Portable soil nutrient detection method based on multispectrum

Abstract

The invention discloses a portable soil nutrient detection method based on multispectrum. The method comprises the following steps: acquiring a soil sample with a preset volume and pre-treating the soil sample; collecting and processing spectral feature information of the preprocessed soil sample by using a multi-spectrometer to form a spectral data set; measuring the nutrients in the soil sample through a standardized chemical analysis method; matching a measured result with the spectral data set, and dividing the data set into a training set and a test set by adopting a random division method; a prediction model is trained through the training set, then the generalization ability of the prediction model is evaluated through the test set, and the prediction model after training and testing is completed is used for achieving prediction of soil nutrients according to spectral feature information obtained by the multiple spectrometers. The device is convenient to operate, and in-situ rapid detection of soil nutrients can be rapidly achieved.

Description

Technical Field

[0001] The invention relates to the technical field of soil analysis, and in particular to a portable soil nutrient detection method based on multi-spectrum. Background Art

[0002] With the continuous development of industrialization and informatization, agricultural output has been greatly improved. As the foundation and important component of agriculture, soil provides nutrients and water for the growth of crops, and is also the medium for the extension and fixation of crop roots. Therefore, timely detection of the content of nutrients such as nitrogen, phosphorus, potassium, and boron in the soil and rapid acquisition of various nutrient information in the soil are of great significance to the reasonable and accurate fertilization of crops and environmental protection.

[0003] However, in the prior art, the nutrient content of soil is measured by traditional chemical analysis test methods, but this method has high cost, low efficiency, long measurement cycle, and will produce chemical waste that pollutes the environment, and cannot achieve rapid detection of soil nutrient content. Compared with traditional chemical analysis technology, near-infrared spectroscopy analysis technology is favored by people because of its advantages such as fast, simple, low-cost, low destructiveness and simultaneous determination of multiple components. Near-infrared spectroscopy, as a green, environmentally friendly, fast and non-destructive analysis technology, can quickly obtain information such as multiple elements in soil in a relatively short time, and can achieve detection without any pretreatment or simple operation of the sample, without chemical reagents, and will not cause secondary pollution to the environment. It can be seen that near-infrared spectroscopy analysis technology has the potential to quantitatively invert the physical properties and component content of the detected object, and is widely used in many fields. It is of great significance to apply it to the field of soil nutrient detection.

[0004] At present, the research on soil nutrients by near-infrared spectroscopy is mainly carried out in the laboratory using a large Fourier transform spectrometer. The soil is transported back to the laboratory for air-drying, sieving and grinding, and then the spectrum is collected. Then, the soil nutrients are modeled and predicted by combining machine learning methods. However, when the established model is used to predict in-situ soil, the prediction accuracy is low and the model fails. Therefore, how to achieve accurate detection of soil nutrients is a technical problem that needs to be solved urgently in this field. Summary of the invention

[0005] In order to solve the above technical problems, this application proposes the following technical solutions:

[0006] In a first aspect, the present application provides a portable soil nutrient detection method based on multi-spectrum, comprising:

[0007] Obtain soil samples of a preset volume and pre-treat them;

[0008] Using a multi-spectrometer to collect and process spectral characteristic information of the pre-treated soil sample to form a spectral data set;

[0009] Determining the nutrients in the soil sample by standardized chemical analysis methods;

[0010] Matching the measured results with the spectral data set, and dividing the data set into a training set and a test set by a random division method;

[0011] The prediction model is trained by the training set, and then the generalization ability of the prediction model is evaluated by the test set. The prediction model after training and testing is used to predict soil nutrients according to the spectral feature information obtained by the multi-spectrometer.

[0012] In a possible implementation, obtaining a preset volume of soil samples and preprocessing them includes: collecting mixed soil samples of the plow layer along a preset route according to the principles of random, equal amount and multi-point mixing; placing the collected mixed soil samples into a grinder for sufficient grinding; screening the ground mixed soil samples through a sieve; and using a tablet press to press the screened mixed soil samples into soil tablets.

[0013] In a possible implementation, the mixed soil samples of the cultivated layer are collected along a preset route according to the principles of random, equal amount and multi-point mixing, including:

[0014] In case of small terrain changes, uniform soil fertility, and small sampling unit area, plum blossom-shaped sampling is adopted, and each sample requires 15-20 sampling points to collect soil and mix it evenly;

[0015] The sampling points should be evenly distributed, and the entire sampling area should be controlled as a whole to avoid sampling in places where fertilizers have been piled, ridges, ditches, and special terrain areas;

[0016] The soil sampling depth and sampling volume for each sampling point must be uniform and consistent, and the ratio of the upper and lower layers of the soil sample must be the same;

[0017] If the amount of mixed soil samples collected is too large, you can use a shovel to repeatedly turn it upside down into a pile, then divide it into four parts in a "cross" from the middle from top to bottom to discard the excess soil.

[0018] In a possible implementation, the collecting and processing spectral characteristic information of the pre-treated soil sample using a multi-spectrometer to form a spectral data set includes:

[0019] A portable multi-spectrometer is placed directly above the soil pressed sheet to collect spectral reflectance characteristics of the soil pressed sheet in different bands, where the bands cover a range from visible light to near infrared;

[0020] The spectral reflectance is measured multiple times for the same acquisition area and the average value is calculated;

[0021] The ratio and product between the features of adjacent spectral bands are calculated and combined with the original spectral features by column splicing.

[0022] In a possible implementation, the calculation formula for calculating the ratio and product between the characteristics of adjacent bands of the spectrum includes:

[0023]

[0024] Multiplication=R i+1 ×R i , i=0,1,2,3,…,16;

[0025] Among them, R i With R i+1 Represents the reflectivity of the i-th or i+1-th band, Ratio i and Multiplication i They represent the spectral ratio characteristics and spectral product characteristics of the i-th band respectively.

[0026] In a possible implementation, the nutrients in the soil sample are determined by a standardized chemical analysis method, including:

[0027] NY / T 1121.24-2012NY / T 1121.24-2012 was used to determine the total nitrogen content in soil samples;

[0028] The total phosphorus content in soil samples was determined using LY / T 1232-2015;

[0029] The total potassium content in soil samples was determined using LY / T 1234-2015;

[0030] The organic matter content in soil samples was determined using NY / T 1121.6-2006;

[0031] LY / T 1228-2015 was used to determine the alkaline nitrogen content in soil samples;

[0032] The available phosphorus content in soil samples was determined using NY / T 1121.7-2014;

[0033] NY / T 889-2004 was used to determine the contents of fast-acting potassium and slow-acting potassium in soil samples.

[0034] In a possible implementation, the prediction model completed by training and testing is used to predict soil nutrients based on spectral feature information obtained by a multi-spectrometer, including: using linear regression analysis to calculate the correlation between spectral features and soil nutrient content by the least squares method, and the calculation formula is:

[0035] y=β 0 +β 1 x 1 +β 2 x 2 +…+β n x n , n = 52,

[0036] Among them, β n represents the coefficient of the nth feature, x n represents the nth feature, and y is the predicted soil nutrient content.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The portable soil nutrient detection method based on multi-spectrum provided in this application is to conduct strict pretreatment of soil samples and then use standardized chemical analysis methods to determine the various nutrient contents in the soil samples. The chemical detection results of soil nutrients are then matched with the corresponding spectral feature data sets, and linear regression analysis is used. The correlation between spectral features and soil nutrient content is calculated by the least squares method to establish a reliable prediction model. By calculating the determination coefficient of the model prediction results, the prediction accuracy and robustness of the model are evaluated, thereby verifying its application value in soil nutrient detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of a flow chart of a portable soil nutrient detection method based on multispectral provided in an embodiment of the present application;

[0040] Figure 2 A schematic diagram of soil sample detection provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The present solution is described below in conjunction with the accompanying drawings and specific implementation methods.

[0042] Figure 1 A schematic diagram of a portable soil nutrient detection method based on multi-spectrum provided in the present application is shown in FIG. Figure 1 In this embodiment, a portable soil nutrient detection method based on multispectral includes:

[0043] S101, obtaining a preset volume of soil sample and performing pre-processing.

[0044] In this embodiment, the mixed soil samples of the cultivated layer are sampled along a certain route according to the principles of "random", "equal amount" and "multi-point mixing". Among them, "random" means that each sampling point is arbitrarily determined, so that all points in the sampling unit have an equal chance to be collected, "equal amount" requires that the depth and sampling amount of each point be consistent, and "multi-point mixing" means that the soil samples collected from each point in a sampling unit are evenly mixed to form a mixed sample to improve the representativeness of the sample. When sampling, "S"-shaped sampling is generally used, which can better overcome the errors caused by agronomic measures such as tillage and fertilization. However, in the case of small terrain changes, relatively uniform soil fertility, and small sampling unit area, plum blossom-shaped sampling can also be used, and each sample requires 15-20 sampling points to collect soil and mix it evenly.

[0045] The sampling points should be evenly distributed, and the entire sampling area should be controlled as a whole. Sampling should be avoided in places where fertilizers have been piled, ridges, ditches, and special terrain. The soil sampling depth and sampling volume for each sampling point should be uniform, and the ratio of the upper and lower layers of the soil sample should be the same. The sampler should be perpendicular to the ground and inserted into the soil to the specified depth. When sampling with a soil shovel, a tillage layer section should be shoveled out first, and then the soil should be shoveled parallel to the section.

[0046] For situations where the number of mixed soil samples collected is too large, see Figure 2 , it is advisable to take 1 kg of mixed soil sample, and you can use a shovel to repeatedly turn it upside down into a pile, and then use a "cross" method from the middle from top to bottom to discard the excess soil. Put the soil sample on a plate or plastic sheet and air dry it in a cool place. After the soil sample is air-dried, use a rolling pin or hammer or other similar tools to break and press it until it is the size of a millet grain. After breaking, use a shovel to repeatedly mix the sample, spread it into a thin layer, divide it into grids, and use a spoon to take soil from each small grid for sample testing.

[0047] The collected mixed soil samples are then put into a grinder for full grinding, and the ground mixed soil samples are screened through a 60-mesh or 10-mesh sieve to ensure the consistency of the sample particle size. The screened mixed soil samples are placed in a tablet press to be pressed into soil tablets for subsequent spectral detection.

[0048] S102: using a multi-spectrometer to collect and process spectral characteristic information of the pre-treated soil sample to form a spectral data set.

[0049] In this embodiment, during the spectral collection stage, a portable multi-spectrometer is placed directly above the soil tablet at a distance of 10 cm to ensure uniform illumination of the light source. The multi-spectrometer will collect the spectral reflectance characteristics of the soil sample in 18 bands, where the bands cover the range from visible light to near infrared. In order to ensure the stability and representativeness of the data, the spectral reflectance of the same collection area is measured multiple times, and the average value is calculated to construct a spectral data set. In this way, the subtle differences of soil samples in different spectral bands can be captured.

[0050] In the data processing stage, in order to more comprehensively reflect the potential information in the spectral data, the ratio and product between the features of adjacent spectral bands are calculated according to the formula, and the columns are spliced ​​to combine with the original spectral features to form a complete spectral feature set containing 18 original spectral features, 17 spectral ratio features and 17 spectral product features. The calculation formula is:

[0051]

[0052] Multiplication=R i+1 ×R i , i=0,1,2,3,…,16;

[0053] Among them, R i With R i+1 Represents the reflectivity of the i-th or i+1-th band, Ratio i and Multiplication i They represent the spectral ratio characteristics and spectral product characteristics of the i-th band respectively.

[0054] S103, determining the nutrients in the soil sample by standardized chemical analysis methods.

[0055] In this embodiment, NY / T 1121.24-2012 was used to determine the total nitrogen content in the soil sample, LY / T 1232-2015 was used to determine the total phosphorus content in the soil sample, LY / T 1234-2015 was used to determine the total potassium content in the soil sample, NY / T 1121.6-2006 was used to determine the organic matter content in the soil sample, LY / T 1228-2015 was used to determine the alkaline-hydrolyzed nitrogen content in the soil sample, NY / T1121.7-2014 was used to determine the available phosphorus content in the soil sample, and NY / T 889-2004 was used to determine the contents of available potassium and slow-release potassium in the soil sample.

[0056] S104, matching the measured result with the spectral data set, and dividing the data set into a training set and a test set by a random division method.

[0057] In this embodiment, the ratio of the training set to the test set is 7:3.

[0058] S105, training the prediction model with the training set, and then evaluating the generalization ability of the prediction model with the test set, and the prediction model after training and testing is used to predict soil nutrients according to the spectral feature information obtained by the multi-spectrometer.

[0059] In this embodiment, linear regression analysis is used, and the correlation between the spectral characteristics and the soil nutrient content is calculated by the least squares method, wherein the calculation formula is:

[0060] y=β 0 +β 1 x 1 +β 2 x 2 +…+β n x n , n = 52,

[0061] Among them, β n represents the coefficient of the nth feature, x n represents the nth feature (1-18 are original spectral features, 19-36 are spectral ratio features, and 37-52 are spectral product features), and y is the predicted soil nutrient content.

[0062] Through model training, the relationship between spectral features and soil nutrient content can be quantified, thereby establishing a reliable prediction model. The performance of the model is verified using a test set. By calculating the coefficient of determination of the model prediction results, the prediction accuracy and robustness of the model are evaluated, thereby verifying its application value in soil nutrient detection.

[0063] In the embodiments of the present application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can mean that A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0064] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0065] The above is only a specific implementation of the present application. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A portable soil nutrient detection method based on multispectral, characterized in that: include: Obtain soil samples of a preset volume and pre-treat them; Using a multi-spectrometer to collect and process spectral characteristic information of the pre-treated soil sample to form a spectral data set; Determining the nutrients in the soil sample by standardized chemical analysis methods; Matching the measured results with the spectral data set, and dividing the data set into a training set and a test set by a random division method; The prediction model is trained by the training set, and then the generalization ability of the prediction model is evaluated by the test set. The prediction model after training and testing is used to predict soil nutrients according to the spectral feature information obtained by the multi-spectrometer.

2. The portable soil nutrient detection method based on multispectrum according to claim 1 is characterized in that: The method of obtaining a soil sample of a preset volume and performing pre-processing comprises: Along the preset route, collect mixed soil samples of the tillage layer according to the principles of random, equal amount and multi-point mixing; The collected mixed soil sample is placed in a grinder and fully ground; Screening the ground mixed soil sample through a sieve; The screened mixed soil sample is pressed into soil tablets using a tablet press.

3. The portable soil nutrient detection method based on multispectrum according to claim 2 is characterized in that: The method of collecting mixed soil samples of the tillage layer along the preset route according to the principle of random, equal amount and multi-point mixing includes: In case of small terrain changes, uniform soil fertility, and small sampling unit area, plum blossom-shaped sampling is adopted, and each sample requires 15-20 sampling points to collect soil and mix it evenly; The sampling points should be evenly distributed, and the entire sampling area should be controlled as a whole to avoid sampling in places where fertilizers have been piled, ridges, ditches, and special terrain areas; The soil sampling depth and sampling volume for each sampling point must be uniform and consistent, and the ratio of the upper and lower layers of the soil sample must be the same; If the amount of mixed soil samples collected is too large, you can use a shovel to repeatedly turn it upside down into a pile, then divide it into four parts in a "cross" from the middle from top to bottom to discard the excess soil.

4. The portable soil nutrient detection method based on multispectral according to claim 1 is characterized in that: The method of collecting and processing the spectral characteristic information of the pretreated soil sample using a multi-spectrometer to form a spectral data set includes: A portable multi-spectrometer is placed directly above the soil pressed sheet to collect spectral reflectance characteristics of the soil pressed sheet in different bands, where the bands cover a range from visible light to near infrared; The spectral reflectance is measured multiple times for the same acquisition area and the average value is calculated; The ratio and product between the features of adjacent spectral bands are calculated and combined with the original spectral features by column splicing.

5. The portable soil nutrient detection method based on multispectral according to claim 4 is characterized in that: The calculation formula for calculating the ratio and product between the characteristics of adjacent bands of the spectrum includes: Multiplication=R i+1 ×R i ,i=0,1,2,3,…,16; Among them, R i With R i+1 Represents the reflectivity of the i-th or i+1-th band, Ratio i and Multiplication i They represent the spectral ratio characteristics and spectral product characteristics of the i-th band respectively.

6. The portable soil nutrient detection method based on multispectrum according to claim 1 is characterized in that: The nutrients in the soil sample are determined by a standardized chemical analysis method, including: The total nitrogen content in soil samples was determined using NY / T 1121.24-2012NY / T 1121.24-2012; The total phosphorus content in soil samples was determined using LY / T 1232-2015; The total potassium content in soil samples was determined using LY / T 1234-2015; NY / T 1121.6-2006 was used to determine the organic matter content in soil samples; LY / T 1228-2015 was used to determine the alkaline nitrogen content in soil samples; The available phosphorus content in soil samples was determined using NY / T 1121.7-2014; NY / T 889-2004 was used to determine the contents of fast-acting potassium and slow-acting potassium in soil samples.

7. The portable soil nutrient detection method based on multispectrum according to claim 1 is characterized in that: The prediction model completed by the training test is used to predict soil nutrients according to the spectral feature information obtained by the multi-spectrometer, including: using linear regression analysis to calculate the correlation between the spectral features and the soil nutrient content by the least squares method, and the calculation formula is: y=β0+β1x1+β2x2+…+β n x n ,n=52, Among them, β n represents the coefficient of the nth feature, x n represents the nth feature, and y is the predicted soil nutrient content.

Images