organic matter content as the first classification of the organic matter content of the zth ecological nutrient soil; Classification optimization module: used to optimize the first classification of organic matter content in the zth portion of ecological nutrient soil according to the influence of salt on organic matter content in the zth portion of ecological nutrient soil, and obtain the second classification of organic matter content in the zth portion of ecological nutrient soil.

Images